Anonymous Assertion

“Anonymous Assertions”1 Sanford C Goldberg Department of Philosophy, Northwestern University 1880 Campus Drive Evanston, IL 60208 USA [email protected] ABSTRACT This paper addresses how the anonymity of an assertion affects the epistemological dimension of their production by speakers, and their reception by hearers. After arguing that anonymity does have implications in both respects, I go on to argue that at least some of these implications derive from a warranted diminishment in speakers’ and hearers’ expectations of one another when there are few mechanisms for enforcing the responsibilities attendant to speech. As a result, I argue, anonymous assertions do not carry the same “promise” of the speaker’s relevant epistemic authoritativeness that ordinary assertions do. If this is correct, the phenomenon of anonymity provides us with a lesson regarding ordinary assertions: their aptness for engendering belief in others, and so for communicating knowledge, depends in general on the very publicness of the act of assertion itself. 1. Background Many commentators have noted the poor quality of much of the commentary to be found on the internet (web sites, blog posts and comments, etc.). It is perhaps not particularly controversial to think that the anonymity that the online world provides is partially responsible for the lamentable quality here. If you can say something under conditions in which no one knows who is saying so, then your tongue might be a bit freer in what you say. No doubt, this phenomenon is not new. (Consider the venerable tradition of bathroom-wall graffiti.) But given the prevalence of the internet in our lives, the role web sites and blogs have come to play in many of our discussions, and the fact that most sites are accessible to 1 An earlier version of this paper was presented at Epistemological Problems of Privacy and Secrecy Conference, Delft University of Technology, the Netherlands, June 8-10, 2012. I would like to thank the various participants for comments made in conversations and during presentations. Special thanks are due to Martijn Blauw, Jeroen de Ridder, Don Fallis, Alvin Goldman, Axel Gelfert, John Hardwig, Klemens Kappel, Peter Ludlow, David Matheson, Pierre le Morvan, Boaz Miller, Seamus Miller, Philip Nickel, and Rik Peels. I would also like to thank two anonymous (!) referees from this journal for their helpful and encouraging comments. 1 everyone in the world, it is plausible to think that the phenomenon of anonymous messages and assertions is much more widespread these days. My aim in this paper is to try both to characterize, and to offer a partial explanation of, the link between anonymity and the epistemic quality of assertions. To do so, I borrow terms from philosophy of language (the norm of assertion) and from epistemology (the epistemology of testimony). My claims will be these. First, the fact of anonymity – the fact that one’s assertion was anonymous – has implications for the warranted expectations that speakers have of their audience, and that the audience has regarding the speaker. In general, when it is mutually known by all parties that a claim was made under conditions of anonymity, this has a diminishing affect on the sort of (assertion-generated) expectations that speakers and hearers are entitled to have of one another. Second, this diminishment of warranted expectations under conditions of anonymity can be explained by appeal to the hypothesis that assertion has an epistemic norm. In particular, assuming that assertion has an epistemic norm, and that this is something that is familiar to all competent speakers of a language, we would predict that some of the expectations generated by the making of an assertion will have an epistemic content: they are expectations pertaining to the epistemic position of the speaker. However, when assertions are anonymous, most or all of the mechanisms that exist for the “policing” of assertions – for ensuring that speakers are living up to the responsibilities associated with asserting – do not apply, and both the speaker and hearer are aware of this (if only implicitly). It is for this reason, I will argue, that we see a diminishment in the warranted expectations 2 that hearers and audiences have towards one another, when an assertion is made anonymously. These claims, I think, point to one cost associated with the privacy or secrecy that the online world affords us. In particular, insofar as the internet provides the opportunity to make claims while keeping one’s identity a secret, this “opportunity” militates against the sort of conditions on which quality communicative exchanges appear to depend. 2. Anonymity Throughout this paper I will be speaking of anonymity, and anonymous assertion. My general approach to anonymity is epistemic. To a very rough first approximation, a speaker or writer S is anonymous relative to a group of people G in a given (spoken or written) assertion S makes when no one in G can knowledgeably connect S to the assertion – more colloquially, when no one in G knows who produced the assertion. I acknowledge that the locution ‘knows who’ is a tricky one, and that as a result my characterization will inherit all of this trickiness. But to a first approximation we can say that no one in G knows who the speaker or writer is, in the sense relevant to anonymity, when no one in G has identificatory knowledge of the speaker or writer that would enable the re-identification of S as the speaker or writer of the message. This first approximation requires immediate modification, in that it is too weak: it allows as cases of anonymity cases we should want to rule out. It is easy to imagine cases in which, as a matter of fact, no one in a given group knows who the speaker or writer is, but not because the speaker or writer hid her identity in any way. Imagine that Sally writes a book, publishes it herself (her own vanity press), puts her name on the cover (she is hiding nothing), but as a matter of fact everyone 3 who reads the book fails to observe the cover, and so as a matter of fact no one among the readers of the book (presently) knows who wrote the book. This is hardly a case of anonymity, yet is satisfies the characterization above. To rectify this situation, I would propose the following slight modification. (I do not pretend that the proposed characterization of anonymity is counterexample- proof, but it will be good enough to use for the purpose of this paper.) The characterization is this: ANON Given a speaker/writer S who produces a message m, S is anonymous (with respect to m) relative to an audience A when (i) no one among the members of A is currently in a position to acquire knowledge to the effect that S produced m, (ii) it is reasonable (for S) to expect that no one among the members of A will come to know something to this effect, and (iii) the explanation for both of the foregoing is that S’s identity, qua producer of m, is in some way “hidden” from the members of A. Note that ANON offers only a sufficient condition, not a necessary condition, on S’s being anonymous (relative to A) with respect to m. ANON calls for some commentary. First, instead of characterizing anonymity in terms of a lack of “knowing who”, ANON’s condition (i) characterizes it in terms of a lack of knowledge “to the effect that S produced m.” I intend for the latter as a stand-in for the former. (Clearly, S is not anonymous merely in virtue of people failing to know her as S, since they may be able to identify her in other relevant ways. Which ways of identification are relevant to anonymity? That anyone can denote her as “the writer of this message” does not suffice to show S