Date: Wednesday, 29.11.23 15:20-17:00 CET
Signup: If you would like to attend the talks please register here to get a Zoom link.
Abstract:
Despite our unique individual experiences, we are surprisingly similar in the images we remember and forget. As a result, our memories are highly predictable just from the images we are viewing, and a deep learning neural network (ResMem; Needell & Bainbridge, 2022) can be trained to make these predictions.
In today’s talk, I will describe two cases in which we are able to successfully predict memory in different populations, with different image sets, and with different tasks. First, I will show that our neural network is able to predict people’s memory performance during a freeform visit to the Art Institute of Chicago (Davis & Bainbridge, 2023). As a result, we show that a majority of the variance in what people remember is predictable and external to the observer (and controllable by the researcher or curator). Second, I will show that we can predict memory in as early as 4 years of age (Guo & Bainbridge, in press).
These findings demonstrate that children’s memory systems become more adult-like and predictable at the age of 4, coinciding with hippocampal maturation. In sum, visual memory can be predicted by a neural network across people, tasks, and images, to reveal insights into the development of the memory system and how it functions in the real world.
