ASSESSING CONTRACT CHEATING THROUGH AUCTION SITES – A COMPUTING PERSPECTIVE Thomas Lancaster
Robert Clarke
Department of Computing University of Central England in Birmingham Perry Barr
Department of Computing University of Central England in Birmingham Perry Barr
Birmingham B42 2SU
Birmingham B42 2SU
United Kingdom
United Kingdom
[email protected]
[email protected]
ABSTRACT
lower moral values may be tempted to use.
The paper studies the use the RentACoder Web site to contract cheat by Computing students. RentACoder is an outsourcing service for computer work which operates under auction principles. Contract cheating is where students have assessed work completed for them on their behalf. The work is original, so not will be detected by the regular anti-plagiarism mechanisms that look for shared commonality. The paper describes the background to contract cheating and discusses a catalogue of 910 bid requests collected by the authors over two and a half years. The UK is seen to supply them with over 25% of contract cheating bid requests. This is largely composed of students outsourcing Java programming assignments; substantial projects are highlighted as a concern. Trends are seen to exist for other countries but are not the same as those identified for UK students.
The contract cheating problem has been considered alongside plagiarism [12], as the work that students are handing in has been written for them and represents the use of another writer’s work without acknowledgement. Well documented techniques to both prevent and detect plagiarism exist [13]. However, the original nature of work produced through contract cheating makes many of the antiplagiarism techniques unsuitable. For instance, using a search engine, such as Google [7], a Webbased detection engine, such as TurnItIn [16], or a source code scanner, such as JPlag [10], will not find contract cheating, as the source from which this has been copied is not available.
Keywords Contract cheating, plagiarism, outsourcing.
This paper presents a background to contract cheating, surveying the literature from a Computing perspective. The results of a study into how Computing students are using the RentACoder online auction site [14] to solicit contract cheating is detailed and trends relevant to UK academics are listed. The results suggest that tutors need to carefully consider their assessment methods.
INTRODUCTION Within the academic discipline of Computing, there are many opportunities available for students to cheat. The expected technological skills of these students means that they should be aware of how to best use electronic resources to assist themselves. The phenomena of contract cheating [1], where original work is completed on behalf or a student for submission, is one that Computing students with 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. © 2007 HE Academy for Information and Computer Sciences
BACKGROUND Contract cheating sites can be categorised into three main categories [12]. These include essay mills, discussion forums and auction sites. A further definition, the feed aggregator, such as RSSMad [15], provides a method of accessing the searching and cataloguing entries to the sites. The categorisations all provide a mechanism through which students can receive original work that they do not have to produce on their own. An essay mill, such as Prof Essays [4], is a site that exists solely to write custom materials for students. A discussion forum, such as Coding Forums [3], allows a student to directly ask online for help. This can lead to original material being provided. Most of the existing contract cheating research and interest is focused on auction sites. This primarily
involves the computer work outsourcing service, RentACoder [14]. Such a service has legitimate contracting uses, for instance a small business without internal computer skills could request an online presence. However students can make use of this service to have untraceable work produced. An auction process works as follows. A customer writes a bid request, detailing their requirements, which is placed on the auction site. A contract cheating student would place details of the assignment that they wanted completing. Primarily these are computing programming assignments, but they may include essays or reports. This bid request is placed out to tender. Coders bid to complete the work required. The buyer chooses one of the bids, usually the lowest priced one from a competent looking coder. Original work is then produced for the student, which they can submit for credit. Students clearly appreciate outsourcing for its low cost and untraceable nature. In one study programming solutions were solicited and submitted for marking [9]. Solutions obtaining “very good marks” cost between $10 and $25 (US Dollars). Essays have been solicited for under $20 [8]. A single quantitative study of the use of RentACoder in this manner exists [1]. An extended subset of the data from this study forms the basis of the study specific to the Computing discipline reported here. The study reported three important results about users employing RentACoder for contract cheating:
subcontractors, outsourcing assignment requirements from multiple students. The results suggest that auction sites will continue to exist. RentACoder states that they allow assignments to be outsourced on the site because they “don't have the resources to devote to ineffective policing“ [11]. However, it is clear that contract cheating is a major revenue source for RentACoder which would explain why RentACoder does not provide visible anti-contract cheating messages. This becomes the job for academics.
