What is it about?
To make sense of the endless stream of events that unfold in front of us, humans need multiple learning abilities: extracting their statistics, predicting their timing, and discovering hidden rules that govern their occurrence. While these learning abilities have historically been studied independently, we have discovered that they are actually deeply interconnected. In practice, we have examined how hundreds of individuals predict future visual stimuli in sequences governed by a hidden rule. We uncovered multiple interactions between statistical inference, temporal prediction, and rule discovery. Most prominently, we found that individuals with longer integration timescales were more likely to discover the hidden rule, and more so for rhythmic sequences. And conversely, individuals who discovered the rule inferred stimulus statistics more accurately. This complex pattern of interaction also emerged in artificial neural networks trained in the same conditions.
Featured Image
Photo by Siora Photography on Unsplash
Why is it important?
Delineating the different facets of human learning is important, but studying how these facets may depend on one another should not be neglected. Understanding the interaction between human learning abilities is needed to improve existing theories of human learning, but also to identify the real-life conditions where these abilities can be expected to work in synergy.
Perspectives
Using a combination of behavioral experiments, quantitative computational modeling of such behavior, together with small neural networks whose structure and function can be selectively manipulated, makes it possible to test and contrast competing theories of human learning.
Lucas Benjamin
Read the Original
This page is a summary of: Sensory integration, temporal prediction, and rule discovery reflect interdependent inference processes, Proceedings of the National Academy of Sciences, February 2026, Proceedings of the National Academy of Sciences,
DOI: 10.1073/pnas.2524629123.
You can read the full text:
Contributors
The following have contributed to this page
