Which Ethical Careers Make a Difference?: The Replaceability Issue in the Ethics of Career Choice

Which ethical careers make a difference? The Replaceability Issue in the Ethics of Career Choice Benjamin Todd Philosophy Thesis for the Masters Degree in Physics and Philosophy University of Oxford 2012 Contact: ben at 80000hours.org Contents 1. Introduction ........................................................................................................................ 3 Part 1 – The Consequences of Career Decisions ................................................................. 8 2. The Naïve View .................................................................................................................. 8 2.1. Introduction .................................................................................................................. 8 2.2. The Counterfactual View.......................................................................................... 10 2.3. The Relevance of the Counterfactual View ............................................................ 12 3. The Simple Analysis of Replaceability .......................................................................... 15 3.1. The Replaceability Ratio ........................................................................................... 15 3.2 The Simple Analysis of Replaceability .................................................................... 15 3.3. Consequences of the Simple Analysis .................................................................... 16 3.4. Simple Analysis Conclusions ................................................................................... 21 4. The Sophisticated Analysis of Replaceability ............................................................... 22 4.1 Problems with the Simple Analysis ......................................................................... 22 4.2. An Example Chain .................................................................................................... 23 4.3. Characterising the Chains ........................................................................................ 25 4.4. The Value Produced by Jobs .................................................................................... 25 4.5. The Leading Order Approximation ........................................................................ 27 4.6. Isolated Industry ........................................................................................................ 29 4.7. Case Study: Taking a Job at a Charity .................................................................... 30 4.8. Case Study: Does Becoming a Banker Have Negative C-Consequences? ......... 40 4.9. Conclusion .................................................................................................................. 43 Part 2 – The Dominance Thesis........................................................................................... 46 5. Other Consequences......................................................................................................... 46 5.1. Introduction ................................................................................................................ 46 1 5.2. Compensation Effects ............................................................................................... 47 5.3. Signalling Effects and Boycotts................................................................................ 50 5.4. Effects on the agent’s feelings and personality ..................................................... 59 5.5. Supply-Demand Effects ............................................................................................ 62 5.6. Conclusion .................................................................................................................. 64 6. Non-Consequentialist Considerations........................................................................... 66 6.1. Introduction ................................................................................................................ 66 6.2. Side Constraints ......................................................................................................... 67 6.3. Virtue ........................................................................................................................... 78 6.4. Integrity....................................................................................................................... 80 6.5. A Real Example .......................................................................................................... 83 6.6. Conclusion .................................................................................................................. 83 7. Analogous Arguments ..................................................................................................... 85 7.1. Introduction ................................................................................................................ 85 7.2. Replaceability in Vegetarianism .............................................................................. 85 7.3. Ethical Investing ........................................................................................................ 87 7.4. Conclusion .................................................................................................................. 93 8. Conclusion ......................................................................................................................... 95 Bibliography .......................................................................................................................... 98 2 1. Introduction It’s often thought that certain careers, such as working for a charity or becoming a surgeon, are ethical because they ‘make a difference.’ Suppose, for example, that a surgeon performs 100 lifesaving operations per year. Intuitively the difference she makes is that she saves 100 lives per year. But there’s a counterargument to this claim. If you don’t take these jobs, someone else will. Let’s introduce a technical term: Replaceable If there’s a job that someone else would take if you don’t take it, then in that job, you’re replaceable. If the person who would replace you would do the job just as well as you would have done, then at a first glance you make no difference at all. The surgeon’s lifesaving operations will still be performed. How, then, are we to understand the consequences of taking a career in which one is replaceable? Almost 200,000 graduates1 enter the workforce each year in the UK, and a majority take ethical considerations into account in choosing their career2. For these people whose career choice is ethically motivated, this question is deeply important. One can have a huge direct impact through many careers. A heart surgeon might be able to perform more than one hundred lifesaving operations per year. A fundraiser who raises $1m for the Against Malaria Foundation would enable 200,000 people to (HESA, 2011) In a survey by The Guardian, over 70% of students said that ethical considerations were crucial in choosing an employer. See (Robinson, 2006) 1 2 3 receive mosquito nets. As I explore in the first part of this thesis, how one analyses replaceability could therefore mean the difference between saving hundreds of lives and saving none. In the second part, I argue that making a difference is not only morally relevant, but that the job-related consequences of career decisions determine which career choice is most ethically preferable. I call this the Dominance Thesis: Dominance Thesis For typical ethically motivated career decisions, the job-related consequences of the decision typically determines which choice is ethically preferable. I define: Job-related consequences The consequences that result from the activities of people in jobs, which they perform as part of the job. Since replaceability determines the job-related consequences of careers, it determines which choice is ethically preferable, so is a crucial consideration for those who are pursing ethical careers. One might find this surprising. Replaceability in career choice was first introduced by Bernard Williams in Utilitarianism: For and Against3 precisely to show that not only the consequences of career decisions are morally relevant. In Williams example, 3 (Williams & Smart, 1973) pp97 4 George is offered a job at a chemical weapons factory. He is told that if he doesn t take the job, a more zealous person will. In that job, George is replaceable. Williams believes that George has reason to not take the job, even if taking the job would have positive consequences (since George would do less harm than the zealous man). Williams gives the second example of Jim and the Indians. Jim is told that if he kills one Indian in a group of twenty, the other nineteen will be let free. If he doesn t kill one Indian, Pedro will kill all twenty. So, for whichever Indian Jim kills, if he doesn t kill them, someone else will. Jim s act of killing the Indian is replaceable. This shows that actions of all kinds can be replaceable, raising similar dilemmas. Replaceable Actions Of an action, if it’s the case that an agent doesn’t perform that action, someone else will, then that action is replaceable for that agent. Williams is primarily concerned with whether consequentialism can explain the intuition that Jim has reason to not kill the single Indian, even though it leads to much better consequences. This is not the concern of this thesis. I m concerned with the normative question of what we should do in typical career decisions. Even if Williams is correct that George shouldn t take the job at the chemical weapons factory, I argue that in typical cases of ethically motivated career choice, consequences determine ethical preferability, making replaceability a crucial consideration. 5 Despite its importance, replaceability is ignored in what s called ethical careers advice by career choice guides and university careers services. This is shame. Each year, thousands of people enter the workforce wanting to make a difference but they are not given the advice they need. You can t know that you re making a difference without the relevant facts about replaceability. The ethics of career choice is also neglected in practical ethics.4 This is also shame. We’ll spend over 80,000 hours in our careers, and career decisions are some of the most important we make. Practical ethics should be able to offer guidance in these decisions. Beyond Williams, who was not concerned with the normative question of which career to take, there’s only one paper which mentions replaceability in career choice5. This paper, by Will Crouch6, is not directly about replaceability, but rather makes an application of it. For this reason, I simultaneously present the first systematic exploration of the ethical issues raised by replaceability in career choice. Outline In the first part, I explore how we should think about the morally relevant consequences of career decisions in the light of replaceability. In §2, I introduce the Naïve View, which states that what’s morally relevant about the consequences of one’s actions is what one does directly. This means ignoring replaceability. I reject this view, arguing that what would have happened if one hadn’t taken the career is The existing recent literature focuses on the question of whether it s permissible to take a non-altruistic career e.g. see (Buss, 2006) and (Care, 1984), as well as small debate about whether many professionals should retire immediately e.g. (Lenman, 2007) 5 A similar point is also made in print in (Tomasik, 2007) 6 (Crouch, Unpublished) 4 6 also morally relevant – the Counterfactual View. In §3, I introduce a Simple Analysis of the relevant counterfactual, which states that the consequences of a career decision depend on the difference between what one does in a career and what one’s replacement would have done. This is the view adopted by Williams and used by Crouch, but it turns out to be too simple. In §4, I introduce for the first time a Sophisticated Analysis of the counterfactual. I explore what this analysis means for the consequences of typical career decisions. In the second part, I defend the moral relevance of replaceability, understood as modifying the consequences of career decisions according to the Sophisticated Analysis. In particular, I argue for the Dominance Thesis. In §5, I consider the other consequences of career decisions, such as signalling effects. In §6, I consider nonconsequentialist considerations, in particular Side Constraints, Virtue and Integrity. I show that none of these considerations gives us reason to ignore replaceability in general. In §7, I consider the general objection that taking account of replaceability in career choice is analogous to making the bad argument that ‘if I don’t buy this meat, someone else will, so it’s permissible to buy it.’ I show that it’s not analogous. Moreover, in cases where it is analogous, like ethical investing, it shows that we should rethink our views about these cases. This means replaceability should be a crucial consideration in all ethically motivated career choices. 7 Part 1 – The Consequences of Career Decisions 2. The Naïve View 2.1. Introduction Let’s consider an example career decision. Sarah the Surgeon Sarah, during med school, decides to become a surgeon. She successfully applies to the surgery speciality, succeeding against several other applicants. Over the course of her career, she goes on to perform 100 lifesaving operations per year. Intuitively, one might think that the consequences of Sarah becoming a surgeon are saving 100 lives per year. But if Sarah didn’t become a surgeon, suppose one of the other applicants would have been offered the place. So Sarah is replaceable. Sarah’s replacement would still perform many of the operations that Sarah performs. One might ignore this fact, because one believes that what s morally relevant is what Sarah does directly as the result of her decision. I call this the Naïve View of consequences: Naïve View The morally relevant aspect of the consequences of taking a career is what one does directly in that career. There are two components to the meaning of doing directly : 8 1. Non-Counterfactual Cause: The agent is part of the causal chain that leads to the result, though the result might still happen if she didn t act7 2. Proximity: The agent is reasonably close in the causal chain to the result and her level of responsibility diminishes with distance from the result Sarah is both part of the causal chain the leads to the 100 saved lives, and since she performs the operations with her own hands, is close to the end result. So, on this view the consequences of her actions include saving these lives. A stricter interpretation requires that the agent is the final agent in the causal chain. The surgeon is indeed the final agent, but this interpretation is too strict. The bankers who contributed to the suffering of the financial crisis (supposing there were some) only did so through a complex causal chain. They were not the final agents in causing the suffering, but it was a consequence of their decisions. A more plausible and common position is that relevance diminishes with distance from the result. No-one would pretend that someone who fundraises to pay for more nurses is not doing any good. Nevertheless, it does seem to be considered especially praiseworthy to be the nurse instead of the fundraiser – directly helping people with one s own hands. Note this condition might not be compatible with certain counterfactual accounts of causation. These accounts, e.g. as developed in (Lewis, 1973), require that for c to cause e, if c doesn t occur, then e would not occur. 7 9 The Naïve View seems to be implicit in much ethical careers advice, which focuses on jobs that directly do good and ignores replaceability. Paradigm ethical careers are those in the 3rd sector8, in which one is directly involved with the activities of altruistic non-profit organisations. They commonly extend to certain public sector jobs, like social work and being a state school teacher, which involve directly helping disadvantaged people. There s little discussion of how one might do good indirectly, for instance through philanthropy or by saving the time of someone who is doing a lot of good. Completing Norman Borlaug s9 chores would not normally be considered an ethical career, but speeding up his research by just a week could have meant that thousands of additional lives were saved by his short-stem diseaseresistant wheat. 