is not anonymous. But this is a hard question to which I do not have anything more to say.) Second, the reason behind ANON’s (ii) is this: an author S who posts under a pseudonym, under conditions in which only S herself (and perhaps her internet provider) knows of S that she is the one posting under this pseudonym, can be said to be posting anonymously (to those in the relevant audience), even though there 4 might be ingenious hackers who, given enough time and resources, could come to know, of S, that she is the one posting under this pseudonym. The point about anonymity here is simply that it was reasonable for S to assume that no one among her audience will come to acquire this knowledge. Third, the idea behind ANON’s (iii) is this: the facts (i) that no one among those in A is in a position to acquire knowledge to the effect that S produced m, and (ii) that it is reasonable for S to suppose that no one among A will acquire such knowledge, are themselves to be explained by the further fact that S’s identity is in some way “hidden” from the members of A. This is to rule out a case like the one just mentioned, where condition (ii) is satisfied in virtue of Sally’s recognition that the readers of her book are systematically uninterested in learning who wrote the book. In such a case, Sally is not anonymous even relative to a readership that is systematically uninterested in (and so ignorant of) her identity as author, even though ANON’s conditions (i) and (ii) are satisfied. Again, I cannot claim that ANON is counterexample-proof; I highly doubt that it is. Still, it is good enough to proceed for my purposes at hand: to explore the epistemic implications of anonymity in assertion. (My hope would be that whatever rectifications are needed to render ANON adequate, these could be made in such a way as to not undermine any of the arguments I am going on to be presenting.) 3. Assertion’s norm and mutual expectations My discussion of the epistemic implications of anonymity in assertion will be shaped by three assumptions regarding the speech act of assertion. (These assumptions are widely – though not universally – endorsed in the literature on assertion.) First, as a type of speech act, assertion answers to an epistemic norm of some sort or other. Second, this feature of the speech act of assertion is 5 common knowledge to all competent speakers of a language. And third, competent speakers of a language can readily and reliably discern when an observed speech act (involving a sentence in a language they understand) is an assertion. In order to simplify matters, I will formulate these three assumptions in one grand assumption. Let us designate this the ‘Common Knowledge regarding the Norm of Assertion’ assumption, as follows: CKNA It is common knowledge that assertion is a type of speech act that is both (i) readily-discernible and (ii) answerable to an epistemic norm of some sort or other. In what follows I do not want to assume anything about what the norm in (ii) requires. This is an issue that has generated a good deal of controversy in the literature. Some say that it requires knowledge, others that it requires adequate evidence, still others that it requires that the hearer occupy some other epistemic position regarding the proposition asserted. 2 I don’t think I have to enter into this controversy in order to make the points I wish to make. For whatever you think the norm of assertion is – whether it is as demanding as knowledge, or something that requires only an adequate degree of evidence – you will agree with the following (at least insofar as you agree that it has an epistemic norm in the first place): in asserting something, the speaker performs an act regarding which it is common knowledge that her act was proper (warranted) only if she had the relevant 2 What I am assuming in connection with CKNA, is a widely-endorsed, but not entirely uncontroversial, set of three claims: (1) as a speech act, assertion is governed by a norm; (2) this norm is epistemic, in the sense that the standard it imposes on appropriate acts of this kind is an epistemic standard; and (3) something in the vicinity of these first two claims are common knowledge among competent speakers. Some reject (1) (see Pagin (2011) for a discussion); some accept (1) but reject (2) (see Weiner (2005), and perhaps Bach and Harnisch (1979)). Among those who accept either or both of (1) and (2), something like (3) usually comes along for the ride (see Goldberg (2011) for an explanation). Still, (1)-(3) form a triad of claims that are accepted by a good majority of those contemporary philosophers who write on assertion. For the variety of approaches that have been taken regarding the nature of assertion, see MacFarlane (2011). 6 epistemic authority. It is this piece of common knowledge, I want to argue, that provides the content of the sorts of expectations that characterize ordinary face-to- face communicative exchanges. (We should be able to agree on this much even if we disagree over what the precise content of the norm is.) The point I wish to make is this: if CKNA is true, we would predict that the practice of assertion is rich with mutual expectations between a speaker and her audience.3 These expectations are grounded in common knowledge of the fact that assertion has an epistemic norm. In what remains of this section I want to develop this point; after doing so I will go on to consider how conditions of anonymity might affect these expectations. Let me begin with the mutual expectations that arise in connection with assertion. Suppose – as follows from CKNA – that in asserting something, the speaker performs an act regarding which it is common knowledge that her act was proper (warranted) only if she had the relevant epistemic authority. Since this is common knowledge, both the speaker and the audience have this knowledge, and both are aware that the other has it as well. 4 The content of this knowledge might be put as a disjunction: either the speaker has the relevant epistemic authority or else the assertion was unwarranted. Of course, ordinarily no speaker who makes an 3 In this paper I will use ‘speaker’ to designate the producer of an assertion, whether spoken or written. Context will make it clear whether the case is one involving a speech act or a written communication. 4 Care must be made in understanding this claim. Some assertions are made under conditions in which the speaker does not know who the audience is (in some clear sense of ‘knows who’). In such cases, there is no audience member a, such that the speaker knows of a that a has the relevant knowledge. Still, the speaker knows that insofar as there is a (linguistically competent) audience, then, whomever is among such an audience, he or she has the relevant knowledge. Of course, if the audience does not know who the speaker was, then we will have to make a similar move: the audience will then know that the speaker, whomever he or she was, has the relevant knowledge. I will return to these more abstract sorts of knowledge when I consider the effects of anonymity on the mutual expectations that hold between speaker and audience. 