Literatu delivers a fresh approach using ML and AI to build a digital, lexical thumbprint for every writer.
Some have called the Bill to amend the Tertiary Education Quality and Standards Agency (TEQSA) Act 2011 great progress. The Bill effectively makes it an offence to provide or advertise academic ‘contract cheating’ services in higher education. Whilst aiming directly at companies that provide contract cheating services, enforcing this legislation against highly mobile online businesses will be like trying to find and knock out a moving target. While our defense to contract cheating is in Parliament, the colloquially termed ‘essay mills’ are moving at warp speed to better cover their tracks whilst probably improving their service.
Interestingly, students who are found to be cheating will continue to be subject to each institutions’ academic integrity policies and sanctions. There are no changes there in the legislation. Higher education providers must keep extending internal academic integrity processes hoping TEQSA has an impact on stopping or slowing the external essay mill supply lines.
Will attempting to stop access to essay mills even be possible? Won’t the essay mills simply morph into something else? In an open discussion with senior educators it’s apparent there is a real problem. “Students can order any level of grade for any topic from contract writers simply by asking for a B+, for example.” Asking for a plausible text tuned to your nominated level of grade is all part of the service.
“When a B level student hands in a B range essay everything appears normal. If the text then passes through essay mentoring and advisory services and Turnitin or other plagiarism checking, it becomes a defacto original student work. This makes it very hard to accuse anyone of anything and is not worth pursuing on a hunch.” Is it possible that current process checks could unintentionally be helping to ‘launder’ contract essays?
What a conundrum educators face, and what a task TEQSA has on its hands. Cheating detection at source is where the fight against contract cheating needs to be refocused. More at source accountability backed up by better detection processes will work.
Every writer has a unique ‘lexical thumbprint’
Is there a way to detect non-original work other than through plagiarism detection? Literatu is proving there is a way and the solution does not involve people or plagiarism engines. Literatu is in the final stages of deploying machine learning and AI capabilities that build a unique Originality Index (OI) from each student’s writing abilities and traits. The new OI link to the author of written texts is akin to a DNA match for humans.
Given an original student text, Literatu builds an OI tagged to the student. All subsequent student submissions are calibrated against the baseline OI. The check is instantly calculated to inform educators that the ‘B’ level text may look like it is from the B level student but in fact has been written by someone else, with B level skills.
Literatu is proving that when we write, our unique combinations of meta cognitive skills influence the way we express opinion, use vocabulary and sentence structures, implement cohesive strategies and rhetorical moves, and use grammar (just for starters). The hypothesis of our research is that writing styles and traits repeat, and can be as unique as speech patterns. Everyone is different when it comes to communication, and tracking writing is in many ways no different to other codified checks like voice oscillations.
Literatu is also proving that writers approach different topic challenges the same way, often by simply re-combining their core literacy skills. When they step outside these skill and lexical competence boundaries, and for example ‘copy and paste’ other text, the identification and spread of differing styles actually amplifies.
Writer-unique lexical attributes are what Literatu identifies and shapes into an Originality Index. When original lexical attributes are replicated there is a high probability of original work. If there is an emerging improvement in writing skills over time, the OI score can be recalibrated with a new baseline. If there is a noticeable gain or drop in the OI score, Literatu gives an estimate on how close to original the text is.
Literatu believes there is a new way to approach the age old problem of monitoring the academic originality and integrity standards of all writers. Existing plagiarism techniques identify copy, paste, similarity and reference incidents. Profiling original writing as a thumbprint of each student’s unique abilities means students must keep their writing real. This approach offers a potentially viable way to reduce the impact of essay mills simply by drastically reducing their customer base.
As with all ML and AI driven models, good data is the key to reliable modelling and outcomes. Literatu Scribo does the initial writing analysis, delivering both an analysis of the current writing skills of the student and 140 data points to the Originality model. Scribo does this calculation the same way every time, in seconds, as part of the live feedback and guidance capabilities Scribo offers students.
The Originality Index starts building when a student submits at least 250 words, preferably from a supervised test writing event. The more controlled the texts and samples from students, the more accurate the index becomes. The model also caters for unsupervised control samples where ‘take-home’ tasks may be the only available samples. Literatu has plans to open an API to the Originality Index models allowing other systems to quickly build originality checking into their workflow.
Higher education really needs to look into smarter ways to reduce the attraction of essay mills and contract cheating to students. Legislation and the pursuit of highly mobile essay mill opportunists may sound viable, but it will be a challenging and thankless process. The essay supply lines and students who use them should no longer benefit from contract cheating.
Mark Stanley is founder and CEO of Literatu. Scribo is a writing support, improvement and validation platform for schools, colleges and universities. https://www.prestoapp.ai