A Spotlight on AI and its Academic Applications (from Nefarious to Benign) by: Margaret M. Kelly
Artificial intelligence is a realm of computer science that aims to create software with humanlike capacities for thought, learning, and adaptation. At the core of AI is machine learning, whereby algorithms are used to help computers learn from massive data sets, performing millions of repetitions instantly to decipher and react to human inputs.
Modern AI Chatbots use Natural Language Processing, which allows computers to process and synthesize human language. In broad strokes, Chatbots use wide-ranging, complex language models to predict what word should come next in a sequence. Given a particular prompt, the Chatbots simulate knowledge and conversation by stringing words together that, probabilistically, make sense in that order. These Chatbots are continually learning and adapting – every time a mistake is identified, the Chatbot learns from it, becoming progressively more sophisticated.
Chatbots have sweeping applications, mostly involving their specialized capacities to perform routine tasks with lightning-like efficiency – so they’re often used for resource optimization/cost reduction functions. In the academic sphere, Chatbot applications are more controversial, and have been cast as a looming threat to academic integrity. Repeatedly, students and researchers alike have been accused of leaning on bots to ghostwrite work.
What has been most concerning is the apparent ability of bot-generated work to fly under the radar, undetected by historically useful and accurate plagiarism detectors. And the bot-generated work is passing – one Wharton professor recently fed a Chatbot final exam questions and, stunningly, the bot passed. Iterations of ChatGPT have also successfully passed the Uniform Bar Exam, with the most recent version, GPT-4, scoring in the top 90%.
In the above cases, Chatbots were ethically engaged, but there have also been reports of rampant cheating across secondary and postsecondary institutions. In a recent survey of 1,000 college and graduate students, 43% of students surveyed indicated they had experience with AI, and 22% of all students surveyed confessed to using AI to complete a graded assignment. When asked whether using AI to complete coursework was cheating, 60% of all applicants responded that it was, but another 15% asserted they’d continue to use it anyway. Nearly one-third of students surveyed also said their teachers had banned students from using AI for coursework, while nearly two-thirds reported teachers have not addressed the use of AI in any clear or consistent way. While the majority of students indicated they had no intention of cheating with AI, these results nonetheless indicate that AI has infiltrated academic experiences in a big way, and educators are scrambling to catch up.
Some accounts are admittedly grim – one teacher distraught about rampant Chatbot use said, “It’s just about crushed me. I fell in love with teaching, and I have loved my time in the classroom, but with ChatGPT, everything feels pointless.” But others take a more optimistic approach to confronting the specter of AI in academics. In January of this year, for example, NPR ran an article detailing one teacher’s approach to students who repeatedly attempted to or succeeded in passing off bot-generated work as their own. The teacher identified his class profile and went with the resistance, so to speak, integrating bots into coursework and removing the taboo around using them, and enhancing the learning experience for students with a newly-leveled playing field.
Regardless of their capacity for positivity, educators have been grappling with the alarming influence of AI on student work in multiple ways. Some have reduced focus on take-home assignments, opting instead to evaluate supervised work, effectively removing the bots from consideration. This approach is all-consuming for teachers, and naturally diminishes the time students spend developing their writing skills. Other approaches involve oral exams (suffering from the same problem), and graded practicums, where students engage in practical applications of what they’re learning — there should likely be more of these in schools, anyway.
The most promising method of neutralizing AI in academic work, though, is simply to filter it out. There are now several AI detection software options on the market, with the front-runners demonstrating remarkable accuracy in identifying bot-generated or even bot-influenced work, and they’re getting better every day. As recently as a few months ago, our team at Advantage Ivy Tutoring conducted trial runs with the leading software in the space, and found it pretty easy to circumvent the filters by simply shifting structure, syntax, and diction, for instance. When we tested the most competitive systems on the market last week, the software caught us every time.
Because most AI-detection software runs a version of AI, we suspect development in this sphere will parallel ever-advancing Chatbot capacities. It’s possible that offices of undergraduate admissions, for example, will establish a filtering process by which every application submitted has to pass a basic AI screen before even making it through the virtual doors. In terms of deterrence, it’s also important for students to note that most degrees from postsecondary educational institutions are, at any point in time, revokable. So, even if their bot-generated work is passing without detection at the moment, that might not be the case for long, and any student experimenting with outsourcing their academic work to AI could therefore be in jeopardy of facing serious (and embarrassing) sanctions down the road.
Ultimately, out of the morass of conflicting reports – some positive, others downright dismal – one thing seems certain: the age of AI in education is upon us. While the specter of AI cheating is real, many students use AI for more benign purposes, like recreation, researching what colleges to apply to, or essay tutoring. So, it’s incumbent on engineers and educators to keep up and adapt – to learn to filter it while also joining it in a cooperative academic effort that could (hopefully) lead to an even better educated, efficient, and well-integrated tomorrow.
Margaret M. Kelly is an educational consultant and application strategist specializing in the college essay at Advantage Ivy Tutoring, where she works with a bespoke team to help qualified students earn spots at the nation’s top colleges and universities. Margaret fell in love with language and learning at Phillips Academy, Andover, after which she earned an A.B. with honors in Political Theory from Princeton University; a J.D. from Virginia Law; and an M.F.A. in Poetry and Literature from the Warren Wilson Program for Writers. She has published flash-fiction, poetry, and legal non-fiction. She lives on a farm in Virginia with her shepherd dog, Huck.
Thank you,
Glenda, Charlie and David Cates