2.2. The Counterfactual View Despite being implicit in much ethical careers advice, the Naïve View isn t a good reflection of what we normally mean by the consequences of a decision. Rather, consequences consist in making a difference. Making a difference means bringing about states of affairs that wouldn t have happened otherwise. Accepting this means taking a Counterfactual View of consequences: Counterfactual View The morally relevant aspect of the consequences of taking a career is the This is clearly the case, for example, in (Burrows, 2006) and in the many on-line ethical careers websites. 9 Sometimes credited with saving one billion lives (Easterbrook, 1997) 8 10 difference between what happened because of you and what would have happened otherwise. The consequences that are morally relevant according to the Counterfactual View are therefore the counterfactual consequences: C-Consequences The c-consequences (counterfactual consequences) of an action are the differences between what happened because of that action and what would have happened otherwise. In Lewis possible worlds semantics, one could understand what would have happened otherwise as what happens in the closest possible in which one doesn t take the job. Adopting the Counterfactual View means rejecting both parts of the Naïve View. Despite this, ‘to make a difference’ is one of the most commonly cited reasons to take an ethical career! So, both views exist in tension in ethical careers advice. I think this is in part genuine inconsistency. Counterfactuals are often ignored. For instance, Nassim Taleb10 has argued that we also fail to take account of how successful people are often not successful in close counterfactual situations when assessing their success, in particular due to Hindsight Bias. I think it’s also the result of a false belief that careers which do a lot good directly are also those that make a difference. I show this is often not the case in §4. 10 (Taleb, 2007) 11 Whatever one believes about the meaning of ‘consequence’, the Counterfactual View needs some weight i.e. the c-consequences of decisions have at least some moral relevance. To see that, consider: Paul and the Paramedics Paul is involved in a car crash. When he regains consciousness, he finds his mother lying next to him, seriously injured. Paramedics have just arrived. Paul pushes the paramedics out of the way, and performs life-saving first aid on his mother. Unfortunately, he’s much less skilled than the paramedics, so causes her permanent spine damage, resulting in paralysis. If Paul hadn’t acted, the paramedics would have performed the lifesaving first aid, and his mother would have fully recovered. Paul s direct action is to save his mother s life. So, according to the Naïve View, the consequence of his action is to save his mother s life, so his action was morally commendable. But it s clearly natural to say that the consequence of Paul pushing the paramedics out of the way is to paralyse his mother. Moreover, this cconsequence is highly morally relevant, and explains why his action was wrong. So, the Counterfactual View must be assigned some weight. 2.3. The Relevance of the Counterfactual View Accepting that c-consequences are morally relevant means that replaceability is morally relevant. Indeed, we can image analogous career decision cases. If Paul takes 12 a job instead of someone who would do much more good in that job, then it also has negative c-consequences. Nevertheless, this example doesn t show how and which c-consequences are morally relevant. One might believe that factors over and above c-consequences are important. For instance, some actions might be wrong in themselves, irrespective of consequences. One might also retain Proximity, i.e. apply greater moral weight to consequences that are caused by the agent and result from a shorter causal chain. This implies partially neglecting replaceability. Proximity is intuitive in that we often act as if indirect ways of bringing about positive c-consequences are less praiseworthy than direct ways. Even if a fundraiser raises enough money to employ nurses to hand out medicine that save 100 lives, we would often find the nurse who saves 100 lives more praiseworthy (assuming both wouldn t have happened otherwise . This could be due to other factors. Perhaps being the nurse is considered more demanding, even if not ethically preferable. Praising the nurse might instead simply be inconsistent with our beliefs about ethical preferability. Overweighting the relevance direct actions might be a useful heuristic. In general it s easier to attribute causality in these cases, thus these consequences are usually more certain. This causes us to mistakenly apply higher weight to close consequences, even when we know the consequences precisely. Fully considered, 13 however, what s important is that the lives get saved. Both the fundraiser and the nurse are necessary to save the lives, so both are equally praiseworthy. Indeed, as we would expect if this were the case, judgements of Proximity often have an air of arbitrariness. An effort to save lives with a malaria vaccine requires everyone from people who distribute the vaccine to those who discovered it, to those who campaigned for research funding. It s odd to say the campaigners are less praiseworthy than the researchers because they are one further step removed in the causal chain from the saved lives. It would be even odder to encourage the researchers to become nurses in order to do good more directly. When it comes to harmful careers, Proximity matters because if one directly causes harm, that action could be prohibited by a side constraint or it could compromise one s integrity. So, although it s implausible that Proximity matters in itself, it often matters due to non-consequentialist considerations. These are covered in §6. Supposing we apply weight to the Counterfactual View, we still need to ask, what are the c-consequences of typical career decisions? In the next section I introduce the Simple Analysis of the counterfactuals, as employed by Williams and Crouch. In §4 I show that a more complex analysis is needed. 14 3. The Simple Analysis of Replaceability 3.1. The Replaceability Ratio What would happen if someone doesn’t take a job depends on the details of the situation. So, the impact of replaceability varies between choices as well as between different analyses. To characterise different answers, I introduce the Replaceability Ratio: Replaceability Ratio, R The ratio between the value of the job-related c-consequences of taking a job, taking into account the fact that one is replaceable, and the positive value one produces directly in that job. ‘Value produced directly’ is intended to be understood as all the good one causes in the job as a result of the activities of the job, whether or not it would have happened otherwise. So if Sarah were not replaceable, and none of those operations would have been performed, R=1. If all of the operations would have been successfully performed anyway, R=0. Adopting the Naïve View, ignoring Proximity, means always taking R=1. 3.2 The Simple Analysis of Replaceability If Sarah doesn’t become a surgeon, someone else would: her replacement. The Simple Analysis claims: Simple Analysis The c-consequences of taking a job consist of the difference between what 15 happens if one takes the job and what would have happened if one’s replacement had taken the job. If Sarah had chosen not to become a surgeon, then someone else would have filled that same job. Suppose that person only performed 90 successful operations per year, then, R=0.1. Williams seems to take this view of replaceability. He says that if George takes the job at the chemical weapons factory, it has positive expected c-consequences. This is because if he doesn’t take the job, it will be filled by someone who will pursue chemical weapons research far more zealously, leading to greater expected harm. Crouch also makes use of this analysis, although he acknowledges in a footnote that replacement can be more complex.11 3.3. Consequences of the Simple Analysis The Simple Analysis is only correct in limited circumstances. Nevertheless, it implies some surprising conclusions about the consequences of typical career decisions, which can still stand on the Sophisticated Analysis. 3.3.1. Doing Less Good Than the Naïve View For most jobs, there’s significantly more than one applicant per place. So, if you don’t take the job, someone else will. In particular, it will be the person who finished 11 (Crouch, Unpublished) §2 16 second in the selection process12. If we assume that on average the selection process sorts people into order of ability and it’s not a field in which there are huge differences in ability, we can expect this person to just be slightly worse at the job than you. So the difference you make is significantly less than what it looks like you’re doing directly. Sarah is probably an example of this kind. One would expect hospitals to be able to differentiate good surgeons from bad ones to some degree. So, ex-ante one expects the second best applicant to be almost as good as Sarah (though this depends on empirical facts about typical differences in ability between surgeons). So, R would be positive but small. 3.3.2. Doing less harm than the Naïve View Similarly, in many harmful jobs you’ll be doing less harm than it looks like you’re doing directly. This is due to exactly the same reasoning: if the selection process works, the person who would replace you would be slightly worse at the job, and so produce slightly less harm than you would have produced. Consider a typical unethical career, like working for an arms company selling contracts to developing world governments. The aim of the selection process will be to choose the person who can sell the most arms for the company. So, it’s likely that the person who just fails to get the job would sell fewer arms, thus producing less harm than the person who gets the job. I m assuming that this counterfactual is not under-specified, as explored in (Hare, 2011), or unknowable in a problematic way, as explored in (Lenman J. , 2000) 12 17 3.3.3. Doing good by taking harmful jobs In some cases, it might do good to take harmful jobs, because you do less harm than the person who would have replaced you. This could be because doing the job well actually makes it less harmful. Examples would be careers that are on average harmful because they require a lot of skill, but if you have enough skill, they’re not harmful. For instance, certain types of stock market investing jobs might produce negative value, because most investors follow the crowd and don’t base their views on economic fundamentals. This makes markets more volatile and less efficient. Skilled investors, however, go against the crowd, providing capital to neglected companies, making markets more efficient. Or perhaps it’s because you’re ethically minded, so you seek to avoid the most harmful activities carried out by people typically in your shoes. For example, some arms company employees bribe foreign government officials to encourage them to buy more arms13. It seems plausible that you could take a top job in an arms company, but veto that practice. Perhaps that would avoid most of the harm normally done by someone in your shoes. In these cases, one takes a job that is harmful but has positive c-consequences, so R is negative. The size of R depends on one’s scope to avoid harmful activities. If one makes a harmful job less bad, -1≤R<0. If one turns a harmful job into a neutral one, R=0. If one turns a harmful job into a value producing job, R<-1. For instance, a BAE Systems executive was recently fined for corruption, see (Evans & Leigh, 2010) 13 18 3.3.4. Reducing the good done by taking good jobs The selection process is not perfect. If you secure a job instead of someone who would be better than you, then your decision has negative c-consequences, although it looks like you’re doing good directly. This means you’re in a situation like Paul and the Paramedics. R is negative, but closer to zero than to -1. Being likable and attractive can help you to succeed at interview. It’s plausible that this often doesn’t correlate equally well with job performance. Similarly, personal connections might give some people an advantage in the selection process unrelated to their ability. If you’re an attractive, likable, connected person, then you might need to be particularly alert to the possibility of getting a job over someone who would perform better than you. If you do, then you’re reducing the good that would have been produced. The more important the job, the worse the effect. Those taking political careers should be particularly alert to this danger. 3.3.5. Doing more good than you appear to do It’s possible for R>1 if your potential replacement would have been destructive, turning a good job into a harmful job. For instance, suppose accountancy normally produces positive value, but in the hands of a corrupt person is very harmful. If your potential replacement would have been one of these people, then by taking the job, the positive c-consequences are greater than what it looks like you’re doing directly. 19 3.3.6. Supporting the arguments in favour of professional philanthropy These conclusions, using the Simple View, are used to support Crouch’s14 arguments in favour of pursuing professional philanthropy over work in the 3rd sector. Professional Philanthropy means pursuing a high earning career with the intention of making large philanthropic contributions to highly cost-effective causes. Crouch first advances the Weak Thesis: The Weak Thesis Typically, it’s ethically preferable to pursue professional philanthropy in a morally innocuous career than to pursue work in the 3rd sector Morally innocuous careers are such that there is no strong non-consequentialist reason not to pursue that career. To support this thesis, Crouch argues that (i) a high earner could typically pay for several 3rd sector workers, having several times as much impact (ii) a high earner can donate to the most cost-effective charities, which are often 10x more effective than the median, whereas a 3rd sector worker will be more restricted in where they work. Either of these arguments is strong enough to demonstrate the Weak Thesis. But in addition, one can argue that by becoming a 3rd sector worker the c-consequences are less than what one appears to do directly. 3rd sector work is typically very competitive, so if you don’t take the job, you would be replaced. Your replacement is likely to be almost as good as you would have been. Replaceability is also used to support Crouch’s Strong Thesis: 14 (Crouch, Unpublished) 20 The Strong Thesis Typically, it’s ethically preferable to pursue professional philanthropy in a moderately morally controversial career than to pursue work in the 3rd sector Morally controversial careers are such that there is a strong non-consequentialist reason not to pursue them. Since morally controversial careers are often very well paid, they are also competitive. So, your replacement would be almost as good as you, causing almost as much harm as you would have done. So, the negative cconsequences of taking morally controversial careers are much smaller than the harm one does directly. Moreover, one can argue that if ethically minded people take morally controversial careers, the c-consequences are often positive. If the job offers no scope to make more ethical decisions, then one’s replacement will be equally harmful. If the job does offer scope to make more ethical decisions, then an ethically minded person can take these opportunities. We would expect these opportunities not to be taken by the average replacement, who’s less ethically motivated. So, the c-consequences are positive. 3.4. Simple Analysis Conclusions The Simple Analysis suggests that the c-consequences of typical career decisions diverge significantly from the good one does according to the Naïve View. R can take any value, though is typically in the range -1 to 1. Since these c-consequences are morally relevant, it’s important to take account of replaceability in typical career decisions. 21 4. The Sophisticated Analysis of Replaceability 4.1 Problems with the Simple Analysis One might feel that one shouldn’t act on the conclusions of the Simple Analysis. This is normally due to the factors addressed in Part 2. Another reason, however, is that the Simple Analysis seems too simple. It doesn’t take account of the complexity of the economy. This is true. In the closest possible world in which Sarah doesn’t become a surgeon, someone else becomes a surgeon: her replacement. But then someone else could take the job Sarah’s replacement would have had; her replacement is also replaced. This could happen again and again. So, in general, the world in which Sarah doesn’t become a surgeon could differ by an entire chain of replacements. I call this the Iteration Effect. Iteration Effect The closest possible world in which one doesn’t take a job will in general differ by an entire chain of replacements. This means the Simple Analysis isn’t a complete assessment of what would have happened otherwise, so it doesn’t tell us full the consequences according to the Counterfactual View.15 In fact, the Simple Analysis can give results that are arbitrarily wrong. For instance, if George takes the job at the chemical weapons factory, he prevents the zealous man from taking the job. But if the zealous man goes on to take an even more harmful job at another factory, then George has had a negative effect. Williams neglects to note this additional complication, since it doesn’t affect his basic point. For someone 15 Although it might be approximately correct for someone who accepts Proximity (§2.1). 