7 assertion does so intending to get the hearer to believe that disjunction. Rather, she aims to get the hearer to believe what she asserted. For this reason we might speak of the act of asserting that p as conveying that the speaker does, in fact, possess the relevant epistemic authority with respect to the truth of the proposition that p.5 (I will sometimes speak of the “promise” of authoritativeness that an assertion conveys.) This would rationalize the speaker’s hope or expectation of being believed: she is presenting the proposition that p as true, under conditions in which she is presenting herself as epistemically authoritative on the matter. We are now in a position to understand two central features of assertoric practice. The first pertains to the audience. Having observed a speaker making an assertion, an audience sometimes responds by querying how the speaker knows, or what her evidence is.6 We can describe such a case as one in which the hearer is requesting the speaker to vindicate the conveyed claim of relevant epistemic authority.7 The fact that such a response is ordinary highlights one class of expectations in play when assertions are made: hearers expect speakers who make an assertion to recognize that they are under some obligation to vindicate their relevant epistemic authority, if called upon to do so. But there is a second feature of our assertoric practice, corresponding to this first feature, albeit pertaining to the speaker. It comes as no surprise to a speaker if, having made an assertion, she elicits one of these responses in a hearer. What is more, the speaker herself will regard herself as under some obligation to respond: it would be curious indeed for a 5 By ‘conveying’ I have in mind something like ‘communicating, as part of the conventional significance of the illocutionary force of the speech act.’ Clearly, what is conveyed in this sense is not part of the content asserted. 6 This aspect of assertoric practice has been noted by many; perhaps the most salient recent example is Williamson (1996). 7 As shorthand I will sometimes speak of the speaker ‘vindicating her relevant epistemic authority.’ 8 speaker to assert something, to be queried regarding how she knows, and to respond with a shrug and then simply proceed as if nothing were remiss. (Stranger still if the speaker responded by manifesting her impression that the audience’s query was inappropriate!) This is not our practice. And audiences know this. Again, this is merely to say that these expectations that speakers and audiences have of one another are mutually known to both. All of this is explicable in terms of CKNA.8 I have just suggested that the making of an assertion generates certain expectations in the mind of the speaker and the audience alike. Both know, and know that the other knows, that the assertion was warranted only if the speaker had the relevant epistemic authority. Consequently, the audience will expect that the speaker will recognize the propriety of the audience’s querying the speaker’s authority; and the speaker, expecting that the audience will recognize the propriety of doing so, regards herself as under some obligation to respond to such a query. For this reason, we can describe these mutual expectations using the language of entitlements and responsibilities. Since a speaker who makes an assertion does something regarding which it is common knowledge that what she did is proper only if she had the relevant epistemic authority, at a minimum the hearer is entitled to believe that the speaker acknowledges that she has the responsibility to possess the relevant epistemic authority. The propriety of the language of entitlements and responsibilities derives from the norm (or rule) governing assertion. Some philosophers hold that the hearer’s entitlement on this score goes beyond the minimal entitlement I have described. 9 In particular, these philosophers 8 Points in this vicinity are highlighted in Benton (2011), which came to my attention after I finished a draft of this paper. 9 See e.g. Ross (1987) and Burge (1993). I should say that I agree, albeit for different reasons than Ross and Burge offer; see Goldberg (2007), Goldberg and Henderson (2006). 9 think that the audience who observes an assertion is entitled to believe that the speaker actually possesses the relevant authority (not merely that she acknowledges the responsibility to do so). To be sure, those philosophers who regard the entitlement as having this enhanced content acknowledge that the entitlement itself is defeasible: a speaker who is clearly incompetent does not entitle an audience to believe that she (the speaker) possesses the relevant authority. Still, these philosophers hold that the entitlement in question is a presumptive though defeasible entitlement to regard the speaker as relevantly authoritative. Such appears to be the view of Angus Ross. Speaking of the rule-governed nature of language use (of which the norm of assertion is a part), Ross writes, It is a quite general feature of rule-governed life that the responsibility for ensuring that one’s actions conform to the rules lies primarily with oneself and that others are in consequence entitled to assume, in the absence of definite reasons for supposing otherwise, that one’s actions do so conform. Thus where the rules are such that one may perform a certain action only if a certain condition obtains, … then to perform the action is to entitle witnesses to assume that the corresponding condition obtains. If that assumption proves false and others act upon it with unfortunate consequences, at least part of the responsibility will lie with oneself for having entitled them to make that assumption. (Ross 1986: 77-78) Ross’s view gives rise to the following picture. Given CKNA, there is a rule governing assertion, to the effect that one must not assert that p, unless one has the relevant epistemic authority – that is, unless one is epistemically authoritative regarding the truth of the proposition that p. So if Ross is right, one who asserts that p entitles others to assume that one is epistemically authoritative regarding the truth of the proposition that p. Now the question before these two views, over what an audience is entitled to assume on observing an assertion, is actually at the heart of the central debate However, I do not want anything I say in this paper to depend on this stronger claim about what audiences are entitled to believe on observing an assertion. 10 in the epistemology of testimony. That debate concerns the conditions under which an audience is justified in accepting what she is told. Both sides can agree that a hearer A is justified in accepting what she is told only if A is entitled to regard the speaker as having been relevantly epistemically authoritative; where the sides disagree is over the conditions on being so entitled. Happily, we need not enter into this debate here, since the point I wish to make at present can be made whatever one’s views on the questions of entitlement and justification. The point I wish to make here is this: there is a gap that needs to be bridged, from what an audience is entitled to believe on observing an assertion, to an audience’s being entitled to believe (being justified in believing) what was asserted. The existence of such a gap is clear on the more minimal account I described above; but it is also clear on Ross’s account. According to the more minimal account I described above, on observing an assertion that p, an audience is entitled to believe that the speaker recognizes that she is under an obligation to have the relevant authority regarding the truth of [p]. Clearly, such an entitlement is not sufficient to justify the audience in believing that p (on the basis of the speaker’s having said so). The audience will also need to be justified in thinking that the speaker has the relevant authority; for it is only then that the audience will be justified in accepting that p (on the basis of the speaker’s having said so). I take this as obvious. Next, consider Ross’ account. On observing an assertion that p, an audience is (presumptively but defeasibly) entitled to believe that the speaker has the relevant authority