CONTRACT CHEATING DETECTION Continued monitoring of contract cheating auction sites, for the purposes of detection, is necessary. Figure 1 shows the Clarke and Lancaster Six-Stage Contract Cheating Detection Process [2]. The process provides an outline framework from within which the data for this paper has been collected. The Six-Stage Process provides a solution for two main problems with detecting contract cheating: •
Auction sites need to be regularly monitored to ensure that attempts to cheat are found and recorded whilst they are available to the public.
•
It is usually not possible to trace the work that the student submits to a source. Instead the source of an assignment specification must be found by a contract cheating detective. This means that tutors must make assignment specifications accessible to a detective.
•
An intensive study of all bid requests placed on RentACoder over a three week period, found that 12.3% of these (99 out of 803 bid requests) represented contract cheating.
Essentially Figure 1 shows six stages, two carried out by a tutor producing an assignment (denoted “T” in Figure 1), four carried out by a detective investigating cheating (denoted “D”).
•
A two month study found that 51.7% of users who placed identified contract cheating bid requests (122 out of 236 users) were “habitual” cheaters, who had previously placed between two and seven (inclusive) bid requests.
•
The two month study found that 2.5% (6 out of 236 users) had made 51 or more bid requests. This suggests that these might not be individual students, but could be third party
In the largely independent publication stage a tutor publishes assignment details, sufficient for themselves to be traced, to a central corpus. If the assignment specification is subsequently found on an auction site the tutor is provided with a nonoriginality report in the investigation stage, where the tutor can carry out a suitable localised procedure.
Figure 1. The Clarke and Lancaster Six-Stage Contract Cheating Detection Process [2].
The detective leads four stages. In collection a corpus of bid requests placed on the auction site is compiled. The identification stage separates those that represent cheating from those that are likely legitimate. The attribution stage finds the institution and tutor involved using the assignment specification corpus and the tutor is contacted during the notification stage. During the process a formal record of evidence is continually compiled.
as an identifiable country (2.3%) may be due to its provision of English language education. This further suggests that once a small number of students in an area make use of a service their peer group follow. Number of Country % institutions USA 203 57.7% UK 75 21.3% Australia 25 7.1% OTHER 21 6.0% Canada 19 5.4% Turkey 9 2.6% Total 352 100% Table 2. Institutions involved with making bid requests.
Figure 1 shows that many stages are labour intensive. The computer symbols shows that only publication and collection can be fully automated. The other stages require manual intervention. At present the process represents an ideal, rather than an actuality. The assignment specification corpus does not exist, meaning that a detective monitoring an auction site has to resort to Internet searching to identify likely associated tutors. This creates problems when assignments are not original; for instance some Computing tutors are prone to reusing standard programming exercises from textbooks. Some students are good at disguising which academic institution they attend. The data reported in this study has been collected during this process to be as complete as possible, but a 100% success rate is not yet possible.
Table 2 shows the number of institutions in each country from which contract cheating attempts have been identified. There is a clear correlation between the percentages in Tables 1 and 2. This again demonstrates the widespread nature of the problem. Number of bid Frequency % requests made 1 to 3 303 50.7% 4 to 8 145 24.2% 9 to 15 71 11.9% 16 to 25 38 6.4% 26 to 59 28 4.7% 60+ 13 2.2% Total 598 100% Table 3. Frequency of RentACoder reuse.
A STUDY OF COMPUTING STUDENTS The authors have compiled a student bid corpus containing 910 RentACoder bid requests collected from March 2004 to December 2006 and identified as traceable Computing assignments. It is informative to study the data collected in order to establish which students are cheating and what they are cheating on. Country Number of cases % USA 462 50.8% UK 231 25.4% Australia 101 11.1% Canada 62 6.8% OTHER 33 3.6% Turkey 21 2.3% Total 910 100% Table 1. Countries involved with making bid requests.