22 interested in the ethics of career choice, however, this is another c-consequence of most career decisions. Crouch also neglects the Iteration Effect. I will show that replaceability only does the work Crouch wants it to do (§3.3.6) given certain empirical assumptions, which plausibly don’t hold in many situations. I will do this by introducing, for the first time, a way to analyse the consequences of typical career decisions in the light of the Iteration Effect. 4.2. An Example Chain If Sarah at medical school takes up the offer to enter the surgery speciality, what would happen? First, there were probably other applicants who met the minimum criteria to become surgeons, but can’t get places. If Sarah takes the offer, then a different person, call them Tiff, can’t become a surgeon. Tiff goes on to take another job, probably in another slightly less competitive speciality, say General Practice. Suppose that Tiff prevents Ursula from taking up her desired GP path. Ursula, who only wanted to work as a doctor if she could be a GP, instead enters medical research. She ends up researching malaria vaccines. Vino, who studied biology, and has her heart set on researching malaria, is unable to take the position Ursula takes. Instead, she takes up a less prestigious research position. She’s the only qualified applicant. 23 Figure 1. An example downwards chain for Sarah World where Sarah doesn’t take offer World where Sarah does take offer Sarah Some other job Sarah Tiff Surgeon Surgeon Ursula GP Tiff GP Vino Researcher#1 Ursula Researcher#1 Vino Researcher#2 24 4.3. Characterising the Chains If someone enters a job, it causes a downwards chain of replacements. The person who would have taken the first job has to take their second preference job. The person who would have taken that job has to take their second preference job. And so on. If someone leaves a job, it causes an upwards chain of replacement. In these, everyone below the first job is able to take a job that they prefer relative to their current job. Replacement chains can end in two ways: 1. Someone enters a no-competition job. This is job that wouldn t exist if it were not for the person who actually fills it. In practice, these are jobs that are created especially for someone or are jobs for which only one qualified applicant can be identified. 2. Someone becomes unemployed, which produces job-related value of zero It shouldn’t be thought that in general replacement chains follow a promotional hierarchy or even jobs in the same industry. In reality, they might involve all kinds of jobs that are completely unrelated, except that the person in the above box would move into them if they lost their current job. 4.4. The Value Produced by Jobs A person in a job produces a certain amount of moral value through the activities of that job, whatever is of value according to one’s meta-ethical and ethical position. For 25 instance, if Sarah becomes a surgeon, the value she produces consists of the value of the 100 operations she performs per year. Notice that the value produced by a job might be completely unrelated to the salary of that job and that different people will disagree about the value of different jobs. For someone who cares primarily about spreading Christianity, priests presumably do work of tremendous value, despite their low pay. For hard-line atheists, these jobs might even be harmful. To calculate the c-consequences of Sarah’s decision in terms of value produced, one needs to estimate the value of each person in each job for both chains and calculate the difference. It can be useful to break the value produced by someone in a job into two terms. One of these terms depends on the job. I call this the opportunity* (the star denotes a technical use of the word): Opportunity* The opportunity* of a job is the job’s potential to produce value. A job’s opportunity* consists of the resources one is given control of, one’s platform to influence others in the job, the number of people one manages and so on. The opportunity* for a surgeon to produce value is determined by the number of surgeries they are able to perform, the facilities they are able to use, the skills of the supporting staff, their options to influence the running of the hospital etc. The second term depends primarily on the person. I call this ability*. 26 Ability* A person’s ability* is their ability to use the opportunity* of a job to produce value. A person’s ability* depends on their skills and their aims. Sarah’s ability* consists of how good she is at surgery, as well as her abilities to inspire her colleagues to be more productive, how much she contributes to the general running of the hospital and so on. Opportunity* is fixed by the job. Ability* is defined so that the product of opportunity* and ability* gives the value produced by that person in that job. In what follows, I’ll apply these ideas to suggest how to estimate the c-consequences of certain career decisions. The Iteration Effect makes it difficult to estimate cconsequences, especially for specific decisions. Nevertheless, it is possible to make good estimates in certain situations, especially if one considers expected cconsequences, which rest on averages. 4.5. The Leading Order Approximation In some situations, one only needs to consider the first couple of steps in a chain. In the limit of considering only the first job, we recover the Simple Analysis. These situations occur when the scope for impact in the first couple of jobs is much higher than the expected impact of the other jobs open to people in the chain. This means we can treat the value produced by people in the other jobs as approximately zero. 27 One example would be becoming an MP in UK. To be elected into office, one must be selected by a party to stand in a constituency. People often seek to enter politics from other careers, like law and journalism. If they are not selected by their party of choice, they might give up their political aims and continue working as lawyers. The impact of the average lawyer is much less than the potential impact of an MP. So, if one wants to calculate the value of becoming an MP, one can sometimes neglect the value produced by one’s nearest rival in their second choice job, and all subsequent steps in the chain. The same could be true when considering the actions of someone with especially large ability*. For instance, some researchers produce work of many orders of magnitude more impact16 than the median researcher in their field. It may be possible just to focus on the chain insofar as it involves these super talented people, approximating all other steps as zero. Cases of high ability* also occur when someone has much more focus on producing value than other people in their field. Owain the Researcher There is a limited amount of funding for artificial intelligence (AI) research from academic institutions. Suppose that the majority of researchers in the field focus on developing AI without regard to the risk of dangerous AI. Number citations is one proxy for impact, and there s evidence that citations approximately follow a power law distribution, e.g. see (Gupta, Campanha, & Pesce, 2005) 16 28 Owain believes that dangerous AI could be extremely dangerous. He takes a job in academia researching artificial intelligence, but unlike his colleagues, he uses the funding to reduce the risk of dangerous AI. If we share Owain’s assumptions about AI, then he is producing far more of value than anyone likely to have the research funding in his place, and anyone in the downward chain of replacements he causes. Indeed, these people might even be producing negative value. 4.6. Isolated Industry Often one only wants to consider one’s impact on a single industry. In this case, the chain can be ignored as soon as it leaves the industry. Isolated industry often arises as an approximation. If one is ethically motivated, then it’s likely that one is considering industries that produce far more value than average. In these cases, one can make the approximation that anyone who drops out of the industry will, on average, enter a job that produces approximately zero value. If one also assumes that if someone leaves the industry, then the chain of replacements doesn’t return to the industry, one can take the chain as ending as soon as someone leaves the industry. If one considers isolated industries, it’s often much easier to make an estimate of the effect of replaceability. The following are some examples. 29 4.7. Case Study: Taking a Job at a Charity 4.7.1. Introduction Crouch claims that in taking a charity job, the c-consequences are typically much smaller than the good one appears to do directly i.e. R<<1. I investigate whether this is true in the light of the Iteration Effect, showing that it’s only true in some circumstances. This discussion also serves as an exploration of the c-consequences of typical career decisions. I’ll assume we’re not in a situation in which it’s valid to make the Leading Order approximation e.g. considering the actions of someone with extremely high ability*, or an extremely high opportunity* position. Consider the following assumption: Constant Ability* Someone’s ability* is constant across the job they currently occupy and the job they would move into if they weren’t able to take that job Constant ability* is clearly a large simplification. Although there are skills which find a wide application across many jobs, motivating the idea that some people are generally more skilled than others, people clearly have different levels of relative skill in different positions. Indeed, people might intentionally alter their relative skill. When George moves to the chemical weapons factory, he aims to reduce his ability, so as to minimise the negative value produced by the position. 30 Nevertheless, often people will move into jobs of a similar kind, so Constant Ability could be a good approximation. We could instead assume: Weak Constant Ability* On average, a person’s ability* is approximately the same in the job they currently occupy and the job they would move into if they lost that job. If what we care about is the expected change in value due to a career decision, then we can use assumptions based on averages. 4.7.2. Equal Ability* Let’s consider Ella, who is considering taking a charity job. She first considers positions that leave little scope for differences in skill; for instance, handing out fliers or basic administrative work. It seems reasonable to think that if someone can’t get an entry level administrative position at a charity, they will seek to fill a similar position at another charity, or they will leave the charity sector. I’ll assume that administrative charity jobs produce much more value than typical administrative jobs, so take the chain to end once someone leaves the sector. So, the chain will consist of similar jobs. We could model this chain by assuming that everyone in the chain has equal ability*. So, the value each person produces just depends on the position they occupy. It doesn’t depend on differences in skill, because there’s little scope for skill. In this model, if Ella takes a job causing a downwards chain, the effect is to shuffle people of equal ability* along the chain. If the resulting number of jobs in the charity sector is the same, then this has no c-consequences. So, R=0. 31 Figure 2. Shuffling people of the same ability* Ella Job1 Person1 Job2 Person2 Job3 Person3 If an extra job is created, the c-consequences are equal to the product of ability* and opportunity* for the created job. So, R could have any value. Suppose we also assume: Declining Opportunity* If one doesn’t get their preferred job, one moves into a job with lower or equal opportunity* This is often not true, because career decisions depend on more than value producing potential. For instance, consider someone who loses their comfortable but low-impact job, forcing them to move into a role of much higher responsibility to pay the bills 32 Nevertheless, it seems to be true that people generally seek the job with the highest responsibility available. So, there’s a tendency for people to move into jobs with lower opportunity* if displaced. So, the assumption we actually make is: Weak Declining Opportunity* On average, if unable to take their first preference job, one will move into another job with lower or approximately equal opportunity*. If we make this assumption, then we expect the opportunity of the created job at the end of the chain to be lower or approximately equal opportunity* to the job Ella takes. So, we expect 0<R≤ 1. Whether Ella’s application leads to job-creation depends on supply and demand for these jobs. As shown in §5.5, Ella probably causes about half of a job to be created, so the situation lies between job-creation and no job-creation. So, overall we expect 0<R≤1, with R closer to 0 than to 1. 4.7.3. Equal Opportunity* Now suppose Ella takes a job at a highly meritocratic environment, like a start-up charity. Also suppose the chain ends whenever someone leaves the organisation (perhaps it’s exploiting a particularly effective niche). In a meritocratic environment, one’s skill is the primary determinant of the resources one is given. So, we could model this situation as if everyone in the chain has approximately equal opportunity*. 33 If no job-creation occurs, Ella’s only effect is to change the level of abilities* in the chain. Specifically, Ella is added to the chain, and the person at the bottom is removed. So, the c-consequences of Ella taking the job are given by the difference between Ella’s ability* and the ability* of the person at the end of the chain, multiplied by the opportunity*. R could then take any value; but suppose we assume: Job Selection Efficiency One can only displace a person of lower ability*. Selection Efficiency is another large simplification. Someone’s ability to get a job could significantly diverge from their ability*. This could be because of biases in the selection process; but it’s also because the aim of most jobs is not directly to produce value. Rather, it’s to produce profit for the company, or some other aim which need not line up with producing what one believes is of value. Nevertheless, it seems like there’s some correlation between someone’s general ability in a job and their ability to produce value in that job. Further, there’s some correlation between someone’s ability to get a job and their ability in the job. So, it seems plausible Weak Selection Efficiency is approximately true. Weak Selection Efficiency On average, if one displaces someone, they’re of lower ability*. 34 Making this assumption, Ella’s ability is expected to be higher than the person at the bottom of the chain. So, 0≤R≤1. If Ella’s ability* is significantly higher, then R is close to one. If job-creation occurs, then Ella simply adds to the pool, and R=1. Averaging the two situations, R is probably close to 1. 35 Figure 3. If Ella doesn’t take the job, equal opportunity* chain Ella Person 1 Ability*=90 Job 1 Opportunity*=1 Person 2 Ability*=81 Job 2 Opportunity*=1 Person 3 Ability*=73 Job 3 Opportunity*=1 Value=90 Value=81 Value=73 Total Value=244 36 Figure 4. Ella does take the job, equal opportunity* chain Ella takes Job 1 Ability*=100 Job 1 Opportunity*=1 Person 1 Ability*=90 Job 2 Opportunity*=1 Person 2 Ability*=81 Job 3 Opportunity*=1 Value=100 Value=90 Person 3 Value=81 Value=0 Total Value=271 Difference in Value=27 R=0.27 37 4.7.4. Declining Opportunity* and Ability* Finally, suppose Ella joins a hierarchical chain, perhaps at a large and established charity. She, on average, displaces people of lower ability* than her into lower opportunity* jobs. This amounts to assuming both Weak Selection Efficiency and Weak Declining Opportunity*. Assuming there’s no job-creation: 1. Ella increases the ability* of the people in the chain, by knocking out the bottom person 2. People move down, so that lower opportunity* jobs become occupied by people of higher abilities* The total effect depends on (i) the length of the chain (ii) the gradient of the decline in ability* (iii) the gradient of the decline in opportunity*. The ‘gradient’ of the decline is the amount by which ability* or opportunity* decreases with each step down the chain. Here are some example inputs to provide a feel for the numbers. I assume ability* and opportunity* decline at equal gradients. 