1 July 2026, by Uğur Turhan

A new study by Sebastian, in collaboration with Romane Cecchi and Stefano Palminteri, was published in the journal Nature Communications.

They investigate how attention biases the perception of option values in human reinforcement learning. A core question was to what extent attention might be the mechanism underlying well-established context effects in learning.

To achieve this, they conducted three experiments with a total of 105 participants and tracked their gaze via eye-tracking. Based on this data, they developed an extended model of reinforcement learning in which gaze duration directly modulates the absolute value of an option. This new model provides more precise predictions, including the aforementioned context effects, than classical approaches that ignore attentional processes, highlighting the influence of attention in value computation.

Link to the full article.