A. Shenhav

Misspecified models create the appearance of adaptive control during value-based choice

The paper reanalyzes data from a decision-making study and challenges the claim that controlled threshold adjustments explain adaptive choice behavior, attributing findings to confounds and alternative models.

Disentangling choice value and choice conflict in sequential decisions under risk

This study disentangles choice value and conflict in sequential decisions under risk using a pig dice game paradigm and cognitive modeling.

Humans can navigate complex graph structures acquired during latent learning

The paper shows humans can infer graph structures from disjoint experiences and link this ability to model-based reinforcement learning.

Rats exhibit similar biases in foraging and intertemporal choice tasks

This study shows rats display similar time preferences in foraging and intertemporal choice tasks, supported by a quasi-hyperbolic discounting model.

The Eighty Five Percent Rule for Optimal Learning

The paper identifies an optimal training error rate of 15.87% (85% accuracy) for learning algorithms in binary classification tasks, applicable to both AI and biological neural networks.

Habits without Values

This paper develops a computational model of habits based on value-free mechanisms, showing it accounts for habit-related behaviors like insensitivity to outcome devaluation and perseveration.