38 No Job-Creation Ability* and Opportunity* Gradients R when length=1 R at the limiting length17 0% 0 0 5% 0.05 0.51 10% 0.10 0.53 25% 0.25 0.57 50% 0.50 0.67 100% 1 1 For length=1, we recover the Simple Model. Including the rest of the chain adds a positive correction, because one shifts higher ability* people into the positions below them. For a gradient of 0%, we recover the equal ability* model. Job-creation Ability* and Opportunity* Gradients R when length=1 R at the limiting length 0% 1 1 5% 0.95 0.51 10% 0.91 0.53 25% 0.81 0.57 50% 0.75 0.67 100% 1 1 It converges because it s the difference between two converging geometric series. The rate of convergence depends on the gradient. For a gradient of 10%, it occurs to within 2 decimal places at length=39. For a gradient of 50%, it occurs at length=7. 17 39 Notice that the results are the same in the limit, because at the limit, the created job is adding no value. However, for length=1, job-creation means adding a significant positive correction to the Simple View. I’d hypothesise that in typical situations gradients are around 5-25%, though this is an empirical question. If this is true, then without job-creation, 0.05≤ R≤ 0.60. So, one makes a significant difference, but it’s significantly less than what one does directly. With job-creation, 0.51≤R≤0.95. So, one makes most of the difference it appears one does directly. In reality, the situation will be somewhere between job-creation and no job-creation, depending on supply-demand dynamics (§5.5). 4.8. Case Study: Does Becoming a Banker Have Negative C-Consequences? Crouch also claims that by entering harmful careers, one does much less harm than it appears one does directly. Let’s suppose that Ella is choosing between taking a charity job, and working in an area of banking where positions are typically value destroying (with negative opportunities*). Let’s assume these banking jobs are especially harmful relative to the average job in the economy, so we can make the Isolated Industry approximation. These assumptions are not implausible given that the NEF has estimated that bankers on average destroy £7 of societal value for every £1 they are paid.18 Then, what are the c-consequences of Ella taking a job in banking? Whether one does harm depends on one’s model of how harm is caused in the banking industry. I’ll consider three options. 18 (NEF, Dec 2009) 40 1. If harm is caused by the mere existence of banking jobs or from something that the individual can’t change, then abilities* are roughly equal i.e. each banker has roughly the same potential for harm. Then R=1 with job-creation and R=0 without jobcreation. In reality, Ella probably has little effect on banking wages, and thus doesn’t increase supply of banking labour, so R≈0 (See §5.5.). 2. If the majority of the harm done depends on a subset of bankers who make terrible decisions or engage in illegal activities, then bankers divide into two groups: (i) a group with abilities* close to zero (perhaps producing positive value), (ii) a subset with positive abilities* (i.e. producing negative value). Again R≈ , unless Ella can displace one of these unethical people out of the industry. In this case Ella has a positive impact: R<0 and perhaps much less than - . It s not clear that Selection Efficiency applies, because these bad people are difficult to identify ex-ante, so you can t expect to knock one out. Nevertheless, supposing that you’re not one of these harmful bankers19, you have some chance of knocking one out. The probability is roughly equal to the relative frequency of these bad people in the pool. 3. If you think that harm caused correlates well with profits produced for the bank, then some people will have negative ability* and others will have positive ability*. It might not be possible to know. The bankers who cause financial crashes often don t intend to do so beforehand. 19 41 Some bankers are profitable; others are not. Selection Efficiency will apply, because banks aim to employ people who make profits20. So, we could model the situation with declining ability* and opportunity*. If we assume everyone has positive ability*, we can use the results from §4.7.4., remembering that R>0 means Ella is producing negative value. What gradient is reasonable? Salary seems like a potential proxy for opportunity*. Its size correlates well with the amount of capital controlled. It increases by 50-100% each time someone gets promoted (see Figure 5). If we assume someone is promoted for every 5 steps of the chain, the gradient is 10-20%. If we assume that people at the bottom of the chain have negative ability*, and Ella has positive ability*, then R values will increase relative to those in §4.7.4., since she s displacing loss-making bankers. Either way, Ella is causing a significant amount of harm, 20 but it s significantly less If profit making potential is predictable 42 than what she does directly. Figure 5. Average Annual Compensation of American Investment Bankers in Order of Seniority21 $4,000,000 $3,500,000 $3,000,000 $2,500,000 $2,000,000 $1,500,000 $1,000,000 $500,000 $0 Concluding, R could be anywhere from negative, to zero, to close to one. It depends on how one thinks harm originates in banking. 4.9. Conclusion Once we move to the Sophisticated Analysis, a much more complete analysis of the counterfactual, it becomes difficult to generalise about replaceability. Many situations will involve chaotic chains that are very difficult to model. In two situations, however, we can develop some understanding of the chain. These are when we can make the (i) Leading Order approximation (in the limit of chain 21 Data taken from http://www.careers-in-finance.com/ibsal.htm 43 length=1 returning to the Simple Analysis) and (ii) the Isolated Industry approximation. In these cases typically 0≤R≤1, with values typically larger than the Simple Analysis. Exactly where it lies depends on empirical facts about the chain, for instance, (i) whether job creation take place (ii) the trend in ability* (iii) the trend in opportunity*. One can see the range of values here (some of the numbers are illustrative): Naïve View Equal opp*, creation Simple Analysis, high relative ability* Equal Opp*, no creation, declining ability* Declining 10% gradient, length 10 Equal Ability*, creation, declining opp* Simple Analysis, low relative ability* Equal Ability*, no creation 0 0.5 1 R R is unbounded both below and above. Situations where R is significantly greater than one seem rare; but could occur for the reason given in §3.3.5. They could also occur when highly able people don’t choose jobs based on impact. For instance, medical school applicants who fail to get a place and enter much higher impact medical research. As shown in §4.8., by entering certain harmful careers, R could be significantly less than -1. 44 Does entering a good industry have good job-related c-consequences? They are likely to be somewhat less than what one appears to do directly, but still significant (R≈0.5). In some situations, the c-consequences could be roughly neutral (R≈0). In others, one will do most of the good one does directly (R≈1). Does entering a harmful industry have bad job-related c-consequences? It depends on how harm is created by the industry. Again, R is likely in the range 0-1, but for an ethically motivated person it might well be R<0. Since R=1 could easily mean the c-consequences have value on the order of saving hundreds of lives, working out R for each career option is vitally important. 45 Part 2 – The Dominance Thesis The lessons of replaceability just learned are often ignored in making career choices, even when they are actually raised. This is odd. If one were given the option of a value-producing career such that R=0 and another such that R=0.5, then it seems one has good reason to choose the second. This could be because people are simply inconsistent. But it could also be because they believe replaceability is not morally relevant in career choice, since it’s outweighed by other considerations. I argue that the job-related c-consequences of career decisions, as understood according to the Sophisticated Analysis, are not only morally relevant, but typically determine which career choice is ethically preferable. This is the Dominance Thesis. It means replaceability is a crucial consideration for everyone whose career choice is ethically motivated. 5. Other Consequences 5.1. Introduction Career decisions have many c-consequences beyond the job-related c-consequences. If these other c-consequences are typically more important or tend to counteract the effects of replacement, then they might give reason to ignore replaceability. This stands quite apart from non-consequentialist views (§6) and general arguments against replaceability (§7). 46 I will argue that although these other consequences can be important, they are highly situation dependent, so frequently don’t give reason to neglect replaceability. Moreover, for someone pursuing an ethical career, they are often not significant. This section will also serve as an exploration of the other consequentialist considerations surrounding career choice. Some of the other c-consequences of career decisions involve: 1. Signalling effects and boycotts 2. The effects on the agent s feelings and personality 3. Supply-demand effects These favour doing good directly. For instance, even if the job-related consequences of taking a harmful job are zero due to replaceability, it might be good to avoid taking it in order to: (1) signal that you want social change; (2) so that you can live with yourself and; (3) to increases the wages of this job, reducing the profits of the company. I will go through these effects in turn, weighing them up against the size of the job-related consequences of career decisions. But before that I will consider another type of consequence, which typically enhances the effects replaceability. 5.2. Compensation Effects Replaceability can be viewed as one of a species of more general compensation effects. Most jobs are performed because a group of people want them performed – they expect the job will create something of value to them. In the counterfactual situation 47 in which you don’t take the job, this demand still exists. So, efforts will be made to bring about the value that you would have created: to compensate for your absence. The most obvious effort is hiring someone else to do the job – replacement. But compensation effects can take other forms. One of the most important is reallocation of tasks. Reallocation of Tasks The process of reassigning the tasks performed by different jobs in order to maximise effectiveness Consider: NHS Budget Cuts Suppose the NHS had to reduce its number of staff and other resources by half. Then the most effective tasks would be allocated to the remaining 50% of the staff. The cost-effectiveness of different interventions carried out by the NHS likely differs by several orders of magnitude. Vaccination programs can produce a year of healthy life22 for under £100, while the NHS funds interventions which cost up to £30,000 to achieve the same benefit.23 So, reducing the resources of the NHS by half could bring about a much smaller reduction in its benefit than 50%. Consider the two plausible premises: 1. The activities of most organisations differ in their effectiveness 22 23 Technically called a QALY, as explained in (Preito & Sacirstan, 2003) (McCabe, Claxton, & A, 2008) 48 2. Most organisations make some attempt to prioritise their distribution of resources to increase their impact If one accepts (1) and (2), then reallocation of tasks will occur in most organisations. The size of the effect depends on (i) the scale of the differences in effectiveness between different activities and their distribution, and (ii) the degree to which tasks can be reassigned. This effect will tend to minimise the difference one makes by taking or not taking a job. Even in the case when you’re not replaceable, it’s not the case that all of the value you create in that position would be lost. Rather, the most important tasks you would have performed will be done by someone else. Here’s an example: Reallocation, no replacement Chris and Mia are equally skilled heart surgeons at a hospital. Suppose Chris didn’t become a surgeon and there was one fewer surgeon at this hospital. Then, the most high priority operations he would have performed would be given to Mia. So, perhaps Mia achieves a higher success rate or works harder. Some of the patients would be sent to other hospitals. Chris’s absence won’t mean that 50% of the benefit is lost. Even if you are replaceable, reallocation could still take place. The most important tasks you would have done might still be reallocated to someone besides your replacements, compensating for any difference in skill, and reducing the effect of your absence. For instance: 49 Reallocation, with replacement Chris doesn’t become a surgeon and is replaced by someone less skilful. Then, the more difficult operations will be assigned to Mia. This will minimise the reduction in the success-rate. Reallocation of tasks applies in most situations to some extent, has potentially large effects acting in the same direction as replaceability. It must also be included in a full analysis of the c-consequences of taking a career. For similar reasons to replaceability, it deserves more attention. 5.3. Signalling Effects and Boycotts Now let’s consider the three types of effects that potentially counteract the importance of replaceability. Williams thought that George shouldn’t take the job at the chemical weapons factory. He considered several reasons a utilitarian might give in order to explain this intuition. These include the signalling effects of the choice – “the effects on the public at large.”24 5.3.1. Personality Signalling Signalling effects can take several forms. One is the effect of signalling a certain type of personality. For instance, in explaining why it’s bad to lie, a consequentialist might point out that just one lie can ruin one’s reputation for trustworthiness, which could hamper one’s life in all kinds of ways. If people started lying prolifically, then all 24 (Williams & Smart, 1973) pp105 50 kinds of social institutions could break down. These negative consequences of lying could explain why it’s normally good to be honest. Personality Signalling One’s actions indicate that one has a certain type of personality, which could influence how one is treated by others. In career choice, it might be that taking a career which involves directly doing good signals that one is a moral person and not corruptible. Conversely, taking a career that involves directly doing harm signals that one is selfish or uncaring. Suppose you’re ethically motivated and pursuing a career that has c-consequences on the order of saving hundreds of lives. Then it seems that the possible danger of people treating you slightly better or worse is a small effect on this scale. The exception would be if it prevents you from being successful with your career. Perhaps being seen as a moral person makes you more likely to succeed. The size of this effect on your career is only likely to be significant is one wants to radically change fields. For instance, taking a charity job might signal that one is a moral person to society at large. But, it’s not going to make you stand out within the 3 rd sector itself! Perhaps one’s friends and family will treat you differently if you signal that you are immoral or moral. This could affect your happiness, which would also affect your career success. On the other hand, it’s not clear that most jobs will cause friends and family to treat you differently. Only a minority of careers are widely considered highly morally commendable or condemnable. If a friend of yours enters a relatively 51 controversial career like investment banking, even if you think investment banks are involved in morally dubious activities, you probably won’t end your friendship. Overall, the effects are not only small relative to job-related consequences, but highly situation dependent. They differ from job to job, and depend on the attitudes of one’s friends, family and colleagues. 