Dealing with contract cheating: a question of attribution Dr Thomas Lancaster & Mr Robert Clarke School of Computing, Telecommunications and Networks Birmingham City University Millennium Point Birmingham B4 7XG thomas.lancaster@bcu.ac.uk robert.clarke@bcu.ac.uk Abstract The issue of attribution, identifying the institutions which students who attempt to outsource work are from, poses a major difficulty for detectives monitoring online sites used for contract cheating. This form of academic misconduct occurs when students get other people to complete assessed work for them. Previous studies on contract cheating have focused on student use of Internet-based outsourcing services. The studies have demonstrated that those sites primarily provide students with work for subjects falling within the computing spectrum. This paper focuses on a study of 627 sample postings made on EssayBay, a commercial site aimed at providing assignment writing solutions for students. The study identifies that students across a range of academic subjects and levels of study, far beyond the computing field, are using EssayBay for purposes of contract cheating. Only 23.7% of the postings investigated are found to be attributable, that is, they can be traced back to the academic institution to which the assignment specification belongs. This suggests that there are issues across the sector with the way that assignments are set and made available for detectives. Based on the study, two factors for measuring the attributability of a posting on a contract cheating are proposed, namely searchability and individuality. Searchability measures how easily a posting can be found using a search engine. Individuality measures how unique an assignment specification is. Generally, both searchability and individuality are necessary to allow assignment specifications to be attributed. The paper concludes by making recommendations detailing how academic departments can combat contract cheating by improving the attributability of the assignment specifications that they release. Keywords Contract cheating, academic integrity, attribution, searchability, individuality, EssayBay. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission. © 2012 Higher Education Academy 1. Introduction Contract cheating is a form of academic dishonesty, where students get others to complete work on their behalf, often for a fee. The term was originally coined by Clarke and Lancaster in 2005 with relation to Internet-based outsourcing services used by students [1]. The intention was to differentiate contract cheating from other forms of inappropriate behaviour that students had been observed undertaking, such as plagiarism. This form of cheating, often carried out by technologically savvy students, has meant that a range of new methods to prevent students from cheating have needed to be developed [2]. Clarke and Lancaster’s initial study [1] focused on the use of RentACoder [3] (now known as vWorker), an outsourcing site originally focused towards computing contractors. The study identified that 15% of postings made on the site in 2005 represented attempts by students to cheat. It also showed that a typical student user posted between 4 and 7 assignments, indicating that students were not being detected by staff and were at least achieving pass grades in the assignments that they had produced for them. A subsequent study showed that 90% of the assignment specifications posted on RentACoder were for subject areas associated with computing and information technology [4]. Sites like RentACoder do provide legitimate services for commercial users. The student use of these sites for cheating is not always condoned. Contract cheating can be differentiated from plagiarism since it is unlikely to be detected by the non-originality engines used by many UK universities, of which the most common is TurnItIn [5]. This is because the work submitted by the student is original (it has been produced to order for them), but it is just not the original work of the student. The original nature of this work makes this form of cheating difficult to detect. Two further problems are worthy of note. First, academics can find it difficult to evidence that a piece of student work is not original, even when cheating is suspected. This is because a definitive source document is usually not available. Second, when an assignment specification is found online, it is often impossible for a contract cheating detective to attribute this to a source institution, allowing them to be notified that one of their students is attempting to cheat. A combination of these difficulties led Clarke and Lancaster to propose a Six-Stage Process for Detecting Contract Cheating [6], although challenges with attribution still appear to pose a major problem. This paper attempts to quantify the difficulties of attribution, with reference to a study of postings made on EssayBay [7]. EssayBay is a UK based outsourcing site focusing explicitly on providing student assignments which has not previously received coverage in the contract cheating literature. The EssayBay exploration is used to introduce the factors of searchability and individuality, which are intended to categorise the level of attributability possible for posted assignments. 