CFL Software

Elliot's Blog

Tales of an intern

Contract Essay Writing: A Perfect Mystery?

On 3rd May 2018, the Quality Assurance agency issued their response to a BBC investigation into contract cheating. Contract essay writing involves a student hiring someone, typically a nameless online essay writer who works for a larger essay mill, to write their essays for them. The issue posed by contract essay writers attacks the very purpose of getting a degree: to show that the student, and the student alone, is capable within their chosen subject. This is because by hiring someone else to write their essays, the student is not showing their capability in their chosen subject. This will harm both the student, who lacks the required skills, and their future employers, who trusted that the student knew their subject.

On the basis of this issue, we must find a way to detect contract essay writing so as to prevent students hiring them and therefore ensure that the student is the author of their work; however, detecting contract essay writing is not as simple as detecting plagiarism: these are original essays, not copied essays from other sources, and therefore cannot be identified by cross-comparisons with a larger database. As a result, the detection of contract essay writing poses a perfect mystery to current methods of plagiarism detection.

So how else do we prove that a student did not write their essay and that someone else did? Simply put, we must develop new methods for the detection of this new problem. During my summer internship at CFL, I have been researching how we might detect contract essay writing. I was armed with CFL’s arsenal of tools and broke down my task of authorship attribution into three steps: firstly, I needed to identify the authorship style of a student using their essays. Then, I needed to introduce a new essay from a different source, which I would hypothetically treat as an essay written by a contract essay writer hired by the student. Finally, I needed to prove that this new essay was not written by the student by showing that a style that differs from the student’s is present in this new essay.

To achieve the first step, I used my own undergraduate essays as my data. I then identified my style quantitatively, using the frequency of specific stylistic markers to measure my style, and found certain features which were consistent across my essays. I had therefore found a way to identify the style of the student.

To achieve the second step, I selected an essay from one of my friends. To achieve the final step, I compared the frequency of the specific stylistic markers that I used to identify my own style to the frequency of the same markers in this essay. As expected, I found that the frequency differed; we had two distinct styles. I had therefore identified a way to detect when a student may have hired a contract essay writer by detecting when the style of a newly submitted essay was different to the previous essays submitted by the student.

The issue then resided in proving beyond reasonable doubt that this student had employed a contract essay writer. I believe that the answer to this issue lies in statistics: specifically, I have been employing Wilcoxon Signed Ranks tests to show that the frequencies of two groups, specifically the frequencies of the selected stylistic markers in my essays and in the essay hypothetically written by a contract essay writer, are significantly different. This is because they result in a p-value of >0.05, meaning there is a 95% chance that these two groups are different due to actual differences and not random chance.

Is contract essay writing a perfect mystery? With current plagiarism detection methods, yes. However, I believe we at CFL can generate a solution to this hard problem and I believe we are making our first steps towards solving this mystery.

Elliot Holmes