5.3.2. Precedent Signalling More commonly raised is that one’s career choices could influence the actions of others. One type of effect is the Precedent Effect: Precedent Effect By taking a certain career, one influences the career decisions of others For instance, working in the 3rd sector could be good because it encourages other people to work in the 3rd sector. As Williams points out, this effect is highly situation dependent, and often small. The extent to which one’s career choice is publicised, one’s level of status and the beliefs of one’s peer group will be important determinants. Few young graduates seem to exhibit much influence over the career decisions of their peers. This is because most career options are well known. At best, if Ella enters investment banking she has a small effect on the extent to which investment banking is regarded as a realistic and socially acceptable option; but since the option to take banking jobs is well known and promoted, the impact seems small. If, however, she takes a very unusual career 52 in a public style, then it seems more likely she will inspire others to do the same. For instance, schools often promote the activities of famous alumni among their current students, which could encourage these students to take similar paths. Precedent effects could be large if one triggers a threshold, contributing to a boycott as explored in §5.3.5. 5.3.3. Political Signalling Career decisions can also signal political beliefs: Political Signalling By taking a certain career, one displays one’s political beliefs, which could influence the actions of others For instance, taking a job in banking (causing harm, for the sake of argument) in 2012 could have negative effects because it signals support for current economic system and all its ills. Political Signalling is, like Precedent Signalling, situation dependent and often weak. 5.3.4. The Replaceability of Signalling Effects The actions that lead to signalling effects are often themselves replaceable. Suppose taking a high profile job in banking has bad signalling effects. But if you don’t take the job, someone else will. That person will create the bad signalling effects in roughly the same way you would have done. They would display to all of their friends that they support capitalism and so on. So, for jobs in which you would be 53 replaceable, there’s apparently little net change in the total amount of negative signalling if you take that job. This is a simplification, because these cases also involve a chain of replacements.25 Suppose you’re considering whether to take a certain 3rd sector job and you believe that 3rd sector jobs have positive signalling effects. If you don’t take it, then someone else will, so apparently you’re having no signalling effect. But if that person will only take a job that directly does good (e.g. will only take 3rd sector jobs), then although they don’t take your job, they’ll take another 3rd sector job. If this increases the total pool of people in 3rd sector jobs, then you might still have a positive signalling effect. It would depend on whether the final person in the chain ends up in a job with a positive signalling effect, and how large this effect is. Moreover, even if you would be replaceable, you could still add to the total pool of applicants for that job. That has some signalling effect in itself. The fact that investment banking jobs continue to have huge numbers of applicants per place even after the financial crisis suggests that many people believe they are not unethical. 5.3.5. The Modification of Signalling Effects All three signalling effects can be modified by how one acts in the other areas of one’s life. Ella’s taking of a job in banking might normally be perceived as (i) selfish (ii) a show of support for capitalism and (iii) setting a harmful precedent for her peers, Signalling effects could in large part be included in the apparatus of the Sophisticated Analysis §4 25 54 causing all three types of negative signalling effect. However, if the job is taken for ethical reasons – reducing the harm done by banking26– then the signalling effects change. Suppose Ella is generally known to be a moral person outside work. Then, the Personality Signalling effect becomes much smaller, since her friends and family still believe she is a moral person. The Political Signalling effect potentially changes from negative to positive. Rather than displaying support for capitalism, taking the job displays a recognition of its harms, and a desire to reduce them. If this is positive, then the Precedent Effect also becomes positive, because it’s good to encourage others to do the same. Since few people enter banking with the explicit intention of making it less harmful (rather than the aim of making a lot of money), the signalling effects are often negative in practice. It’s easy to confuse this with the signalling effects being negative in general. If a strongly ethically motivated person enters banking, however, the signalling effects could be very different. 5.3.6. Organising Boycotts One might respond that if these arguments are taken seriously in general, it would stifle the chance of social change, leading to worse c-consequences than we took the Naïve View. Here’s an argument to that effect: 26 Or making large philanthropic donations 55 Tish the Wheat Trader 1. Tish believes that wheat trading jobs on balance produce negative value, so a world without them is better than a world with them. 2. However, she takes replaceability into account, and reasons that the perfect replacement condition holds. Perfect Replacement Condition: if she takes the job, someone else equally good would take it, so the c-consequences of the decision are zero 3. So, Tish takes the wheat trading job, believing that it s ethically neutral. 4. Universalization Test: If everyone reasoned like Tish, then people will continue to apply to these jobs, and wheat trading will continue to exist. 5. If everyone instead avoided jobs that directly produce harm, then wheat trading would cease to exist 6. So, the world in which everyone reasons like Tish is worse than the world in which everyone acts as if the Naïve View is true 7. So, one should in general not take replaceability into account First note Perfect Replacement is an unlikely condition, and it’s especially unlikely that everyone would believe it. Normally, R>0, so taking a harmful job still causes some harm through the job-related consequences. Tish will also have a small but negative Precedent Effect, since she adds to the total number of people applying to the job. But let’s assume it holds in this case and everyone believes it. 56 We then apply a universalization test. Failing this test would imply it’s wrong to act according to the principle according to some ethical systems, like rule consequentialism. It’s also taken to entail that one shouldn’t campaign to raise awareness of the idea, but notice that this doesn’t actually follow unless one expects to persuade everyone to believe it (or makes a more complex argument). Let’s assume everyone believes the condition and that universalizability matters. Considered at an instant, agents would continue to apply to wheat trading jobs. But in reality, agents can communicate. This means there is some chance of organising a boycott of wheat trading. Tish should take into account that there’s some (perhaps very small) chance that her career choice hampers the boycott through the Precedent Effect. This is an additional negative effect of taking the career, which means it’s no longer ethically neutral. So, we’ve actually shown that Perfect Replacement should only be believed if understood as: Perfect Replacement is true, if everyone else’s actions are fixed. If one takes into account the chance of organising a boycott, then it can’t be the case that everyone truly believes Perfect Replacement. Note that we haven’t shown anything about Imperfect Replacement, which is the normal case. 57 A similar type of example been explored by Shelly Kagan.27 For instance, consider his example: Polluting Factory A factory emits a gas which is thinly scattered across the surrounding area so that no person breaths in more than one of the molecules. One molecule has no effect on one’s health. However, there are thousands of similar factories. Inhaling thousands of molecules has a very negative impact on one’s health. Each factory owner apparently makes no difference, since one molecule never makes a difference. The end result is very destructive, far offsetting the value produced by the factories. We could now suppose Tish believes that a boycott is possible, but that her choice has a negligible effect on its chances of success: Vague Replacement Condition If one extra person takes a wheat trading job, it won’t affect whether a boycott of wheat trading happens or not. This condition seems implausible in the case of career choice. In reality, Tish does make some difference to the chances of a boycott, it’s just very small. Kagan points out that if the difference is small enough, then it will make an imperceptible difference to people’s happiness. It’s unclear that imperceptible differences in happiness could matter. In this case, we could mount a second response. Perhaps Tish’s action has a high probability of making no difference, but a small probability of making a significant difference. It seems plausible that social movements involve 27 (Kagan, 2011) 58 tipping points28. To organise a boycott requires some people to start, which could encourage yet more people to boycott it. Ex-ante, there’s a small probability that Tish’s career decision triggers a boycott of wheat trading. This is why the Vague Replacement condition is false. 5.4. Effects on the agent’s feelings and personality 5.5.1. Introducing the effects Williams also considers the fact that George feels that he doing evil by working at the chemical weapons factory. Similarly, Jim might feel revulsion at having to kill an Indian. I call these Emotional Effects. Emotional Effects Career choices could lead to positive or negative emotions for the agent and the people the agent interacts with. Emotional effects are often raised as a reason to overlook replaceability. If there’s a correlation between doing good directly and feeling good, then it’s a reason to favour doing good directly. Also note that most people regularly interact with tens or hundreds of other people outside of their job. If their job makes them miserable, then it could negatively affect these people too. 28 For a detailed model, see (Granovetter, 1978) 59 Emotional effects are linked to two other kinds of effects on the agent. First, jobs that involve doing good directly could be easier to stick with, so have a lower risk of burn-out29: Burn-out Risk The chance that someone gives up on a career because they find it too demanding. If being involved in harmful activities makes you unhappy, then you might have a high risk of burn-out, reducing the expected-value of the choice. Second, pursuing jobs that directly cause harm involves a risk of Corruption: Corruption Risk The risk of corruption is the risk that the agent’s moral beliefs or ability to act on those beliefs are negatively changed by their working environment. The people one works with have a subtle influence on one’s moral beliefs over time. Moreover, the types of activities one is engaged in could alter one’s dispositions to act in the future. For instance, some results in psychology suggest that thinking about money increases one’s disposition to be materialistic.30 5.5.2. Situation Dependence Before going into Emotional Effects and Burn-out risk in more detail, note that all three effects are highly situation dependent. While one person might find wheat trading repulsive, another might not, even if they also believe it’s harmful. The 29 30 Burn-out and Corruption are introduced and explored in (Crouch, Unpublished) (K. D. Vohs, 2006) 60 danger of moral corruption also varies greatly between people and between jobs. Some people will even inspire their colleagues to act more ethically. Furthermore, one can often purposely alter the situation in order to offset these affects. For instance, if one is worried about being corrupted or burning out, one could (i) associate with a group of like-minded people (ii) make a public declaration of your intent to act ethically. The negative emotional effects of a career could be counteracted by focusing on the good one can do in other aspects of one’s life. 5.5.3. Emotional Effects If one becomes significantly less happy, and that impacts the happiness of the hundred people you most regularly interact with, that might be good reason to avoid an ethically neutral job. But if hundreds of lives turn on the decision, then that outweighs the Emotional Effects. After all, the value lost to 100 people through being less happy is small compared to the value lost if 100 people die. Williams argues for the same point. Even if George hates his work at the chemical weapons factory, it’s a minor consideration compared to the suffering that could result from the weapons developed by the zealous researcher. Emotional effects might lead to other effects that are significant. For instance, negative emotional effects could increase the risk of burn-out or corruption. They could also hamper one’s effectiveness in a job. This could be very significant, and is a good reason not to take a job that one hates even if seems very high impact. 61 5.5.4. Burn-out The burn-out risk of a career path will generally lower its expected value. Whether this should change your career decision depends on what job you would go into if you burn-out, and how much good you do before you burn-out. Consider someone choosing between professional philanthropy and 3rd sector work. The risk that they burn-out might be higher in professional philanthropy (depending on the person, and note there’s a risk of burning out in the 3rd sector too!). But, if they believe the annual value produced by professional philanthropy is much higher than through 3rd sector work, it might be worth attempting. Depending on the nature of the highearning job, even if they burn-out, they might still be able to enter a similar 3rd sector job. Burn-out risk also shows that being overly zealous in the pursuit of maximising cconsequences is unlikely to be a good strategy. 5.5. Supply-Demand Effects 5.5.1. Introducing the effects One’s career choices have an effect on the supply and demand for certain jobs. Being willing to take a certain type of job increases the supply of labour for that type of job. This has two effects: (i) it decreases equilibrium wages, decreasing industry costs, and (ii) the industry will hire more people to do that kind of job. 62 For markets assumed to be in equilibrium, one can use neoclassical economic models to estimate the size of these effects, provided one knows the elasticities of labour demand and supply for the relevant industry. The elasticities are difficult to estimate empirically and a subject of intense debate in economics31. Nevertheless, one can make rough estimates. If the elasticity of labour demand is -1 (a reasonable central estimate32), then since the elasticity of supply ranges from 0 to infinite, increasing the labour supply 1% causes an increase in the quantity of labour supplied from 0% to 1% at the new equilibrium.33 This means that adding yourself to the supply causes, on average, between zero and one unit of job-creation in the industry. Most likely, the true figure lies somewhere in the middle.34 So, the Iteration Effect chain ends somewhere between job-creation and no job-creation. It also causes a -1% to 0% change in the wage rate. This means that adding oneself to the supply on average means that wages reduce so that the industry’s costs decrease by zero to one (average) person’s salary. If the elasticity of labour demand is -0.5, which would be more typical of jobs with high qualifications, then a 1% increase in supply, causes a 0-1% increase in quantity supplied (most likely 0.5-0.66%), and a -0.5%-0% change in the wage rate. This meta-study shows that elasticity of supply estimates range from below zero to over 2: (Evers & R. de Mooji, 2005) 32 This well-cited book includes a study of manufacturing jobs, finding values of -0.9 to -1.4 (Hamermesh, 1996). 33 I assume both elasticities are linear since I m considering small changes. 34 It s .33% to .5% if the elasticity of labour supply is from .5 to , which is in line with the best estimate in this meta-study: (Evers & R. de Mooji, 2005) 31 63 It’s difficult to estimate the significance of this effect compared to job-related consequences. It depends on the value produced per unit of profit, compared to the value produced by someone in a job. 5.5.2. Non-equilibrium, Imperfect Markets The above only applies is the labour market is in equilibrium and describable by neoclassical economics. In reality, labour markets are frequently out of equilibrium and not perfectly competitive. Real markets involve persistent unemployment, suggesting the average job is over supplied. This would mean one’s true impact on wages is less than what I’ve just estimated. Some industries support wages in excess of the rates required for markets to equilibrate. The financial services sector currently has large excess wages of ~50%35 suggesting that adding oneself to the supply has almost no effect on wages or jobcreation. This supports the intuition that hugely competitive application processes mean that a single applicant has effectively no impact on wages. 