Individual differences reveal shared computational structure in perceptual decision-making

A six-dimensional representation of individual differences in perceptual decision-making in the Bernoulli Clicks Task links two temporal evidence-weighting dimensions, two choice-outcome history dimensions, plus side bias and signal-to-noise ratio.

• Why it matters: Fixed-architecture models often treat people as mere parameter variations, leaving the mechanistic structure behind behavioral variability unclear.
• What they did: The study learned a low-rank neural network using population geometry to infer shared computational components and participant-specific weights from trial-level choices, then tested out-of-sample prediction across primary and replication datasets.
• The result: The six-dimensional organization replicated in an independent dataset, implying that diverse decision heuristics can be explained as different weightings over a small set of shared computational components.

AI summary of the abstract · may contain errors · not medical advice · read the paper.