regarding the truth of [p]. Still, this entitlement is only presumptive; in order for it to hold, it must be the case that there are no relevant defeaters. From this we see that the absence of defeaters is a further condition on 11 the audience’s being entitled (full-stop) to regard the speaker as having been relevantly epistemically authoritative, and so is a further condition on the audience’s being justified in believing that p (on the basis of the speaker’s having said so). Nor is this the only further condition. In particular, the mere lack of defeaters does not suffice to turn the presumptive entitlement (to regard the speaker as authoritative) into an entitlement full-stop, and so does not suffice to render the audience justified in believing that p (on the basis of the speaker’s having said so). For suppose we have an audience who suffers from extreme gullibility: he would have been oblivious to the presence of defeaters had they obtained, and so would have believed whatever he was told. The mere fact that there were no defeaters – the speaker spoke in a way that would have struck an ordinary person as competent and sincere, her claim would have struck an ordinary person as not implausible on its face – does not seem sufficient, by itself, to entitle the audience to regard the speaker as relevantly authoritative. 10 Now it might be too much to require hearers to have positive evidence that there were no defeaters, in order for them to enjoy an entitlement to regard the speaker as relevantly authoritative. Even so, the following seems reasonable: a hearer is not entitled to regard the speaker in this way unless the hearer herself is counterfactually sensitive to the presence of defeaters, such that, had there been any relevant defeaters, she would have attended to them, and so would have refrained from so regarding the 11 speaker. In this way we see that, whether we accept the minimal account of the hearer’s entitlement (described above), or Ross’s account, the following point holds: 10 This is a point that Elizabeth Fricker has made; see Fricker (1994). 11 For a defense of this sort of view within the framework of an anti-reductionist account in the epistemology of testimony, see Goldberg and Henderson (2006). 12 what an audience is entitled to believe on observing an assertion, by itself, does not suffice to justify the audience’s belief in what was asserted. This is what I mean when I say that there is a gap that needs to be bridged, from the former to the latter. The existence of this gap is important. I am now going to argue that, when we consider how this gap is actually spanned, we will be appealing to considerations that are (typically) largely absent in cases in which an assertion is made anonymously. 4. The Epistemology of Assertion The question before us is this: how is it that, from what you are entitled to assume merely in virtue of having observed someone else make an assertion you understood, you come to be justified in accepting the assertion? Here I go over various of the things that bear on this question. (I will return, briefly, to the debate between the minimalist view and Ross’s view of assertion-based entitlements.) This will pave the way for the next section, where I argue that many – though not all! – of these things are absent in cases in which an assertion is made anonymously. First, some background. As consumers of others’ speech, audiences have various ways of assessing the credibility of an assertion. 12 Some of these pertain to the audience’s assessment of the speaker, others pertain to the audience’s assessment of the act of assertion itself, still others pertain to the plausibility of the content asserted, and yet still others pertain to the context of the communication. What is more, many of these assessments are broadly subcognitive, 13 while some 12 Or perhaps we should say the cognitive systems of audiences have various ways of assessing the credibility of an assertion. 13 To the extent that we rely on subcognitive processes in our credibility judgments, there are real questions regarding how reliable those processes are; social psychology appears full of evidence that women and members of minority groups have their credibility systematically discounted. In this connection, see Fricker (2007) for a very robust defense of the idea that this constitutes an epistemic injustice to those who are the victims of this 13 others are the result of conscious deliberation. In what follows I will use “assessment” broadly, to cover any of the (subcognitive or cognitive) resources that a subject (or her cognitive system) brings to the task of discerning the credibility of an assertion. Consider first the audience’s assessment of the speaker and her assertion. The audience will typically assess such things as how sincere and confident the speaker seemed in the assertion; how well-placed she was to have the knowledge in question, relative to what she is presently asserting (including her relevant background, if it is known by the audience); and what sort of motives she might have had in speaking as she did. In the latter respect, the audience might assess such things as whether the speaker had any present motive for insincerity, as well as what “pressures” there are on the speaker to speak responsibly. Since the topic of the “pressures” bearing on a speaker to speak responsibly will loom large in my subsequent discussion, I want to dwell on this topic a bit here. These “pressures” derive from what I will describe as the mechanisms for the “policing” of assertoric responsibility. By this I include all of the practices and institutions that encourage speakers to live up to their assertoric responsibilities, and/or enforce the repercussions for those who do not do so. These “pressures” can take many forms, but I will focus primarily on those that involve negative consequences the speaker would (likely) face if she is found to be speaking irresponsibly.14 Such pressure on a speaker increases in proportion to (a) the severity of the consequence, as well as (b) the chances of her being “found out.” Of prejudice-based diminution in the credibility their words are assigned. Still, these are among resources that audiences bring to bear in assessing testimonial credibility. 14 Let us understand speaking irresponsibly as amounting to violating the norm of assertion. One can quibble with this identification – if the norm is knowledge, it would seem that not all forms of violating this norm are cases of irresponsible assertion (consider Gettier cases) – but these differences, though real, are not germane to the present discussion. 