5.6. Conclusion As shown, none of these effects provides a general reason to neglect replaceability. They can be highly relevant if one believes that most careers are roughly neutral in their c-consequences. This belief could explain why we typically put a lot of weight Estimated by comparing wages between jobs requiring similar skills (Philippon & Reschef, 2009) 35 64 on them in careers advice. For someone ethically motivated however, the difference in the job-related c-consequences between different choices could easily be on the order of saving hundreds of lives. Then, except insofar as they indirectly impact the job-related c-consequences, these other effects are outweighed. 65 6. Non-Consequentialist Considerations 6.1. Introduction Although, as shown in §2, c-consequences are morally relevant, according to nonconsequentialist views they are not the only factor that’s morally relevant. Perhaps it’s just wrong to take certain careers, because they violate rules we should all respect. Or perhaps taking harmful careers compromises our integrity; while doing good directly is virtuous. Indeed, Williams uses the fact that George is replaceable precisely to show that not only c-consequences are morally relevant. Consequentialists reject the relevance of these factors. So, for consequentialists, I’ve already shown that replaceability is a crucial consideration. Nevertheless, I will show that despite Williams’ example, the c-consequences of career decisions typically dominate which is ethically preferable even for non-consequentialists. I will explore three non-consequentialist considerations: side constraints, virtue and integrity. I show that neither outweighs c-consequences in general. Before I begin, note that since I’m only concerned with the ethical preferability of different options, demandingness is not relevant to my argument. If the most preferable career is highly demanding, it might be permissible not to take it, but it will remain the most preferable option. 66 6.2. Side Constraints 6.2.1. Careers in the Mafia Deontological non-consequentialist theories involve the addition of Side Constraints: Side Constraints Principles that prohibit certain actions, no matter the c-consequences Common Side Constraints are principles like ‘never use people as a means to an end’ or ‘never torture.’ If one adopts Side Constraints, then they outweigh c-consequences whenever they apply. One can argue that Side Constraints are relevant by applying replaceability to an illegal career: Lorenz the Mafia Boss If Lorenz doesn’t work for the Mafia, someone else will, since in his town unemployment is high and lots of young men want the money. However, since he’s ethically motivated, he will aim reduce the harm that would have been done otherwise by reducing the number of shootings. Supposing this is the career with the best c-consequences open to him, should Lorenz take the career? Most people find this career impermissible, despite the net positive c-consequences. Presumably this is because entering the Mafia violates Side Constraints. In fact, there are good consequentialist reasons to think such a situation is extremely unlikely. Although in §5 I argued that the other c-consequences of career decisions 67 are typically outweighed by the job-related c-consequences, this doesn’t seem to be the case for illegal careers. Let’s compare a typical ‘unethical’ career, banking, with this example. 1. Since Lorenz is ethically motivated, it s difficult to imagine he could really go through with the career. He would burn-out. There are, however, ethically motivated bankers, suggesting the same is not true of banking. 2. Careers in the Mafia are rare and illegal, so the signalling effects of an extra person attempting to join the Mafia even if Lorenz doesn t increase the number of people in the Mafia) are strong. The same is not true of banking, which employs thousands of young graduates each year and is not illegal. 3. Working for the Mafia carries significant risks of violent death or imprisonment, which would be extremely stressful for Lorenz and all his friends and family, so there are large negative Emotional Effects. 4. Criminal organisations, which offer job positions, are often labour constrained due to the risks associated with these careers. So, criminal careers are often not replaceable. So, it’s not replaceability that caused the problem; rather, it was neglect of the other consequences. Suppose the career in the Mafia really is the career with the best consequences for Lorenz. A consequentialist must bite the bullet and concede that he should take the career. If one still finds it impermissible to take the career, then nonconsequentialist factors must be relevant. 68 The career in the Mafia is apparently analogous to a Transplant Case36. This involves a doctor killing a healthy patient in order to use their organs to save the lives of five others. This clearly seems impermissible. A non-consequentialist explanation is that it violates the rights and autonomy of the health man. Similarly, Lorenz shoots different people from his replacement, in order to reduce the total number of shootings. 6.2.2. Side-Constraints in Typical Career Decisions If one adopts Side Constraints, how are c-consequences ethically relevant? A plausible position is that so long as Side Constraints are obeyed, actions are ranked in order of ethical preferability by their c-consequences. Suppose there are three actions open: A1, A2 and A3. One adopts Side Constraints, which prohibit A1. The other two actions are permissible. But suppose the c-consequences of A3 are much better than A2. Then, A3 is ethically preferable to A2. A3 might be a moral obligation or superogatory. This model gives the right answer to Paul and the Paramedics. Both performing first aid and letting the paramedics work are (presumably) permissible, but letting the paramedics work is ethically preferable. How much impact might Side Constraints have on career choice? Let’s define37: 36 37 As introduced in (Thomson J. J., 1985) After (Crouch, Unpublished) 69 Morally Innocuous Career A career which lacks a strong non-consequentialist reason against taking it. In particular, taking a Morally Innocuous Career doesn’t involve violating a Side Constraint. It seems plausible that most careers are morally innocuous. Any nonharmful career is unlikely to violate a Side Constraint. Most careers are not seen as harmful, so most career are morally innocuous. For these careers, c-consequences determine which is ethically preferable. Side Constraints have the best chance of being relevant when considering unethical careers. But even considering careers that are often claimed to be unethical, it’s not clear that taking the career violates Side Constraints. One should distinguish the act of taking a career from the morality of the typical people in that career. For instance, it’s plausible that banking is considered unethical because in the recent past many bankers have been unethical. This doesn’t mean that entering banking is unethical. Actions that violate Side Constraints are widely regarded as wrong. It’s not widely considered wrong to take most unethical careers, like being a trader at a bank, working for big oil or being a corporate lawyer. Rather their status as unethical is controversial. This suggests that even taking typical unethical careers does not violate Side Constraints. At the least, taking these careers is certainly not considered as bad as killing or torturing someone. It’s less clear that taking them, however, is not like violating a minor Side Constraint like not lying. 70 It might be that most people are wrong about the moral status about these careers. The mechanisms by which they cause harm are often complex. Since they don’t involve directly harming, people are unwilling to declare them as morally reprehensible as burglary. Nevertheless, if they were sure about the mechanisms by which harm was caused, they would find them equally bad. So, although there’s some reason to think that even taking typical unethical careers doesn’t violate Side Constraints, it’s unclear. We don’t need to be clear, however, because even if we consider a career such that taking it does violate a Side Constraint, there are still three situations in which taking the career is still permissible in light of the c-consequences: 1. Emergency Situations 2. Pareto Improvements 3. Trolley Cases I provide a preliminary discussion of each, arguing that Side Constraints are often outweighed by the c-consequences of career decisions even when they are indeed violated. 6.2.3. Emergency Situations Threshold theories permit Side Constraints to be overwhelmed when the cconsequences are large enough. Many deontologists admit the existence of thresholds due to Ticking Bomb cases. In these, one is asked to choose between saving a large number of people and violating a Side Constraint, such as torturing 71 someone. The number of people can be made arbitrarily large, making it implausible to hold onto the Side Constraint. The response of deontologists is to postulate thresholds at which point c-consequences overwhelm Side Constraints. If replaceability means that the c-consequences of taking a career are very positive, then it could be permissible according to these theories to enter the career even if it would violate a Side Constraint. For instance: George and the Bioweapons Lab George is offered a second job at a bioweapons lab. He suspects the lab is working on an extremely dangerous and risky virus. He believes that by working at the lab he can delay the progress of the research by working less efficiently than the typical researcher. Even a small delay to the program has high expected c-consequences. Suppose George thinks that contributing to this research violates Side Constraints, such as profiting from the suffering of others. If the expected c-consequences were large enough, it might overwhelm the Side Constraints. Careers that involve espionage or sabotage could also be of this kind. The German military leaders who died attempting to assassinate Hitler during WWII are often considered heroes. Spies who lie and kill for the country often celebrated too, at least in fiction. It’s plausible that in entering politics, one might expect to have to eventually violate side constraints. Nevertheless, this could easily be outweighed by the potential to benefit millions of citizens. 72 6.2.4. Pareto Improvements One reason it seems to be impermissible to kill the healthy patient in the Transplant Case is that it violates that patient’s rights. This is because the surgeon changes who is harmed. This is not the case with Jim and the Indians. If Jim asks the Indians what he should do, the one who he is going to kill is indifferent, because he will be killed anyway. The other 19, however, will be saved, so prefer Jim to kill the single Indian. So, some of the Indians benefit and none is made worse-off by Jim’s action38, thus the action is a welfare Pareto improvement. Since Pareto improvements don’t involve acting against anyone preferences, it’s widely accepted that they are permissible, even if they might violate a Side Constraint (such as killing). Williams himself accepts that it’s preferable for Jim to kill the single Indian. It’s not plausible, however, that entering many harmful careers brings about a Pareto improvement. One will act differently from one’s replacement, which means that ultimately different people will be harmed: the banker will place slightly different deals and George will develop different weapons. Even if it does, it will be extremely difficult to know that it does. Our actions in careers have very complex effects. If anyone is made worse off by an action to any degree, then it’s not a Pareto improvement. More plausibly, however, taking a harmful career in order to have positive cconsequences could often be an ex-ante Pareto improvement. This means that no-one 38 Ignoring Jim himself! 73 faces a negative expected-benefit before the action, even if it turns out to not be a Pareto improvement ex-post. For instance, suppose Jim is told that if he doesn’t kill a single Indian, Pedro will kill 10 of the 20 at random. Ex-ante, each Indian prefers Jim to kill the single Indian, because that increases their chance of survival from 1/2 to 19/20. Ex-post, however, it could be true that the Indian Jim kills would not have been killed by Pedro, so Jim’s action made that Indian worse off. Ex-ante Pareto improvements still have the attractive feature that no-one’s desires prior to the action need to be over-ridden. Since the harms caused by many careers are so indirect, it’s often impossible to know who will be harmed. If someone is planning to do significantly less harm than their replacement, however, then it’s plausible that no-one faces a negative expectedbenefit. This is particularly likely if we restrict our attention to the victims of unethical careers. For instance, investment banking is considered bad because of the suffering inflicted on the poor by financial crises. The impact of any single banker on the details of a financial crisis is unknown. So, if a banker expects to lower the risk of financial crisis, then each poor person in the society might face an expected-benefit. In fact, some poor people will be made worse off by the change in timing of the crisis, but this is not known in advance. One problem with using ex-ante Pareto improvements is that although taking the career might be one, it might be necessary to perform actions that aren’t ex-ante Pareto improvements during the career. For instance, it might be known by an oil company is going to harm several villages in the developing world but it’s not known which. Graham will harm fewer villages than his replacement if he gets the 74 job, so all of the villagers prefer Graham to take the job. Nevertheless, once Graham is actually in the job, he has to pick which villages will be harmed. The action of picking is not an ex-ante Pareto improvement. My intuitions about what Graham should do are unclear. One answer is that he should resign if he finds himself in this situation. Alternatively, it could be reasonable to treat the career as a single action. 