Memory is predictable

Despite our unique individual differences, there's a surprising consistency across people in their memories, where we tend to remember and forget the same images. This suggests that certain images are more memorable than others, and this effect is so pervasive that neural networks can predict people's memories based on images alone. In this talk, I will demonstrate how predictable people's memories are, even in real-world scenarios like remembering pieces from an art museum (Davis & Bainbridge, 2023; Chen et al., 2025 preprint) or in which social media posts go viral (Peng & Bainbridge, 2026). I will then present a framework suggesting that images that are easier to process may be those that end up being most memorable (Kramer et al., 2023), and that we may need to move outside traditional Euclidean spaces to understand how we represent our memories (Lee et al., 2024 preprint).

Online: https://utoronto.zoom.us/j/82573199520