The frequency through which each of the 910 bid requests have reused RentACoder is shown in Table 3. Since some users post more than one bid request the total number of unique buyers identified is 598. For each of these, their total number of bid requests made on RentACoder has been checked. There has been no attempt to verify that all of their previous bid requests represent contract cheating attempts, but it seems likely that many were. Three observations of note can be made:
Some background to the corpus is immediately useful. Table 1 shows the country from which assignment found as a bid request was identified. Although it is not surprising that most attempts (50.8%) to contract cheat originate from the USA, a sizeable minority (25.4%) originate from within the UK. This is far in excess of the expected relative size of the two countries. The appearance of Turkey
•
50.7% of users (303 out of 598) made between one and three bid requests. This is less than in the earlier study [1]. The apparent reduction may be due to an influx of new users who are just finding out about the opportunity or getting established on the site. It may also be that the use by Computing students differs from that of the wider student body. A more encouraging
view would be that cheaters are being caught and so long term use never gets established. •
•
There are still habitual cheaters. 36.1% (216 out of 598) made between four and fifteen requests, a larger range of repeat values than in the previous study. The percentage is enough to classify these users as “repeat cheats”. Likely third party subcontractors have a continued existence. 2.2% of users (13 out of 598) made 60 or more bid requests. This is far in excess of the number an individual student would be expected to make.
Of particular interest to Computing academics are the type of assignments that users are placing on RentACoder. Table 4 shows the frequency of each
type of Computing subject in the student bid corpus, along with the country from which the outsourced assignment originated. It is sorted by the overall frequency of each type of assignment. Programming assessment features highly in the corpus. 60.0% bid of requests (545 out of 910) are for a form of programming. Two types of assignments dominate the results. 20.3% request C or C++ programming (185 out of 910). 19.8% are for Java programming (180 out of 910). The third place result, Operating Systems and Utilities assignments, represents only 7.6% (69 out of 910), a substantial drop from the first two results.
Type Australia Canada OTHER TURKEY UK Programming - C/C++ 18 13 8 3 21 Programming - Java 21 24 2 3 52 Operating Systems & Utilities 4 3 2 3 6 Programming (unspecified) 1 5 1 7 Database - implementation 7 5 3 22 Web Programming 8 1 2 2 8 Major Project 2 5 22 Database - design 3 1 9 Systems analysis 3 2 14 Networking 6 1 1 8 OO Modelling 1 1 1 14 Computer Science 1 2 7 Algorithm analysis 3 Programming - Visual Basic 1 1 2 4 Assembler 1 1 2 Web Design 4 1 1 7 Programming - MathLab 1 1 5 Programming - Prolog 2 1 3 Programming - Cobol 2 1 Excel 3 1 Multimedia 1 1 1 6 Programming - OpenGL 3 1 1 Mathematics 1 3 Programming - graphics 1 1 Hardware 2 1 2 Artifical Intelligence 2 1 Programming - C# 2 Programming - Pascal 1 Formal Methods 2 Graphics - Application 1 Programming - Haskell 2 1 Programming - J2ME 1 1 Programming - Maple 1 Programming - MathCad 1 Programming - ML Programming - SML Total 101 62 33 21 231 Table 4. Number of assignments of different types requested.
There are noticeable regional differences between the assignment solutions requested. Although C and C++ programming dominate the requests from the USA, comprising 26.4% of regional bid requests (122 out of 462 requests) this is not reflected in the other countries. Outside the USA Java assignments provide most overall requests. In the UK, these are 22.5% (52 out of 231) of the requests. The UK has an identifiable trend for outsourcing major projects, such as a final year undergraduate project. 9.5% (22 out of 231) bid requests from UK RentACoder users are for project work. Tutors
USA 122 78 51 41 16 14 3 13 7 9 8 8 14 9 12 2 7 8 7 5 4 4 6 1 2 3 3 1 2
1 1 462
Total 185 180 69 55 53 35 32 26 26 25 25 18 17 17 16 15 14 14 10 9 9 9 8 8 6 5 5 4 3 3 3 2 1 1 1 1 910
largely understand the importance of this work and can attempt to dissuade cheating by requiring regular submission of components, such as a specification, interim, draft and final report, Students appear to be outsmarting them by asking for work to mirror this process. The practice of setting major projects that look like small commercial systems, becoming common in many institutions, is an ongoing cause for concern. They can increase the possibility for contract cheating as they are hard to distinguish from legitimate commercial systems.