6.2.5. Trolley Cases In the Trolley Case39 a run-away trolley is hurtling towards five people on a track. You have the option to divert the trolley onto another line, so that it only kills one person. Unlike the Transplant Case, many people have the intuition that it’s permissible to divert the trolley, killing the one person. The Trolley Case is not a Pareto improvement, and many different arguments have been put forward as to why it’s permissible to divert the trolley. Thomson40 has argued that it’s because by diverting the trolley, one is diverting a pre-existing threatening force by acting on the force. In the Transplant Case, however, the surgeon creates a new threat in killing the patient. The Trolley Case is like a pilot diverting a crashing plane into a lowly populated area. Different people will be harmed, but it seems clearly better to minimise the destruction, even if the ratio of harm to benefit is significantly smaller than five to one. 39 40 First raised in (Foot, 1967) (Thomson, 1986) 75 Entering a harmful career with the intention of making it less harmful often seems analogous to the Trolley Case. Careers roles that produce negative value are preexisting threats. This is because if you don’t act, someone else will take the career and cause the harm, just as the trolley would hurtle into the five people. It’s ethically preferable for you to take the position and divert the harm that would be done. Similarly to manipulating the trolley, one acts on the threat to yield the benefit. As a banker, one makes different deals. As an oil engineer, one makes different plans. One doesn’t directly harm, as with killing the healthy patient or Lorenz shooting different people. A source of dis-analogy is that these career choices seem to be dilemmas of acquiescence rather than of necessity.41 Diverting the trolley is a dilemma of necessity, because the trolley is not an agent. Rather, it’s a mindless harmful force. The danger that one’s replacements will be more harmful than oneself because they are less ethically minded is a dilemma of acquiescence. It involves an agent’s decision to act unethically. Woodard claims that this provides additional reason not the take the harmful career. It’s unclear whether this outweighs the other ways in which the Trolley Case is analogous. Moreover, it’s not true of all cases. For instance, if one’s replacements would be more harmful because they are less well informed, then it would no longer clearly be a dilemma of acquiescence. Switching the path of the trolley is also justified with an application of the widely accepted Doctrine of Double Effect (DDE). DDE states that it’s permissible to harm if (i) it’s in the pursuit of a much greater good and (ii) one doesn’t intend to cause the 41 Explored in (Woodard, 2009) 76 harm. One doesn’t intend to kill the single person on the track. Rather, their death is an unintended side-effect of one’s efforts (diverting the trolley) to save the five people. This is not true of the Transplant Case, which seems to involve an intention to kill the healthy person. Many legal, but unethical careers, however, don’t involve intending to do harm. A banker selling complex securities doesn’t intend to harm anyone (unlike a fraudulent banker). She intends to earn a profit, and would earn the profit in a less harmful way if it were possible. It’s an unintended side-effect that these securities contribute to financial instability. The Mafia, however, intends to kill people in order to intimidate others. If the c-consequences are great enough, then DDE could justify entering an unethical career provided the activities don’t involve intending to do harm. Note that, under DDE, it’s permissible to create a threatening force as long as one doesn’t intend to do harm. For instance, tactically bombing a munitions factory to prevent war might be permissible, even though an unintended side-effect is to kill civilians. This is particularly relevant to professional philanthropy, because pursuing harmful careers can enable you to save thousands of lives through your donations. It’s unclear whether the impact of donations can be included in the job-related cconsequences, making R<<0, and thus similar to a Trolley Case, or whether it might be that R>0 so more like a Tactical Bombing case. 77 6.2.6. Conclusion Side Constraints aren’t relevant to most career decisions, so c-consequences dominate. Even taking some unethical careers probably doesn’t violate Side Constraints. Moreover, even if they do, it could still be permissible to take the career if it’s an emergency situation, ex-ante Pareto improvement or analogous to a Trolley Case. Answering exactly when it’s permissible requires a much more detailed application of the latest ethics of permissibility.42 6.3. Virtue Virtue Ethics places moral relevance in the character traits of agents, which is a natural but completely different way to think about ethics from consequentialist or deontological theories. I’m not able to do justice to Virtue Ethics, but I discuss it because it’s often thought that careers doing good directly enable one to express more virtuous traits. Sarah should persist in becoming a surgeon because this enables her to express a character that’s compassionate, courageous, honest and so on. The opposite seems to be the case for harmful careers. This suggests that replaceability is not relevant in Virtue Ethics. C-consequences must still be important in Virtue Ethics, however, due the example of Paul and the Paramedics (§2.2). Any reasonable ethical system must have something to say about why it’s bad for Paul to paralyse his mother. The importance of c-consequences is embodied in virtue of beneficence – the virtue of helping others and preventing harms. By taking the job at the chemical weapons factory, George can 42 As developed, for instance, in (Kamm, 2007) 78 (by supposition) prevent harm, so he’s acting beneficently. All else equal, it’s better to be more beneficent rather than less, so careers with better c-consequences are to be preferred. Is all else equal? It seems the majority of morally innocuous careers provide equal opportunities to express one’s virtuous character. The difficult case involves harmful careers. Even then, it’s not clear taking the job compromises George’s ability to express a virtuous character. Rather, so long as he’s not corrupted43, he would show considerable courage and determination by placing himself in an environment he finds repugnant in order to help others. Further, virtuous character traits are not blindly followed rules. Rather, one embodies beneficence or honesty with a degree of practical wisdom. Consider44: An honest person cannot be identified simply as one who, for example, always tells the truth…for one can have the virtue of honesty without being tactless or indiscreet. Ordinarily taking the harmful job is perhaps not virtuous, especially if it’s for personal gain. However, if lives are at stake, George need not blindly follow these rules in order to be a virtuous person. A strength of Virtue Ethics is that it encourages a holistic assessment of someone’s morality. If George embodies virtuous character traits throughout the rest of his life, we might judge that he remains a virtuous person even if his career involves some 43 44 A more important danger in Virtue Ethics SEP, (Hursthouse, 2012) 79 activities that would normally not be considered virtuous. We know, for instance, that George is morally opposed to chemical weapons and is ethically motivated. In this context, it seems highly virtuous for George to take the job and seek to reduce the suffering caused by chemical weapons. A wide variety of careers are potentially virtuous. This leaves c-consequences and the virtue of beneficence as the primary determinant of differences between choices. 6.4. Integrity It’s often considered morally wrong to take careers doing harm because they violate one’s integrity. Williams argues that this is why George shouldn’t take the job at the chemical weapons factory, and moreover, that this fact can’t be explained merely in terms of c-consequences. Integrity might also be proposed as reason to take careers that directly do good, even if they don’t make a difference. One might think that integrity just consists of a particularly strong form of Emotional Effect on the agent (§5.5.3.). Williams argues that Emotional Effects are unlikely to be strong enough to show that George shouldn’t take the job. After all, George is just one man, and the work at the chemical weapons factory is likely to (by supposition) to contribute to the violent deaths of many more people. A consequentialist would respond that in this case, George should take the job. For Williams, there is a morally relevant consideration over and above the c-consequences. 80 For Williams, integrity consists of the intrinsic value of staying true to one’s life projects. George is morally opposed to chemical weapons, and so shouldn’t involve himself in their manufacture. But it’s plausible that not being involved in chemical weapons research is not among the life projects of many other people. In that case, integrity’s not a consideration and they should take the job if in a similar situation. The relevance of integrity varies from person to person and from job to job. Cconsequences remain highly ethically significant because they determine what should be done whenever integrity is not relevant. Integrity can, like Side Constraints, be outweighed. Williams thinks that Pedro is making an attack on Jim’s integrity in forcing him to kill an Indian – it’s among Jim’s life projects not to kill innocents. But Williams thinks it’s right for Jim to kill the single Indian. It’s not worth Jim preserving his integrity at the expense of the lives of 19 Indians. Harmful careers could easily involve harming people on this scale. So, if one can enter one of these careers with an expectation of bringing about less harm than would have happened otherwise, it’s plausible that integrity is outweighed. In fact, it’s quite plausible George is a case of this sort too. It seems that George should take the career despite his integrity45. Integrity based objections to replaceability often turn on a confusion. It’s often unclear exactly what someone’s life projects are. We’re told that George is morally opposed to chemical weapons, so his life projects include not contributing to research into chemical weapons. But, perhaps George’s life project is actually preventing the use of chemical weapons. The first implies that he shouldn’t take the job. The second 45 Overlooking that it might not be a Pareto improvement like Jim s action 81 implies that he should take the job, because by taking the job he can hamper the development of chemical weapons, reducing the scale on which they are likely to be used. The second project seems to be the better project. Williams distinguishes lower-order projects, like preventing the use of chemical weapons, from higher-order projects, like the utilitarian project of searching for the greatest happiness of the largest number of people. It’s plausible that whatever George’s higher order projects are, they are better served by the project of preventing the use of chemical weapons on people than the project of never researching chemical weapons. After all, what’s bad about chemical weapons isn’t that we know about them, it’s that they might be used to harm people. Raising integrity could even be question-begging in this context. Our career is one of our life projects. I’m making an argument about which career is ethically preferable. Thus, I’m making an argument about which life projects are best to have. So, responding to my argument by saying that it conflicts with our life projects begs the question. Williams responds that a moral theory which requires us to change our most deeply held life projects is too demanding. I, however, have set aside considerations of demandingness,46 so integrity isn’t relevant. But see (Ashford, 2000) for a response 46 82 6.5. A Real Example These examples might be considered unintuitive, but consider a real life case. Oskar Schindler47 took a job producing weapons for the Nazi Government during World War 2. This sounds like a career that would certainly violate Side Constraints and Schindler’s integrity. But Schindler is widely acclaimed as a hero, as celebrated in the book Schindler’s Ark and the film Schindler’s List. His biographer called him, “one of the most remarkable Righteous Gentiles in the Holocaust.”48 This is because he purposely made the factory less efficient, so that it would produce fewer arms than it would have produced otherwise. He also used his position and money to save the lives of around 1000 Jews. Schindler could have fled the country, but he stayed in his position, helping the victims, earning our moral praise. 6.6. Conclusion Side Constraints and Integrity turn out to normally not be relevant to career choice. Moreover, when Side Constraints are relevant, they are often outweighed by cconsequences, so it’s permissible to take the career with the best c-consequences anyway. Virtue Ethics doesn’t judge individual actions as right or wrong. Rather, certain careers enable one to be far more beneficent than others. Lacking a reason to believe these careers are much less virtuous in other ways, it’s preferable to take them. 47 48 This example is borrowed from (Crouch, Unpublished) (Crowe, 2007) pp.624 83 Thus, in the majority of ethically motivated career decisions, even if these views are relevant, it’s typically preferable to pick the option with the best job-related cconsequences. 84 7. Analogous Arguments 7.1. Introduction A final type of objection to taking account of replaceability is that it’s analogous to supposedly bad arguments like, ‘if I don’t buy the meat, someone else will,’ and ‘if I don’t invest in the arms company, someone else will.’ In this section, I explore the analogies and disanalogies between the cases. Rather than give us reason to ignore replaceability, these examples deepen our understanding of it. 7.2. Replaceability in Vegetarianism 7.2.1. A Bad Argument Consider a moral position in which eating meat is wrong due to the animal suffering it leads to. Consider: Supermarket Meat Bob purchases some beef at the supermarket, which is about to sell out. He argues that if he doesn’t buy the meat, someone else will, so it won’t affect the welfare of the animal in question either way. This is a bad argument because it neglects Bob’s supply-demand effect. In reality, Bob causes a small expected increase in beef prices, causing a small increase in quantity of beef supplied. Using realistic figures for fed-beef49, buying an extra 1 unit of beef, causes an extra 0.8 units to be supplied to the market. So, there’s a grain of truth in Bob’s argument. He doesn’t expect to cause all of the suffering associated 49 Taken from: (Langemeier & Thompson, 1967) 85 with 1 unit of beef. Rather, by raising the prices of beef, he on average causes someone else to eat slightly less. So, he only causes the suffering associated with 0.8 units. Note that similar arguments are made in the context of burning fossil fuels; for instance, ‘if I don’t buy this plane ticket, someone else will.’ These arguments fail for the same reason. Note also that the supply-demand effect is different in career choice, because the job one causes to be created will generally be occupied by a different person. 7.2.2. A Good Argument Pepperoni Pizza at a Party Bob is, unknown to his friend, vegetarian. Upon arriving at the friend’s party, he discovers his friend has bought a large quantity of pepperoni pizza. He reasons that if he doesn’t eat the pepperoni, someone else will. Either way, it won’t contribute to animal suffering – there’s more than enough food for everyone to fill up. He makes it clear that he would prefer vegetarian pizza in the future. Since this meat has already been bought, and in the future Bob’s friend won’t buy more meat on his behalf, Bob doesn’t have a supply-demand effect. So in this case, it’s true that he doesn’t directly contribute to animal suffering by eating the meat. 86 Many people have the intuition that he still shouldn’t eat the meat. This could be explained by signalling effects. By being public about his vegetarianism, he could encourage others to become vegetarian. As with the career choice signalling effects, these will depend on the details of the situation and Bob’s friends. It might be that by being a reasonable vegetarian who doesn’t want to let food go to waste, he’s more persuasive. Eating meat this time might make it harder for him to resist temptation at other times. Forming the habit of never eating meat, might be the best way for Bob to be a vegetarian. This is analogous to the Corruption Effect in career choice. Bob will need to assess the size of these effects in the light of his situation. 7.2.3. Eating Meat Not a Good Analogy These two examples have no analogue with the job-related c-consequences. Buying meat does not produce a stream of value in the same way as occupying a job does. The job-related c-consequences are often the most important factor in career choice, so buying meat is also a bad analogy. 7.3. Ethical Investing 7.3.1. Introduction Should universities invest in arms companies? Oxford’s recent investment of £630,000 in Northrop Grumman, a company that profits from cluster bombs, was 87 considered headline-worthy50. In 2010, Oxford founded the Socially Responsible Investing Review Committee, after pressure from the SRI Campaign. The campaign states, “by modifying…its investments in line with its values, an institution can recognise its moral responsibility, and have a practical positive impact in the world.”51 It appears that investing in arms companies is considered unethical due to its negative impact. But if Oxford doesn’t buy the shares, someone else will, so prima facie Oxford makes no impact at all – or else there is something wrong with replaceability. Ethical investing is a good analogy with career choice, so it’s worth exploring in some depth. Rather than reject replaceability, I show that we should rethink our views of ethical investing. This also illustrates that the ideas in this thesis have wider application. Equities (stocks and shares) are fractional ownerships of businesses and are traded in the stock market. Ownership consists most importantly of voting rights to appoint directors, who make key business decisions. Equities are issued by businesses as a way to raise capital. Socially responsible investing is, most often, the practice of not buying the equities of businesses that are involved in unethical activities. This is called negative screening. The definition of ‘unethical activity’ varies, but generally involves activities that are perceived to lead to significant harm e.g. arms manufacture; gambling and tobacco 50 51 As reported in The Independent (Taylor & Porter, 2011) As taken from the archived SRI campaign website: (SRICampaign) 88 companies. Ethical investing can also involve positive screening (investing in companies perceived to have positive social return) and shareholder activism (attempting to use voting rights to positively shape company behaviour). The SRIR Committee is open to activism, but so far the only action it has recommended is that Oxford sells its investments in arms companies.52 7.3.2. Supply-Demand Effects One effect of being willing to buy the equities of arms companies is to increase demand for these equities. The market for equities, however, is not like the labour market. According to the Efficient Market Hypothesis (EMH), equity prices are at in equilibrium about a fair value. So, ignoring transaction costs, equity owners are indifferent to holding the equity. This means that by purchasing the equity, one has no effect on its price. The EMH is hotly debated53. An alternative model is that each equity holder has a different price at which they are willing to sell. If one wants to purchase an equity, one must bid up the price until enough people are willing to sell. So, purchasing an equity raises the price. In reality, small marginal buyers have approximately no effect on prices. They are known as price-takers. As the size of one’s purchase grows, one has more and more impact on prices. 