Researchers Uncover Similarities Between Human and AI Learning

Strategic Insights -

As humans, we learn and adapt to new situations in distinct ways, often incorporating both quick, flexible learning and incremental learning to improve our understanding over time. New research at Brown University has found striking similarities between human learning and artificial intelligence learning, shedding light on the complexities of human cognition and the development of more intuitive AI tools. Led by Jake Russin, a postdoctoral research associate in computer science, the study found that by training an AI system to integrate two types of learning – flexible and incremental learning modes – it interacted similarly to working memory and long-term memory in humans.

Key Takeaways:

  • Researchers at Brown University discovered that humans and AI integrate both flexible and incremental learning modes, with the two types interacting similarly to human working memory and long-term memory.
  • The study found that AI systems can perform in-context learning after meta-learning through multiple examples, revealing a trade-off between learning retention and flexibility.
  • The research team, led by Jake Russin, used meta-learning to tease out key properties of the two learning types and demonstrated that humans and AI share similar dynamics between the two learning modes.
  • The team's work suggests that the two learning modes interact similarly to human working memory and long-term memory, resolving the ambiguity in how humans and AI integrate the two learning types.
  • The study was published in the Proceedings of the National Academy of Sciences and provides new insights about human learning and the development of intuitive AI tools, particularly in sensitive domains such as mental health.
  • The research was supported by the Office of Naval Research and the National Institute of General Medical Sciences Centers of Biomedical Research Excellence.

Statistics:

  • The researchers used meta-learning to train an AI system that integrated two types of learning – flexible and incremental learning modes – and found that the AI system's ability to perform in-context learning emerged after it meta-learned through multiple examples (12,000 tasks).
  • The AI system successfully identified new combinations of colors and animals after meta-learning through 12,000 similar tasks, demonstrating the potential of AI to learn from multiple examples.
  • The study found that human working memory and long-term memory interact similarly to the AI system's integration of flexible and incremental learning modes, with quicker, flexible in-context learning arising after a certain amount of incremental learning has taken place.
  • The research team's work showed that analyzing strengths and weaknesses of different learning strategies in an artificial neural network can offer new insights about the human brain.
  • The study's findings have implications for the development of intuitive and trustworthy AI tools, particularly in sensitive domains such as mental health.

Sources:

  • "Researchers Uncover Similarities Between Human and AI Learning" (Brown University news article, Sept. 4, 2025)
  • Proceedings of the National Academy of Sciences (PNAS)
  • Office of Naval Research
  • National Institute of General Medical Sciences Centers of Biomedical Research Excellence