FUTURE DIRECTION It is clear that Computing students are outsourcing their work to auction sites. The relatively large proportion of UK students involved suggests that this of great local concern. The local Computing community needs to take the opportunity to be seen to be tacking the contract cheating problem, in particular within the noted problem areas of programming assignments and project work. Assessing a robust system for monitoring and detecting contract cheating is vital. The Six-Stage Process, shown in Figure 1, provides a workable model but needs software support and a recognised service within which to operate. Advances in tracing individual students, such as setting individually identifiable versions of assessments [6], are needed. Alternatively a viva examination could be used, particularly useful for assessing project work. Detection must be used alongside methods of discouraging contract cheating. A number of recommendations exist [12]. Using alternatives to coursework such as practical programming examinations [5] makes outsourcing less viable. UK academics must continue to develop robust anticontract cheating processes and procedures. More studies are needed about how and why students contract cheat. Although many of the students cheating are on Computing degrees, contract cheating is not solely a Computing problem. But it is the Computing tutors who are best placed to develop the software to aid in the detection process. It is hoped that a community led approach to tackle the problem and raise awareness can be initiated.
REFERENCES [1] Clarke, R., and Lancaster, T., Eliminating the Successor to Plagiarism? Identifying the usage of contract cheating sites, Proceedings of 2nd International Plagiarism Conference (2006). [2] Clarke, R., and Lancaster, T., Establishing a Systematic Six-Stage Process for Detecting Contract Cheating, submitted to The Second International Conference on Pervasive Computing and Applications (2007).
[3] CodingForums.com - Web coding and development forums. Get help on JavaScript, PHP, CSS, XML, mySQL, ASP, and more!, http://www.codingforums.com. [4] Custom Essays Writing Service – ProfEssays, http://www.professays.com. [5] Cutts, Q., Barnes, D., Bibby, P., Bown, J., Bush, V., Campbell, P., Fincher, S., Jamieson, S., Jenkins, T., Jones, M., Kazatov, D., Lancaster, T., Ratcliffe, M., Seisenberg, M., Shinner-Kennedy, D., Wagstaff, C., White, L., and Whyley, C., Laboratory exams in first programming courses, Proceedings of 7th Annual Conference for Information and Computer Sciences, 224-228 (2006). [6] Fincher, S., Barnes, D., Bibby, P., Bown, J., Bush, V., Campbell, P., Cutts, Q, Jamieson, s., Jenkins, T., Jones, M., Kazatov, D., Lancaster, T., Ratcliffe, M., Seisenberg, M., Shinner-Kennedy, D., Wagstaff, C., White, L., and Whyley, C. Some good ideas from the Disciplinary Commons, Proceedings of 7th Annual Conference for Information and Computer Sciences, 153-158 (2006). [7] Google, http://www.google.com. [8] Gusmaroli, D., The Cybercheats Making a Small Fortune, Daily Mail, June 17, 2006, 12-13 (2006). [9] Jenkins, T., and Helmore, S., Coursework for Cash: The threat from on-line plagiarism, Proceedings of 7th Annual Conference for Information and Computer Sciences, 121-126 (2006). [10] JPlag – Detecting software https://www.ipd.uni-karlsruhe.de/jplag.
plagiarism,
[11] Kom, J., Cheating Students Outsource to Lowest Bidder, The Ottawa Citizen, September 25, 2006 (2006). [12] Lancaster, T., and Clarke, R., “The Phenomena of Contract Cheating”, to appear in Student Plagiarism in an Online World: Problems and solutions, Idea Group, Inc. (2007). [13] Lathrop, A., and Foss, K., Student Cheating and Plagiarism in the Internet Era – A wake up call, Libraries Unlimited Inc, (2000). [14] Rent A Coder: How http://www.rentacoder.com.
software
[15] RSS Mad, http://www.rssmad.com. [16] Turnitin, http://www.turnitin.com.
gets
done,