52 53 (SRIR-Committee, 2010) For a summary of critical views, see (Malkiel, 2003) 89 Oxford was approximately a price-taker in the context of its purchase of £630,000 of Northrop Grumman equity, which amounted to less than 0.01%54 of the total quantity available. Supposing Oxford could increase the market price of Northrop equity, it’s not clear this effect would last. Only 12% of the U.S. equity market is ethically invested55, so the remaining 88% would buy additional units of Northrop equity, profiting from the difference between the new price and fundamental value. A sustainable price decrease could only happen if so many people boycott the stock that it exhausts the capacity of non-ethically motivated arbitrageurs. Let’s suppose Oxford did increase the price, what effect would that have? The main mechanism by which it could lead to more harm is that higher equity prices are taken as an indication that if the company were to issue more equity, it would fetch higher prices. This makes the company more credit-worthy, and so better able to grow its business. Even then, if Northrop is made more successful at the expense of another cluster bomb company (a compensation effect), it’s not clear that more cluster bombs will be created or used. However, we expect that on average the cluster bomb industry is made a little more credit-worthy, which could help it to grow. So, there is potentially a small increase in harm. 54 55 Based on a market capitalisation of $15bn (USSIF, 2010) 90 7.3.3. The Costs of Negative Screening Prima facie, negative screening reduces Oxford’s investment opportunity set, reducing its scope for investment performance. Whether negative screening does in fact reduce performance is a subject of intense debate.56 If it does, then since Oxford uses money with the ethical aim of educating students and producing research, it has a negative effect. This is analogous to professional philanthropists refraining from taking high earning but unethical jobs. 7.3.4. Boycott vs. Control If a very large number of people refused to buy Northrop stock, it would force the price down, making the company less credit-worthy. If a significant number of ethically minded people bought the stock, however, they might be able to use their voting rights to positively influence investor behaviour. The analogy to owning the equity and engaging in shareholder activism is taking the job and seeking to make it less harmful. For someone seeking to have a ‘practical positive impact’57, the question is whether it’s easier to have impact through control or through seriously hampering its activities through a boycott. This will vary from situation to situation. I’d suggest the control option is often easier. One achieves some influence immediately and a block of ethical investors can elect a director with only 20% of equity. The coordination required for such actions is more complex than boycotting the stock, but still simple. Oxford could publish lists of the companies it invests in, requesting help from other 56 57 For instance, this study argues that it has no effect: (S. & Walkshausl, 2011) (SRICampaign) 91 ethically minded investors. In a market where only 12% is ethically invested58, it will be difficult to hamper a company to any degree through boycott. In most cases, negative screening means passing up the opportunities to change unethical companies, and instead buying companies whose managements are already socially responsible. 7.3.5. Signalling Effects Signalling effects seem to be the most significant negatives. In particular, it demonstrates Oxford is not boycotting the stock, making a boycott less likely. However, in the light of the previous subsection, it might be better for Oxford to seek to influence harmful companies anyway. Oxford could also declare that it would boycott the stock if a significant proportion of the market joined in. Buying the stock could also have a Political Signalling Effect – it shows that Oxford supports arms companies. As with all signalling effects, its potency depends on how well publicised it is and on how influential Oxford is on other investor (relative to the person who would otherwise hold the equity). It seems Oxford could largely negate the signalling effect by making it clear that while it will invest in any listed company, it doesn’t condone the manufacture of cluster bombs, and that in fact it wants to exert an ethical influence on Northrop through shareholder activism. 58 (USSIF, 2010) 92 7.3.6. Conclusion Replaceability in investing is similarly complex – it’s not clear that buying the equities of harmful companies causes harm, just as it’s not clear that taking harmful jobs causes harm. There’s reason to think Oxford should actively seek to buy the equity of unethical companies. The importance of each effect is different from career choice. Supply-demand effects are close to zero. The important effects revolve around the effect on Oxford’s investment returns, Political Signalling, the chance of a boycott, and the potential gains of control. 7.4. Conclusion The Supermarket Meat (and plane ticket) argument turns out to be misguided due to the large supply-demand effect. It’s not a good analogy, since there are no job-related c-consequences and the supply-demand effects are different. Eating the pepperoni pizza, however, could be the right thing to do. It depends on the details of the other effects. The analogy with ethical investing is closer. Rather than suggesting that replaceability in career choice is a bad argument, it suggests that we should rethink the practice of negative screening. The following table summarises the cases. 93 Application Job-related Consequences Supply-Demand Precedent Effect/Chance of Boycott Depends on R Influences jobcreation and wages Small negative, depends on publicity and stated intentions N/A Increases price, causing extra meat supply N/A Potential benefits of control Choosing a harmful career Buying meat at a supermarket Negative or positive, depending on stated intentions. Corruption Risk Integrity Harm Permissible? Depends on person Depends on personal projects Potentially, if R<0 Small negative, Small negative depends on whether friends are influenced by your choices Ditto Presumably violated No No effect Ditto Negative or positive, depending on stated intentions. Ditto Depends on personal projects No harm (except potential indirect effects) No effect if a price-taker, but potentially increases equity prices Small negative, depends on publicity and stated intentions Ditto Unlikely Depends on projects if relevant at all Often no harm. Permissible if use control, or higher investing returns used ethically Eating meat at a party Buying the stock of a harmful company Political Signalling 94 8. Conclusion In seeking an ethical career, it’s not enough just to find a career directly doing good. In Part 1, I showed that what one does directly gives little indication of the cconsequences of typical career decisions. Understanding these c-consequences is vitally important. The activities of many careers produce value on the scale of saving hundreds of lives, so taking one job over another can be a matter of life or death for hundreds of people. I introduced a Simple Analysis of the counterfactuals needed to estimate the c-consequences, but showed that this must be replaced with the Sophisticated Analysis. Applying the Sophisticated Analysis showed that the c-consequences of typical career decisions range from none of the good one does directly, to all of it. Taking a career that directly does good can even have bad-consequences; while taking harmful careers can make a positive difference. Failing to take account of replaceability, therefore, means being radically misinformed about the cconsequences of one’s career decisions, possibly to the detriment of hundreds of people. We can all agree that it’s morally relevant if one career saves more lives than another. But in Part 2, I argued that the job-related c-consequences of career decisions for the ethically motivated are not only morally relevant, but typically determine which option is ethically preferable: the Dominance Thesis. The other c-consequences of career decisions are often small for any career. For someone pursuing an ethical career, they are likely to be outweighed. Similarly, non-consequentialist factors often 95 don’t apply or are outweighed when they do. Replaceability in career choice is not analogous to other arguments that sound similar, like ‘if I don’t buy this meat, someone else will.’ This means that an assessment of the c-consequences of different career options, as determined by replaceability, is enough to work out which option an ethically minded person should take. So, replaceability is a crucial consideration in choosing an ethical career. Despite this, it’s rarely taken into account in ‘ethical careers’ advice. Each year, thousands of people do not make the most of their ethical career due to a lack of information about and understanding of replaceability. We need a program of empirical research, aiming to apply the analysis developed in §4. For each industry, we need to know about the typical patterns in ability* and opportunity*, whether entering the industry leads to job-creation, how labour flows between this industry and others, and so on. The extent to which reallocation of tasks can occur within an industry is vital. It’s also important to investigate the extent to which the other tools of economics can be applied to the problem. This will enable us to start to work out the areas where ethically minded people can have the most impact. More philosophical work could also be done considering demandingness; as well as in applying the latest in the ethics of the permissibility of causing harms to career choice. 96 There are signs of interest in these issues. I hope this interest will continue to grow. Rather than settle for making a difference with our careers, let’s aim to make the most difference. 97 Acknowledgements I’d first like to thank Toby Ord, who was a great supervisor and put in many hours of work reading and discussing early drafts. He also came up with some of the important concepts, and was the first person to tell me about the Iteration Effect. Next, I’d like to thank Will Crouch, who helped me to develop the initial plan for the thesis, did some early work on the Sophisticated Analysis of Replaceability and provided very helpful comments on the penultimate draft. I’d also like to thank Brian Tomasik, the first person to write about replaceability in ethical career choice, who commented on an early version of several sections. Mike Webb and Ronan McDonald also provided detailed comments on two sub-sections. Thanks to Mark Lee and Andreas Mogensen for reading several sections. I’m also grateful to Seb Farquar and Owain Evans for proof reading several sections of my final draft. Any mistakes remain my own. Finally, I’d like to thank all my friends and fellow volunteers at 80,000 Hours, who showed interest in replaceability, and helped me to see where the important issues might lie; as well as my friends who put up with me constantly talking about writing my thesis. 98 Bibliography Ashford, E. (2000). Utilitarianism, Integrity and Partiality. The Journal of Philosophy XCVII(8), 421-439. Burrows, G. (2006). Ethical Careers Guide. New Internationalist Publications Ltd. Buss, S. (2006). Needs (Someone Else's), Projects (My Own), and Reasons. Journal of Philosophy CIII no.8, 373-402. Care, N. S. (1984). Career Choice. Ethics 94, no. 2, 283-302. Crouch, W. (Unpublished). The Ethics of Career Choice. Crowe, D. M. (2007). Oskar Schindler: The Untold Account of His Life, Wartime Activities, and the True Story Behind the List. Basic Books. Easterbrook, G. (1997, Jan). Forgotten Benefactor of Humanity. The Atlantic Magazine. Evans, R., & Leigh, D. (2010, Dec). Judge 'astonished' by corruption denials as he fines BAE £500,000. Evers, M., & R. de Mooji, D. v. (2005). What Explains the Variation in Estimates of Labour Supply Elasticities? CESIFO Working Paper No. 1633. Foot, P. (1967). The Problem of Abortion and the Doctrine of Double Effect. Oxford Review, Number 5. Granovetter, M. (1978). Threshold Models of Collective Behaviour. American Journal of Sociology, Vol. 83, No. 6, pp. 1420-1443. Gupta, H. M., Campanha, J. R., & Pesce, R. G. (2005). Power-law distributions for the citation index of scientific publications and scientists. Braz. J. Phys., vol.35 no. 4a. Hamermesh, D. S. (1996). Labour Demand. Princeton University Press. Hare, C. (2011). Obligation and Regret When Tehre is No Fact of the Matter About What Would Have Happened if You Had not Done What You Did. Noûs, 190206. HESA. (2011). High Education Statistics Agency, Performance Indicators Index, Table E1a 2009/10, available www.hesa.ac.uk. Hursthouse, R. (2012). Virtue Ethics. The Stanford Encyclopedia of Philosophy. 99 Jamison, Breman, Measham, Alleyne, Claeson, Evans, et al. (2006). Disease Control Priorities in Developing Countries 2nd Ed. OU Press and The World Bank. K. D. Vohs, N. L. (2006). The Psychological Consequences of Money. Science 314, 1154-6. Kagan, S. (2011). Do I Make a Difference? Philosophy & Public Affairs, 39, no. 2. Kamm, F. (2007). Intricate Ethics: Rights, Responsibilities, and Permissible Harm. Oxford University Press. Langemeier, L., & Thompson, R. G. (1967). Demand, Supply and Price Relationships for the Beef Sector, Post-World War II Period. Am. J. Agr. Econ. 49, (1 Part I): 169-183. Lenman, J. (2000). Consequentialism and Cluelessness. Philosophy & Public Affairs Vol. 29, No. 4, 342-370. Lenman, J. (2007). Why I have no plans to retire: In defence of moderate professional complacency. RAtio 20, no. 2, 241-246. Lewis, D. (1973). Causation. The Journal of Philosophy Vol. 70, pp. 556-567. Malkiel, B. G. (2003). The Efficient Market Hypothesis and Its Critics. CEPS Working Paper No. 91. McCabe, C., Claxton, K., & A, J. C. (2008). The NICE Cost-Effectiveness Threshold: What it is and What that Means. Pharmoeconomics, 26 (9):733-744. NEF. (Dec 2009). A Bit Rich: Calculating the Real Value to Society of Different Professions. Philippon, T., & Reschef, A. (2009). Wages and Human Capital in the U.S. Financial Industry: 1909-2006. NBER Working Paper No. 14644. Preito, L., & Sacirstan, J. A. (2003). Problems and solutions in calculating qualityadjusted life yaers (QALYs). Health and Quality of Life Outcomes, 1:80. Robinson, O. (2006, July). Planning for a Fairer Future. The Guardian. S., L., & Walkshausl, C. (2011). Vice vs. Virtue Investing Around the World. SSRN. SRICampaign. (n.d.). www.timdavies.org.uk/srioxford/home.asp.html. SRIR-Committee. (2010). Recommendation to Council on Divestment from Arms Manufacturing Companies. Taleb, N. (2007). Fooled by Randomness: The Hidden Role of Chance in Life and in the Markets. Penguin. 100 Taylor, J., & Porter, L. (2011, Oct 17). Oxford University invests £630,000 in US firm that profits from cluster bombs. The Independent. Thomson, J. J. (1985). The Trolley Problem. 94 Yale Law Journal, 1395-1415. Thomson, J. J. (1986). Killing, Letting Die, and the Trolley Problem. In W. Parent, Rights, Restitution, and Risk: Essays in Moral Theory. Cambridge: Harvard University Press. Tomasik, B. (2007, September 20). Making money and making change not mutually exclusive. The Pheonix. USSIF. (2010). 2010 Report on Socially Responsible Investing Trends in the United States. Williams, B., & Smart, J. (1973). Utilitarianism: For and Against. CU Press. Woodard, C. (2009). Pedro's Significance. The Southern Journal of Philosophy, Volume 47, Issue 3, 301-319. 101