2. Commercial cheating using EssayBay This paper focuses primarily on the use of EssayBay [7] between March and July 2010. In August 2010, EssayBay was reporting as suffering a technical failure. Currently EssayBay operates a reduced service. EssayBay provides the following explanation for the services that it offers: “For the first time ever EssayBay connects writers with buyers that need custom essays, term papers, theses and dissertations and allows writers to offer their services directly based on the strength of their reputation and past results. Much like an auction, projects are posted by buyers, and interested writers bid on them. Buyers can see the qualifications of the writer, the guarantees they offer, and the feedback they have received from other buyers, before they choose to accept a bid.” [8] The EssayBay approach differs from many sites used by students for contract cheating which do not explicitly advertise themselves as cheating sites. Indeed, some outsourcing sites previously used by students for contract cheating may be disapproving. For example, recently RentACoder [3] appear to have adopted a policy of not allowing coursework requests to be submitted and regularly removes postings made that are of that nature. EssayBay itself is tailored towards student use for cheating purposes. A typical posting on EssayBay allows a student to define the academic subject from 78 options, select the academic level of the assignment and indicate the grade that they require. The student can also add a description. Unlike previous studies [1, 4] which showed a large bias in contract cheating attempts towards computing subjects, EssayBay is used by students across a range of disciplines. During the period between March and July 2010, 627 postings made on EssayBay were analysed to identify the types of assignments placed on the site. The postings were selected to examine the full range of subject areas that are involved by students who are attempting to contract cheat. No attempt was made to record and analyse every post. An analysis of the sequence numbers of the postings suggests that this sample represents around 10% of postings made during that period. This suggests that slightly over 2,000 posts a month were made on EssayBay by students looking to have work completed for them. An initial problem of classification of subject areas arose from the options made available by EssayBay. This was a list of 78 subjects ranging from Accounting to Transport, of which a student could choose multiple options. These categories proved too wide to allow the aggregation of small subject areas. Students also did not seem to always identify the most appropriate category, particularly where multiple subjects were selected. Therefore, the postings were reclassified using a hierarchical classification based upon the UCAS/JACS2 21 subject areas. A manual decision was made to identify the most appropriate category in each case. Figure 1 shows a breakdown of the 627 postings as classified into the UCAS/JACS 21 subject areas. This sample only identified postings from 15 of the 21 subject areas. Key results from Figure 1 show that the most commonly posted assignments were from the subject category “Business & Admin studies”. These comprised 24.5% of all the postings analysed. Postings from “Mathematical and Computer Science” amounted to 5.0% of the total. These figures seem to be in line with the number of students studying these disciplines. This suggests that students from all disciplines are prone to use sites like EssayBay. Figure 1: Analysis of the EssayBay postings by UCAS subject category Figure 2 shows an analysis of the 627 postings by the academic level indicated in the postings. In most cases, this is the level selected by the student, although a few cases were noted where this had been moderated down by the researcher, for instance from Masters to Undergraduate Level. It appears that some students think that they will get a better result if the requested level is boosted. Level Count % International Bacc. 2 0.3 PhD ? 2 0.3 GNVQ 4 0.6 Advanced Placement 5 0.8 PhD 10 1.6 High School 11 1.8 GCSE A/A2 15 2.4 Masters 133 21.2 Undergraduate 445 71.0 Total 627 100.0 Figure 2: Analysis of the EssayBay postings by academic level At 71.0%, undergraduate assignments made up the bulk of the postings, as could be expected. More seriously, 21.2% were for postings at Masters Level, including final dissertations. Further, 10 postings were recorded as being at PhD Level, with another 2 likely at this level. Examples of PhD level assignments included research proposal documents and literature reviews. Requests for complete PhD theses were observed, but it is not known if any of these were successful outsourced and submitted. A final analysis attempted to attribute the source institution for the 627 postings. This was successful in only 23.7% (149 out of 627) of the cases, indicating the difficulties posed by attributability. Figure 3 shows a breakdown of these identified assignments by the country of the institution attributed. The results in Figure 3 are reflective of the fact EssayBay is an English language site, used primarily by users in English speaking countries. EssayBay does allow students to indicate their country of original, but this has been seen to be a poor predictor for purposes of attributability. The small number of attributed postings may reflect some courses run in the English language in countries other than the four identified that could not be attributed. Row UCAS Subject Category Australia Canada UK USA Total B Subjects allied to Medicine 3 1 2 2 8 C Biological Sciences 1 1 7 9 F Physical Sciences 2 3 5 G Mathematical & Comp Sci 16 16 H Engineering 6 7 13 K Architecture,Build & Plan 1 3 4 L Social Studies 1 6 7 14 M Law 6 4 10 N Business & Admin studies 13 14 9 36 P Mass Comms and Documentation 1 1 2 4 Q Linguistics, Classics & related 2 6 1 9 R European Langs, Lit & related 1 1 V Hist & Philosophical studies 1 7 8 W Creative Arts & Design 1 2 3 X Education 1 8 9 Column Total 38 2 63 46 149 Figure 3 – Analysis of the attributed EssayBay postings by country of institution Figures 1 to 3 demonstrate widespread use of EssayBay and similar sites across all academic disciplines and at a wide range of academic levels. Figure 3 further demonstrates the difficulties of attributability of assignments placed on sites such as EssayBay. 