GhostDetect® is a tool for analyzing stylometric differences between two writing samples. It attempts to identify inconsistencies between a writer’s typical style of writing and another source. The underlying assumption behind this tool is that the greater the stylometric differences in writing samples, the less likely they are to be from the same author. By tracking divergences in authorship style, GhostDetect® therefore spotlights possible ghostwriting.
Academic ghostwriting is a significant and underappreciated problem in higher education. Unlike ordinary plagiarism, which copies and pastes publicly available textual sources, ghostwritten material arises from specific private arrangements between ghostwriters and cheating students, who then intend to pass the ghostwritten work off as their own.
Because academic ghostwriting illicitly copies other people’s work, it is a form of plagiarism. But because ghostwritten material is unique, leaving no internet footprint, it eludes conventional plagiarism checkers, such as Turnitin, which depend on comparing a written text with other already existing texts. With ghostwriting, no such other texts exist.
Academic ghostwriting poses a significant threat to the integrity and effectiveness of higher education. To redress this problem, AcademicInfluence.com has not only developed this tool but also, in June 2022, published a full-scale book on the topic: Dave Tomar’s The Complete Guide to Contract Cheating in Higher Education. The tool and book together empower educators to deal effectively with academic ghostwriting.
GhostDetect® attempts to differentiate the written work of distinct authors, and thus to place a much needed check on academic ghostwriting. Its results, however, must be interpreted with care. It cannot provide a smoking gun. Its results must be the start and not the end of any efforts to determine whether a student has cheated by employing a ghostwriter. It cannot, by itself, incriminate a student of cheating or provide grounds for disciplinary action.
Any tool like this cannot completely rule out false positives (i.e., mistakenly ascribing distinct authorship) or false negatives (i.e., mistakenly ascribing identical authorship). For instance, the same author may write in very different styles depending on genre and context (possibly leading to a false positive). At the same time, distinct authors may write in a similar style (possibly leading to a false negative). As such, similarities or differences in style exposed by this tool cannot decisively establish identical or distinct authorship.
Applying GhostDetect® is, however, a crucial first test for uncovering ghostwriting. Sharp divergences in style, when detected by this tool and purported to come from the same author, call for, at a minimum, some explanation and closer examination.
Use of this tool is largely self-explanatory. We welcome feedback on how to improve it.
Flury-Riedwyl faces assign graphical representations to points in multidimensional space. Because GhostDetect® associates multiple stylistic measurements with each writing sample, any sample can be represented as a point in multidimensional space and thus take the form of a Flury-Riedwyl face.
In their 1981 paper ”Graphical Representation of Multivariate Data by Means of Asymmetrical Faces,” statisticians Bernhard Flury and Hans Riedwyl extended Herman Chernoff’s groundbreaking 1973 work on using faces to display multivariate data. These came to be known as Chernoff faces, and subsequent faces were named after their inventors.
Although Flury and Riedwyl allowed in their work for both symmetrical and asymmetrical faces, asymmetrical faces tend to be perceived as oddities, distracting the observer. We have therefore limited the Flury-Riedwyl faces in GhostDetect® to those that are symmetrical. Moreover, we give users the option of also making the faces smile (since the smile/frown axis tends to have an overwhelming influence on facial perception).
Because humans are hardwired to make sense of facial patterns at a glance, faces such as these provide a quick method to compare degrees of variation between two writing samples. Facial features can communicate profoundly on a subconscious and emotional level. Care must therefore be taken not to over-interpret the faces that appear in GhostDetect®. For instance, any emotional meanings perceived in our Flury-Riedwyl faces are incidental to the data inputs, and bear no connection to the emotional tone or voice of the writings or the authors.