14 course, these pressures work only if the speaker herself is broadly knowledgeable of (a) and (b); and so whatever knowledge an audience has of these pressures on the speaker are relevant to the assessment at hand only insofar as the audience regards the speaker as aware of these pressures, and as sensitive to them. Considerations of this “outside pressure” sort are particularly important when it comes to an audience’s assessment of written testimony. You don’t know the journalist who wrote the article you just read in the NY Times; yet you probably regard her, if only implicitly, as highly reliable. This sense of yours is backed, at least in large part, by your (perhaps only implicit) awareness that the NY Times has a reputation to protect, and that it will go to great lengths to do so, especially when it comes to hiring and exerting editorial control over their journalists. (It is because these practices enforce a higher degree of reliability among its writers that the NY Times has a greater claim on our attention, epistemically speaking, than does the National Enquirer.) There can be little doubt but that a system of very extensive social relations, perhaps together with our (merely implicit) awareness of these relations, warrants the credibility we assign to a good deal of the written reports we encounter. Of course, professional sanctions are not the only form of potential pressure that is brought to bear on speakers. There are all sorts of other pressures that can be brought to bear against an irresponsible asserter. These include a diminishment in the trust such a speaker is accorded, the consequent diminishment in the roles that are open to her to play in the deliberations of the various communities of which she is a member, other forms of loss of status (and perhaps loss of friends and partners), and the moral disapproval of one’s peers. I just spoke of the audience’s assessment of the speaker and her assertion (with particular attention to the pressures that can be brought to bear on the 15 speaker for asserting irresponsibly). Moving on, let us consider next the audience’s assessment of the plausibility of the content of the assertion. Obviously, the audience is better positioned on this dimension to the extent that he himself has ample background knowledge on which to draw. If you are an expert in radiation, then you are in a good position to assess the plausibility of the latest claim made regarding radiation levels near various sites in Japan; if you are not an expert, you are more helpless in this regard. Still, it would seem that common sense itself often suffices to enable an audience to make at least some assessment of the plausibility of an asserted content. And this, together with other background knowledge that the hearer brings to bear in the assessment, is what he has to go on. The purpose of these sorts of assessment is to bridge the gap I identified above. Let us use ‘E’ to designate the proposition that, according to one’s favored theory, the hearer is (perhaps merely presumptively) entitled to believe merely in virtue of observing the speaker’s assertion; 15 and let us use ‘P’ to designate the proposition asserted by the speaker. The gap, then, is that between the audience’s (perhaps merely presumptive) entitlement to E, which she enjoys in virtue of having apprehended the speaker’s assertion, and the materials that determine the justification of her belief that P, formed on the basis of that assertion. The schema for bridging this gap will depend on whether one’s epistemology is internalist or externalist. If one’s epistemology is internalist, the gap will be regarded as to-be- bridged by the audience’s reasons or evidence: her reasons or evidence, together with her observation of the assertion and her (perhaps merely presumptive) entitlement to E, must rationalize the transition to the belief that P. If one’s epistemology is externalist, the gap will be regarded as to-be-bridged by factors 15 According to Ross’s account, E is the proposition that the speaker is relevantly authoritative regarding P. According to the minimal account, E is the proposition that the speaker acknowledges that she is responsible for being relevantly authoritative regarding P. 16 which render reliable the transition, from her observation of the assertion that P, to her coming to believe that P. (Presumably, these factors involve but are not exhausted by the audience’s background beliefs and her counterfactual sensitivity to things “not appearing right” in the way of assertion.) To see how this works in a concrete case, let us assume that we want an account, not just of the audience’s justified belief that P, but her knowledge that P, formed through her acceptance of the speaker’s assertion that P. In presenting such an account I will assume Ross’s account of the assertion-generated entitlement, as well an externalist account of epistemology. Both of these assumptions are made for the purpose of illustration only: together they place the weakest demands on an audience, and so render the gap more easily bridgeable than on other assumptions. Consequently, if we can show that even here bridging the gap is a non-trivial affair, the point will be all the stronger if we have other epistemological views (in general, or regarding assertion in particular). So imagine that S, aiming to communicate that P to A, asserts that P in A’s presence, and that A apprehends this assertion with understanding. That is, A knows that S has asserted that P. Then, from Ross’s model, A is presumptively entitled to regard S as relevantly authoritative regarding the truth of the proposition that P. Clearly, A is not thereby in a position, merely in virtue of this presumptive entitlement and the other facts of the case, to know that P through accepting S’s assertion. For one thing, S’s assertion might be false. For another, even if S’s assertion is true, this might be a matter of luck. Let us suppose, though, that neither of these things are true: S’s assertion was knowledgeable. This still does not suffice for A to be in a position to come to know that P through accepting S’s assertion to that effect. As noted above, if A’s acceptance is to be justified, it would seem that A cannot simply 17 depend on the knowledgeableness of S’s assertion; she needs to protect herself by being counterfactually sensitive to signs of incompetence or insincerity. That is, A must be such, that, were S’s assertion to have been incompetent (unreliable) or otherwise insincere, A would (likely) have picked up on this, and would not have accepted this assertion. How does A attain this sort of counterfactual sensitivity? Well, assertions that are insincere or incompetent often (though not always) exhibit features that give this away: the speaker didn’t look her audience in the eyes; the speaker seemed nervous; what she said just didn’t make sense; etc. In other cases, while the assertions may not have those tell-tale signs of incompetence or insincerity, the audience has background knowledge which, applied to the case at hand, sets off bells in her mind. This is precisely the job that is done by the two sorts of assessment I mentioned above. It should not be particularly controversial to say that this sort of “counterfactual sensitivity” is a large part of our dealings with other speakers’ assertions. But it is worth noting that the fact of an audience’s counterfactual sensitivity reverberates for speakers as well. This is for a simple reason. Speech acts are performed with practical aims in mind. The speech act of assertion is typically performed with the aim of getting one’s audience to believe what one asserts to be the case.16 When this is one’s aim, one is well-served to have in mind how the candidate audience is likely to react to one’s assertion. Insofar as one anticipates that one’s audience will be counterfactually sensitive in this way, one will adjust one’s assertions accordingly: if one is aware that one can’t get away with just anything, presumably this will have a bit of a constraining effect on one’s linguistic behavior. 16 That is, this is what Austin would have called the “perlocutionary aim”. 18  What I am calling the “epistemology of assertions,” then, is that part of the practice of assertion which, informed by the rich stock of mutual expectations that hold between speaker and audience, constrain the audience’s acceptance of assertions, and the speaker’s making of assertions, in virtue of the epistemic expectations and background knowledge the various parties bring to the communicative exchange. 