“When using AI, we need to be asking ourselves what kind of individuals we want to be and what kinds of teams we want to be part of,” said psychologist Robert Goldstone (Indiana University) during the recent event hosted by the American Academy of Arts and Sciences, How Can AI Impact Human Memories?
Moderated by cognitive scientist Suparna Rajaram (Stony Brook University), this event brought together neuroscientist James DiCarlo (MIT) and Goldstone to discuss the duality of artificial intelligence (AI) use and memory: this technology can help researchers understand how memory works but using AI can also change the way people use memory.
When framing the importance and relevance of the interdisciplinary conversation, Laurie Patton, President of the American Academy, spoke about memory and being human. “There are countless stories where an individual loses their memory and begins to lose their sense of self,” she said. “What memory is, how it develops, and what it means to remember something is still being researched. Advances in AI technology may be changing how this work is done.”
In her opening remarks, Rajaram expanded on the place of AI stating, “the question is no longer whether AI will change our lives; rather how AI will change and is already changing our lives.” She asked, “Can advances in AI help us generate models of human memory and cognition in general to better understand and develop helpful interventions?”
DiCarlo and Goldstone are working to answer these questions.
DiCarlo’s research uses AI models to study and understand the human brain, likening the mechanical systems to biological ones. When asked what good AI technology can contribute to memory and cognition, he said, “Once you have an engineered model of precisely how the neural mechanisms support the cognitive aspect…they allow you to do things that you otherwise couldn’t do.” For example, his team will invert models—thinking about how models can be used to activate neurons deep in the brain and ask the model to design images to do so.
While using AI tools may help us develop objective understandings of memory, the experience of using memory is much more subjective and can be shaped by AI.
Goldstone studies human collective behavior and how people do and should interact with each other to form successful teams. He has extended his research to include teams that form between humans and AI systems. When asked about the collaboration between humans and AI systems and the consequences to memory, Goldstone cited a study showing that individuals who used a large language model when writing an essay had less structured brain activity, poorer recall of essay ideas, and less ownership of ideas compared to people who worked unassisted. “When learning, it’s usually better to retrieve the knowledge on your own…this is one of the reasons why using a writing assistant like a large language model may hurt a person’s memory. People aren’t getting practice retrieving the information on their own.”
As the conversation turned to the pros and cons of using AI technology, Goldstone explained his approach as thinking about AI in terms of whether it makes someone feel bigger or smaller. “I’m not at all attracted to…self-driving cars because I really like driving. If technology takes away my driving…the sphere of my life has been shrunk a little bit.” But there are other times, he continued, where he is programming an experiment and can get a user interface working in a fraction of the time it used to take, making him feel empowered.
When it comes to using AI, both panelists agreed: these tools are important, but using them can lead to cognitive laziness, the urge to forgo critical thinking, and diminished memory retrieval. As this technology proliferates, educators will have to reconsider what knowledge and which skills are valuable and how to pass those on to the next generation.
This conversation was part of the Academy's work on Artificial Intelligence, which includes publications, events, and essays.