A Normative Theory of Decision Making from Multiple Stimuli: The Contextual Diffusion Decision Model

A team of researchers has published a groundbreaking paper on arXiv titled 'A Normative Theory of Decision Making from Multiple Stimuli: The Contextual Diffusion Decision Model.' The work, led by Michael Shvartsman, Vaibhav Srivastava, Narayanan Sundaram, and Jonathan D. Cohen, introduces the CDDM, a computational theory that formally extends the influential Diffusion Decision Model (DDM). While the DDM excelled at modeling simple two-choice decisions, the CDDM tackles the far more complex reality of decisions influenced by dynamically changing, multiple sources of information, grounding its approach in Bayesian inference and a simple neural network architecture.

The CDDM's power lies in its ability to unify several long-standing psychological paradigms under a single model. The researchers demonstrate that the CDDM can successfully account for data patterns in five classic context-dependent tasks: the Flanker, AX-CPT, Stop-Signal, Cueing, and Prospective Memory tasks. Crucially, it can recover consistent qualitative patterns across these diverse tasks using the same set of core parameters. This allows the model to perform normative analyses, exploring optimal policies for response and memory allocation, and provides a robust, generalizable framework for studying how contextual information is integrated into the decision-making process.