5. Assertion under conditions of anonymity I have been developing the idea that the practice of assertion – both the making of assertions by speakers, and their acceptance by audiences – is shaped in large part by the mutual expectations between speaker and audience, as well as the background knowledge the various parties bring to the communicative exchange. I now want to argue that a good deal of the expectations in play regarding ordinary assertions are unwarranted, and much of the background knowledge needed for the purpose of assessment will be unavailable, when it comes to assertions that are made anonymously. The first point I wish to make is straightforward: when assertions are anonymous, there is little that the audience has to go on in assessing the assertion for credibility. (As a result, the audience’s “counterfactual sensitivity” to defeaters will be correspondingly diminished.) To see this, let us consider how the anonymity of an assertion bears on the conditions for credibility-assessment. Consider (what was noted above) the four types of information that the hearer might bring to bear to the task of assessing assertions for reliability. These include: (1) information regarding the speaker; (2) information regarding the act of communication (or assertion) itself; (3) information regarding the content of the communicated message; and (4) information regarding the context of the 19 communication.17 Cases can be found in which the hearer has ample information in each of these categories: consider for example a case of face-to-face speech with a good friend on a topic that is mutually familiar to speaker and hearer alike. What is more, it is easy to think of cases in which the hearer has significant information from some but not all of these categories. Cases of “stranger testimony” on a topic that is unfamiliar to the hearer – cases which are common in the literature in the epistemology of testimony – are cases in which the hearer typically enters with little or no background information to bring to the task of assessment; but when the exchange is face-to-face, the hearer can acquire some information regarding both (1) and (2) from the exchange itself, and this is often enough to enable her to reach an assessment. (The speaker looked normal, spoke with apparent confidence, on a topic regarding which people typically can be expected to be both sincere and competent, etc.) Cases of reading newspaper articles are typically cases in which the hearer/reader has ample information in (4), and from this she can often draw relevant inferences regarding (1) from the fact that the speaker/writer’s article is published in the high-reputation newspaper. And so forth. But now consider assertions made under conditions in which the speaker/writer is anonymous relative to the hearer. As in “stranger cases,” so here too the hearer will start out with minimal background information with which to assess the speaker; but unlike stranger cases, the anonymity of the exchange ensures that there is significantly less information to be gleaned from the exchange itself with which to perform the assessment. Since the speaker/writer is unknown to the hearer/reader, the latter has no background information from categories (1) or (2). To be sure, if the message comes across as curious in any way – it exhibits 17 These sorts of information are not separate, of course: sometimes the assessment is made on the basis of knowledge of the speaker’s relation to the subject-matter of her testimony, connecting at least (1) and (3). 20 emotional volatility, or anger, or incoherence; the sentences are ungrammatical, or otherwise unintelligible; or something else just doesn’t seem right – then the audience, having come across the message, can draw inferences regarding the speaker’s competence and/or sincerity. But if what is written gives no basis for an inference of this kind to be drawn, then the audience is left with little in the way to go on regarding (1) and (2). What is more, anonymity may obscure the context in which the communicative act itself was produced, and/or may obscure what the speaker/writer anticipated in the way of the context in which the message would be encountered by the hearer/reader. In such cases, the hearer has little information from (4) either. In such cases where background information from (1), (2), and (4) is minimal even after the hearer has observed the message, the only information that the hearer has to go on is the background information she has on the subject- matter of the message. To be sure, if the message is on a topic on which the hearer has ample background knowledge, she may have enough background information to assess the reliability of the message: perhaps the message is on a topic on which the hearer can draw on her expertise, and so can discern attempts to “fake” such expertise; perhaps the message is on a topic in which the hearer knows that saying something that even achieved prima facie plausibility was something that likely could come only from someone who knew what she was talking about; and so forth. What is more, insofar as the message is extended (involving lots of individual assertions constituting a whole report), it can be assessed for internal coherence, and there will be an increasing number of points at which the hearer’s general background information can be brought to bear on the report’s overall plausibility. But there can also be cases in which the hearer has little or no relevant expertise on which to draw, and the report offers too little information from which to draw any 21 substantial inferences regarding its internal coherence or its overall plausibility. Such cases represent the sorts of cases in which the hearer’s background information is the least helpful in the task of assessment, with the result that, were the hearer to accept the report, she would be most epistemically dependent on the speaker/writer.18 It is clear that the epistemological perspective of the audience in such a case is seriously diminished, relative to the credibility-assessment task. For this very reason, it would appear that the audience’s counterfactual sensitivity to defeaters would be diminished as well. To see this, imagine that, randomly surfing the web one day, you come across a site that presents several pieces of information about a place – or a person, or a time period – about which you know virtually nothing. Nothing on the site gives you pause – the sentences don’t ramble, there would appear to be no vested interest that you can discern, it appears to be internally consistent and coherent, there are no claims that clash with your commonsense and the little relevant background knowledge that you bring to bear. Still, you don’t know who wrote this, you have no idea whether this site is monitored (and if so by whom), and so forth. How confident are you that, were the information on the site simply made-up, you’d recognize this, and would refrain from believing any of it? It would seem that your “counterfactual sensitivity” is seriously diminished. 18 It is interesting in this respect that the cases used in the literature for highlighting maximum (epistemic) dependence of a hearer on speaker are “stranger cases” rather than cases of anonymity. If the point is merely to highlight hearer epistemic dependence on speaker, the use of stranger cases is curious, since all else equal, hearer dependence is typically greater in anonymity cases than in “stranger cases.” Below I will be suggesting something that can explain the focus on “stranger cases” in this connection: we think of assertion as a public act, where the publicity of the act is not merely in the fact that it is an act produced in such a way that others can observe it, but also in the fact that the speaker herself is public in a sense to be identified – and this latter sort of publicness is central to assertion’s playing the knowledge- and information-spreading role that in fact it plays. Or so I will be arguing. 