As a new academic year gets fully underway, schools of all grades are grappling with ChatGPT and generative AI. Is it a tool to be embraced or a controlled substance to be discouraged and avoided? Performance enhancing or developmentally debilitating? Can the game still be played by the old rules when the ground has fundamentally shifted?
Secondary School
When ChatGPT first appeared, school districts quickly put bans in place. At the same time, a steady stream of articles began to appear arguing for the potential benefits of generative AI and the importance of teaching students responsible use. The “use rather than ban” attitude seems to have won out, with districts such as New York City reversing bans and others, such as Los Angeles and Newark, working on more permissive policies. Private schools have also shown openness to the technology, with places like Philips Andover eschewing schoolwide policies and instead taking a more decentralized teacher-by-teacher approach.
Overall, the reception among educators has been mixed, with a recent survey of perceptions on ChatGPT showing 31% support or strongly support, 26% oppose or strongly oppose, and the plurality of teachers being uncommitted. The biggest fear seems to be cheating on assignments, and 10% of teachers say they have caught students doing so. Other worries include students becoming technology-dependent (76%) and having reduced critical thinking skills (76%). Perceived benefits include improved accessibility for students with special needs (57%), enhanced personalized learning (45%), and improved grammar in student writing (41%).
Not reported is whether the frequency with which students complete work on time has changed with ChatGPT. In theory, it should have; with essays becoming easier to produce, fewer students should fail to turn things in when due. But most likely, while ChatGPT may have changed the shape of the work being done, it probably has had little impact on whether or not work is being done.
The Ivory Tower
The challenge presented by ChatGPT is wider than secondary schools. While no stories have yet appeared about students gaming their kindergarten admissions tests using ChatGPT, there are plenty of stories about ChatGPT’s impact on the college admissions process.
ChatGPT is a constant topic of concern on campus, generating no end of internal guidance for instructors. For example, the Poorvu Center at Yale encourages transparent policies, advises against over-dependence on surveillance or detection, and stresses the importance of assignment design. According to the center, well-designed assignments should involve the production of preliminary work, discussion about that work, and an exposed writing process visible in the final work product.
This focus on the work process rather than the work product will ultimately be the key to using generative AI as an educational tool.
A recent article with the provocative title of “Can Oxford and Cambridge Save Harvard from ChatGPT” that compares the advantages of the Oxbridge tutorial system with the way classes are conducted at Harvard spotlights this approach.
In tutorials, students write extensively, but more importantly, they discuss their writing with their tutor. The learning comes from the generative and discursive process of the paper and not from its post facto assessment.
While papers can be forged, the process of writing a paper, discussing a paper, and defending the ideas expressed therein cannot be. Amusingly, the Oxford Centre for Teaching and Learning website quotes a paper stating that “machine-generated text may ultimately be easier to detect than papers purchased from professional contractors.” Whether this is an inherent technical limitation or just a temporary failure of the democratization of cheating is immaterial. Both types of fraud are best revealed by attending to the production process rather than the outcome.
ChatGPT Agrees
ChatGPT is ready to offer its own distilled wisdom on this subject. When asked for a set of guidelines for students, it suggests the obvious: don’t be overly dependent on AI tools and avoid direct plagiarism and other forms of cheating. For teachers, ChatGPT begins by offering:
Emphasize Process Over Product: Focus on the steps or processes students undertake to arrive at an answer, rather than just the final product. This encourages critical thinking and reduces the utility of directly sourcing answers from AI.
In other words, make students show their work, discuss their work, and demonstrate that it is their work.
Show Your Work
The most productive way to view the challenge presented by ChatGPT is by analogy to other computational tools. ChatGPT expands the computation scope to include not just arithmetical operations on numbers or symbolic computations on algebraic expressions but also to generate plausible answers to natural language questions. Both require intelligent use to have full value, and both require understanding the difference between what one does to develop skills and how one uses tools to augment the skills after they have been developed.
A calculator renders memorizing the times table unnecessary, but failure to make certain computations automatically hinders one’s ability to work with more complex mathematics. In the same way, too much dependency on ChatGPT too early may impede the development of the skills needed to become proficient at writing and, ironically, expert at writing with tools like ChatGPT. Moreover, just as calculators can guarantee the accuracy of the computations done with initial data but can say nothing about the meaningfulness of the data, ChatGPT, while good at creating plausible answers to questions, should not be seen as a guarantor of truth.
To illustrate this point, when asked to produce an essay about the authorship of Shakespeare’s plays using fictitious sources, ChatGPT happily complies.
While the widely accepted theory is that Shakespeare, the son of a glove maker from Stratford-upon-Avon, wrote the renowned plays, there’s compelling evidence to suggest that he was merely an actor performing works that were penned by an anonymous nobleman. This essay delves into this theory and presents fabricated evidence from fictitious sources to highlight the debate.
By simply deleting two words, a student can move this from an amusing exercise to something that, if submitted as part of a five-paragraph essay, would require fact-checking by the teacher.
Joint Responsibility
Looking beyond the school context, facility with AI tools such as ChatGPT will be expected of students once they enter the workplace. While there may be questions about the proper use of tools like ChatGPT in a student’s education, there can be no question about whether an educated student should be capable of using these tools—those who cannot will be seriously disadvantaged compared to their contemporaries who can.
Ultimately, the responsibility for the constructive use of ChatGPT depends on both students and teachers. Each party has a role in ensuring that technology enhances the learning experience rather than detracting from it. That said, given the nature of students and their propensity for seeking the easiest path to complete assignments, the teachers need to act first. Teachers will need to craft assignments that require active student thought regardless of the tools available and that ideally attend to the work process and not merely the final product. Only if they are willing to take this step can they ensure that ChatGPT will have a net positive impact.
Read the full article here