3. Characterising the attributability of assignments The difficulty of attributing assignment specifications to institutions was raised by Lancaster and Clarke as part of their Six-Stage Process [6]. Informal observation with other data sets has shown typical attributability rates of around 20% to 30%. This is also reflected in the EssayBay data set. The difficulties with attributability for a contract cheating detective are widespread, particularly since students may try to disguise an assignment specification by stripping out information that easily identifies its source. Successful attribution often depends on finding many small clues which lead to one academic institution. The types of clues available to a contract cheating detective are varied. Some examples are useful here. Students can post the same assignment specification more than once, either on different sites, or when an outsourcing attempt has failed. Standard rubrics used on multiple modules across an institution can be observed. The unique module codes used in many UK and Australian universities can also help, although this method can be compromised with the volume of modules tagged with codes like CS101. Finally, although the student’s reported location may not always be reliable, this can provide a useful additional indicator of the location of a university. The observation of the difficulties with attribution has led to the proposal of two attributability factors, searchability and individuality. Searchability is intended to measure the extent to which the posted assignment specification, or the institution from which it was taken, can be found online. This may be using a search engine such as Google [9] or a non-originality engine such as TurnItIn [5]. An assignment which could not easily be found online would receive a low searchability score. Individuality is intended to measure how well an assignment specification can be distinguished from other specifications on a similar topic. A standard assignment taken from a textbook would receive a low individuality score. For a posting to be attributable, high scores are generally needed for both searchability and individuality. An assignment specification taken from a textbook might end up being used at multiple institutions, thus receiving a high searchability score, but a low individuality score. This would not be attributable. Similarly, a tutor may aim to design out cheating opportunities by automatically generating unique data sets for each student in a class. This would give a set of assignment specifications, each with a high individuality score, but unless sample text was made available on the web, these would obtain low searchability scores. The exact dividing line for high and low attributability is fuzzy, even when the searchability and individuality factors are taken into account. Continued study is needed to define a clear mapping across these properties. 4. Recommendations to improve the attributability of assignments The analysis shows that it would be possible for detectives to forewarn many more academics that their students were contract cheating if the level of attributability were increased. Although attribution does not always identify which student in a cohort was cheating, it does tell staff that they should look for suspicious submissions of student work. It can also help persuade staff to design out cheating opportunities in the future. Left unchecked, contract cheating has the potential to seriously undermine confidence in the standards of academic awards. It is recommended that methods of improving attributability are explored at both departmental and institutional level. In particular, the anti-cheating and plagiarism measures in place for every assignment should be formally reviewed, documented and disseminated during quality processes. The reuse of old assignment specifications and the use of standard examples from textbooks should be discouraged. Practical measures to improve searchability for assignment specifications should be considered. These include introducing unique tags, such as module codes in file names and company names in a case study. Misspellings and idiosyncratic word orders can be used. The use of metadata fields and watermarking should be considered where practically possible. Assignment specifications should also contain standard wording, which is searchable. Specifications, including individual items such as student projects, should be published and made accessible to search engines and not hidden inside private Virtual Learning Environments. 5. Conclusions This investigation of EssayBay shows one example of the commercialisation of cheating by students. Although EssayBay is currently operating a reduced service, contract cheating continues on a variety of other outsourcing sites, and inside essay mills. The contents of these sites are not easily observable by detectives. A rising number of postings requesting the proof reading of essays and dissertations have also been observed, a behaviour which can at best be considered semi-legal. Continual improvements of anti-contract cheating procedures are needed. In particular, the issue of attributability needs to be explored further, with reference to developing watertight processes to make postings on contract cheating sites searchable and identifiable. A suitable scale on which the attributability factor can be academically discussed needs to also be finalised. 6. References [1] Clarke, R. & Lancaster, T. (2006). Eliminating The Successor To Plagiarism? Identifying The Usage Of Contract Cheating Sites, in Proceedings of 2nd Plagiarism: Prevention, Practice and Policy Conference [2] Lancaster, T. & Clarke, R. (2007). The Phenomena Of Contract Cheating, in Student Plagiarism In An Online World: Problems And Solutions, in Roberts, T. S. (editor), Idea Group Inc, Hershey, Pennsylvania [3] VWorker.com: How Work Gets Done, Guaranteed! (2012). http://www.vworker.com [Accessed 17 February 2012] [4] Lancaster, T. and Clarke R. (2007). Assessing Contract Cheating Through Auction Sites – A Computing Perspective, in Proceedings Of 8th Higher Education Academy Annual Conference For Information And Computer Sciences [5] TurnItIn: Leading Plagiarism Checker, Online Grading and Peer Review (2012). http://www.turnitin.com [Accessed 17 February 2012] [6] Clarke, R. & Lancaster, T. (2007). Establishing A Systematic Six-Stage Process For Detecting Contract Cheating, in Proceedings Of 2nd International Conference On Pervasive Computing And Applications [7] Term Papers and Custom Essays | The World's Only Marketplace for Custom Written Papers (2012). http://www.essaybay.com [Accessed 17 February 2012] [8] EssayBay | About EssayBay (2012). http://www.essaybay.com/about.php [Accessed 14 February 2012] [9] Google (2012). http://www.google.com [Accessed 17 February 2012]