OpenAI has launched a new “study mode” feature on ChatGPT to help encourage “cognitive learning”, in an attempt to address concerns that widespread AI use is damaging students’ education.
With some studies estimating nine in 10 students are using AI in their assessments less than three years since large language models were popularised, the tech giant said the “study mode” option will encourage students to “work through questions step by step, instead of just getting an answer”.
Academics from Stanford University have helped develop the feature that OpenAI said “combines Socratic questioning, hints, and self-reflection prompts to guide understanding and promote active learning” and can be tailored by students to reflect their attainment level.
While the tool is unlikely to prevent students’ cheating, its launch reflects a pressure on AI companies to introduce some guard rails on their products.
Robbie Torney, senior director of AI Programs at Common Sense Media, which worked alongside OpenAI to develop the tool, said that study mode “encourages students to think critically about their learning…instead of doing the work for them”.
“Features like these are a positive step toward effective AI use for learning. Even in the AI era, the best learning still happens when students are excited about and actively engaging with the lesson material.”
The tool focuses on improving “cognitive outcomes and retention of information” and can help “hard bake” good educational practice into student learning, James Donovan, head of education and cognitive outcomes research at OpenAI, told Times Higher Education.
He explained that the Stanford academics had worked with the company to develop a pedagogical framework to “reinforce information, enable active cognitive engagement, help support scaffolding for retention purposes, and ensure there’s more of an internalisation process being triggered”.
While the model goes “some way” towards discouraging student cheating, a “holistic” approach is needed from universities, Donovan said.
“The function of study mode offsets a little bit the risk of cheating in that it forces cognitive engagement. The whole point of it is to use Socratic dialogue and to make sure that people are actually doing what they’re doing.
“But the long-term goal of addressing cheating in the classroom is probably more of a holistic thing that needs to happen with departments and with institutions, beyond just a single feature.”
Donovan said OpenAI is working on other features to help detect AI cheating but “we can’t go all in on those solutions yet because the false positive rate is too high”.
Ultimately, he said that OpenAI hopes that the “thrill” of learning and achievement could be enough to pull students away from using AI for “productivity” purposes, and instead become a model based on self-improvement.
“Our hope is once you start using study mode, there is a feedback loop where you internalise information better, and you’ll see that actually you learn faster.
“Even though it might feel more efficient to just chuck stuff into ChatGPT and get a huge readout, because you don’t retain that necessarily as well or internalise that as well it might take longer for you to actually get to the point that you’re performing natively, outside of ChatGPT, at the level that you want,” Donovan said.
“Whereas I think the more people use study mode, the more confident they’ll get with the material, and the closer to a sense of mastery.”
juliette.rowsell