22  In short, from the epistemic perspective of the audience, it is clear that anonymous assertions present a challenge: most of the ordinary ways of assessing reliability are not available. This alone might lead epistemologists to conclude that anonymous assertions are epistemically perilous creatures: since one does not have much to go on in making a credibility-assessment, one’s counterfactual sensitivity to defeaters is diminished, and one is limited in the background information one might cite in any effort to render rational the move to accept the assertion. What is more, the hearer is aware that the various social practices that we have for “policing” assertions – the mechanisms whereby one’s community encourages (enforces?) speakers to live up to their assertion-generated responsibilities – are inapplicable. In these contexts it is far from clear how any belief formed through such acceptance might be doxastically justified (let alone knowledgeable when true). I will return to this point shortly. 6. The diminishment of mutual expectations and its effects on the pragmatics of anonymous assertion In this last section, however, I want to focus on a dimension of anonymity that standard positions in the epistemology of testimony might well overlook: the perspective of the speaker, the producer of assertions. It is here, in connection with what I will call the pragmatics of anonymous assertion, that the implications of anonymity are most profound; and it is here, too, that we can begin to appreciate the sense in which it is the very publicity of ordinary (non-anonymous) assertions that render this speech act apt for the communication of knowledge. As noted above, ordinary speakers recognize the various mechanisms that might be brought to bear on them, were it to be discovered that they asserted unwarrantedly. A speaker who asserts anonymously will recognize that no such 23 mechanisms can bear against her – or at any rate that it is highly unlikely that any such mechanisms will bear against her – for the simple reason that no one (for whom the assertion was anonymous) will be in a position to warrantedly connect her to the assertion she made anonymously. This will enable her to evade having to take responsibility for any unwarranted assertions. At a minimum, even otherwise- scrupulous speakers might lose some of the energy they have for self-scrutiny and self-policing; such speakers might give in to occasional tendencies to exaggerate or otherwise speak in self-serving ways (after all, no one will be checking up on them). To be sure, a speaker who goes too far – her tales are too tall, her exaggerations too big, her claims too implausible on their face – might well fail to get others to believe what she says. But while such an outcome might fail to advance her interests, still, it is no skin off of her back, so to speak, since it would not affect the trust she is accorded when she speaks in plain daylight. In short, such a speaker will quickly recognize that she is asserting in a context involving few or no mechanisms available for the encouragement (or enforcement) of assertoric responsibility. It should be equally clear that conditions of anonymity affect not only what the speaker can expect from an audience, but also what an audience can expect from a speaker. This is because any reasonable audience will recognize the situation confronting the anonymous speaker. Such an audience will see that, when mechanisms for the enforcement and encouragement of assertoric responsibility are minimal or nonexistent, the speaker has less to worry about in making unwarranted assumptions – her doing so will not adversely affect her self-interest in any significant way. An audience who recognizes this will also be aware that it is imminently reasonable to assume that the speaker herself recognizes this. Such an 24 audience, I submit, has, and appreciates that she has, generic grounds for skepticism regarding the truth of assertions presented anonymously. To be sure, these generic grounds for skepticism do not ensure that it is never warranted for an audience to accept an anonymous assertion. If an audience has some independent way to confirm that a particular anonymous assertion is credible – say, it’s on a topic on which she (the audience) has extensive background knowledge, the assertion’s content is plausible relative to that background knowledge, and she (the audience) recognizes that it is unlikely that one could simply luck on to such a plausible claim unless one really knew what one was talking about – then it can be reasonable for her to accept the anonymous assertion. (Here the generic grounds she has for skepticism are trumped by her specific grounds for regarding this particular anonymous assertion as credible.) Still, she is left with generic grounds for skepticism regarding anonymous assertions as a class. Of course, the speaker, too, is in a position in which to anticipate that the audience, reasoning in this fashion, will come to have diminished epistemic expectations regarding anonymous assertions as a class. Stronger, both speaker and audience alike are in a position to appreciate that the other will likely reason in this way. But for this very reason, these diminished expectations on both sides are likely to be mutually acknowledged. And this strikes me as an interesting outcome. For if I am right that it is the warranted mutual expectations of speakers and audiences that frame the epistemology of assertion, then to the extent that assertions made anonymously constitute a class of assertions regarding which there is a diminishment in the mutual expectations that are warranted, to just this extent the belief-worthiness potential of these assertions is diminished. 25  What I have in mind is this. I just noted that an audience has pro tanto grounds for generic skepticism regarding the whole class of anonymous assertions, and that this point is mutually familiar to any competent speaker and audience. But for precisely this reason, the epistemic threshold for warranted acceptance of an assertion is systematically harder to meet in cases in which the assertion was anonymous than it is in cases of “ordinary” assertion. More specifically, I submit that the following is a possibility: (P) Two hearers who are exactly alike with respect to their relevant degree of informedness, their competences relating to the comprehension and assessment of testimonies, and the background information they bring to the scene may nevertheless be such that the hearer who observes an “ordinary” (face-to-face) assertion, by S, to the effect that p is warranted in accepting it, whereas the hearer who observes an anonymous assertion that p (by S, albeit unbeknownst to the hearer) is not warranted in accepting the anonymous assertion. What might explain this? Since the speaker and the asserted content are the same in both cases, the epistemic difference in (P) cannot be chalked up to any difference in S’s epistemic position regarding the proposition she asserted (= the proposition that p). Nor can we explain the epistemic difference in (P) by appeal to a difference in the hearers’ epistemic perspective on the proposition asserted, since by hypothesis both hearers have the same relevant degree of informedness and bring the same background information to the scene. The epistemic difference in (P), rather, lies in the audiences’ epistemic position regarding the credibility assessment of the speaker and the asserting of the content. But we must also keep in mind that the audience’s overall