On February 23rd, the American Academy of Arts and Sciences convened a public program titled How Can AI Impact Human Memories? The event brought together leading scholars to examine how artificial intelligence (AI) is reshaping not only the storage and retrieval of information, but also the nature of human cognitive function and memories. Rather than debating if AI will influence daily life, the discussion centered on how it will do so and its potential impacts on research, education, and society.
Suparna Rajaram, PhD, a SUNY distinguished professor of Psychology at Stony Brook University, moderated the conversation. She is well known in the FABBS community, elected as a Fellow to several FABBS member societies, including the American Psychological Association (APA) and Association for Psychological Science (APS).
The forum’s guests featured James (Jim) DiCarlo, PhD, a neuroscientist at Massachusetts Institute of Technology (MIT) known for his work on visual processing and computational models of the brain, and Robert Goldstone, PhD, Indiana University Bloomington, a cognitive scientist whose research explores learning, categorization, and collective intelligence. Each speaker offered a distinct outlook on how AI may shape human cognition, memory, and learning.
DiCarlo provided an engineering and neuroscience perspective on the relationship between AI and human intelligence. He highlighted how machine learning models can support intelligence through computational representations of sensory processing. DiCarlo shared research showing how vision models using AI can help scientists better understand how the brain processes visual information and may even help unlock and modulate the creation of new neurons in the brain. He took an optimistic view, suggesting that AI systems may eventually replicate aspects of human functionality in ways that benefit society.
DiCarlo described memory as an adaptive system that has helped humans survive over time. He explained that tools such as writing reduced the need to memorize large amounts of information, which allowed humans to reorganize memory use toward higher-level thinking and efficiency. He suggested AI may play a similar role by reducing certain cognitive burdens and enabling new forms of intellectual growth.
Goldstone focused more on the social and cognitive implications of AI. He cited research showing that individuals who used large language models (LLMs) to write essays demonstrated poorer recall, less ownership of their work, and reduced brain activity compared to those who wrote without AI assistance. Goldstone warned that AI can blur authorship in memory, making it difficult for people to remember what ideas they generated themselves. Because of this, he stressed the importance of retrieval-enhanced learning, a form of active recall that strengthens memory more effectively than simply reviewing or re-exposing oneself to information.
Goldstone also raised concerns that society may underestimate AI’s potential to reshape people’s sense of purpose and daily lives. He discussed potential risks associated with AI systems, including hallucinations, bias in generated language, and an overemphasis on prediction rather than interaction and categorization. At the same time, he expressed hope that AI could be designed to enhance cognitive function rather than diminish it, allowing students to expand their capacity to learn. He suggested that educational practices may need to evolve, mentioning alternatives such as oral examinations or in-class quizzes to reduce AI-assisted cheating to better evaluate student understanding.
Both speakers concluded on a note that balanced caution with optimism. DiCarlo and Goldstone agreed that while the future of AI is uncertain, there is a real opportunity to guide its development in ways that align with human goals and values.