epistemic inferiority in assessing cases of anonymous assertion is mutually familiar to speaker and hearer alike. So a speaker who asserts anonymously should expect that a hearer who encounters her anonymous assertion will recognize his (the hearer’s) own epistemic inferiority relative to the position he is in with respect to “ordinary” assertions. And this, I 26 think, means that one who asserts anonymously should not expect – is not warranted in expecting – that the audience’s assertion-generated expectations will be the same as in cases of ordinary assertion. Whereas in ordinary cases the hearer expects the speaker to recognize both the responsibilities attendant to the making of an assertion and the various mechanisms in place for “encouraging” speakers to live up to those responsibilities, the hearer who observes an anonymous assertion has grounds for thinking that there is little or no external “encouragement” putting any pressure on the speaker. It is in this way rational for the hearer to adjust her confidence in the trustworthiness of the assertion, absent reasons that address these concerns (e.g. of the sort suggested above). I have been arguing, in effect, that anonymity saps assertion of some of “the promise” of epistemic authoritativeness that ordinary (non-anonymous) assertion conveys. How might we square this with CKNA itself, that is, with the hypothesis that it is common knowledge that assertion answers to an epistemic norm? I think the answer is this: while the norm of assertion remains constant, the fact of anonymity itself provides a hearer with a generic reason for skepticism about the reliability of the assertion, with the result that, absent reasons with which to address this sort of skepticism (e.g. of the sort discussed above), the hearer is unwarranted in thinking that the speaker satisfied the norm of assertion. On such an account, it is by serving as this sort of reason that anonymity saps assertion of some of “the promise” of epistemic authoritativeness that ordinary assertion conveys. However we decide to model the phenomenon, though, this is perhaps the most controversial of the claims I am making: insofar as there is a diminishment in the mutual expectations that are warranted in connection with anonymous 27 assertion, the assertions themselves no longer convey that the speaker has the sort of epistemic authority that would be needed to warrant outright belief. I think that this point tells us something interesting about ordinary assertions: much of the epistemico-pragmatic potential of assertion as a speech-act type – much of what makes assertions apt to generate belief in hearers, and so apt for the transmission of knowledge – derives precisely from the fact that assertion is a public act. By speaking of assertion as a public act I have in mind two features of our ordinary assertoric practices. First, assertions are acts that are made “in public,” by one whose identity can be known and tracked over time. Second, assertions are acts that are addressed to an audience of other epistemic subjects, who can be expected to have certain expectations regarding the asserter’s epistemic perspective on the asserted content, who will rely on the mechanisms for ensuring general conformity with these expectations, and who can be expected to hold the speaker to her assertoric responsibilities. Insofar as assertions are anonymous, the basis of this “publicness” no longer exists; and my hypothesis is that the result is that the class of anonymous assertions do not have the belief-worthiness potential of ordinary assertions. Even if I am correct in thinking that anonymous assertions do not have the belief-worthiness potential of ordinary assertions, still, it is worth noting that this fact need not doom anonymous assertion from ever being usefully employed. Clearly, there are all sorts of motives one might have for remaining anonymous in one’s assertion, not all of which are nefarious. (Perhaps the most common non- nefarious motive is the fear of retribution when what is asserted goes against the interests of the strong or the powerful.) What is more, regimes are imaginable which allow for (some) anonymity while still preserving the belief-worthiness of the 28 anonymous assertions. One such regime would be the “security wall” model. On this model, someone or some group is authorized to serve as a both a security wall (behind which the speaker’s anonymity is preserved), but also as a filter (only allowing worthy asserters or worthy assertions to pass, and so to see the light of day). Insofar as the assertions that pass through the filter are seen as “backed” by the authoritativeness of the person or group who constitute the security wall, and insofar as members of that group can be held responsible for the warrantedness of assertions that so pass, the security wall model might preserve the belief- worthiness potential of anonymous assertion. Another regime might follow the “Wikipedia” model: we develop practices whereby all assertions made on a particular web site (for example) are anonymous, so the barrier to entry remains low; but anyone is entitled to anonymously correct any assertion that is made, so the “cost” of correction is low; and constant or at least regular vigilance by the anonymous masses holds out the hope that errors are soon corrected. Since the Wikipedia phenomenon has been extensively studied already, 19 we might well learn about the conditions under which this sort of regime produces reliable information, and the conditions under which it does not. My present point is only that it offers a model in which anonymity is compatible with the sort of belief-worthiness potential associated with ordinary (public) assertion. Work Cited Bach, K. and Harnish, R. 1979: Linguistic Communication and Speech Acts. MIT Press, Cambridge.) Benton, M. 2011: “Two More for the Knowledge Account of Assertion.” Analysis 71:4, 684-87. Burge, T. 1993: “Content Preservation.” Philosophical Review 102:4, 457-488. 19 See e.g. Fallis (2008). 29 Fallis, D. 2008: “Toward an Epistemology of Wikipedia." Journal of the American Society for Information Science and Technology, 59: 10, 1662-74. Fricker, E. 1994: “Against Gullibility.” In B.K. Matilal and A. Chakrabarti, eds., Knowing from Words, pp. 125-61. Amsterdam: Kluwer Academic Publishers. Fricker, M. 2007: Epistemic Injustice. Oxford: Oxford University Press. Goldberg, S. and Henderson, D. 2006: “Monitoring and Anti-Reductionism in the Epistemology of Testimony.” Philosophy and Phenomenological Research 72: 3, 576-93. Goldberg, S. 2007: Anti-Individualism: Mind and Language, Knowledge and Justification. Cambridge: Cambridge University Press. Goldberg, S. 2011: “Putting the Norm of Assertion to Work: The Case of Testimony.” In J. Brown and H. Cappelen (eds.), Assertion: New Essays, pp. 175-96. Oxford: Oxford University Press. McFarlane, J. 2011: “What Is Assertion?” In J. Brown and H. Cappelen (eds.), Assertion: New Essays, pp. 79-96. Oxford: Oxford University Press. Pagin, P. 2011: “Information and Assertoric Force.” In J. Brown and H. Cappelen (eds.), Assertion: New Essays, pp. 97-136. Oxford: Oxford University Press. Ross, A. 1986: “Why do we believe what we are told?” Ratio 28:1, 69-88. Weiner, M. 2005: “Must We Know What We Say?” Philosophical Review 114: 227- 51. Williamson, T. 1996: “Knowing and Asserting.” Philosophical Review 105: 489-523. Sanford Goldberg is Professor and Chair of the Philosophy Department at Northwestern University. He is author of Relying on Others: An Essay in Epistemology (Oxford University Press, 2010) and Anti-Individualism: Mind and Language, Knowledge and Justification (Cambridge University Press, 2007). 30