Reveal the Distractions of the Irrelevant Features using a Neurobiologically Plausible Cognitive Robotics Model

In complex scenes, human perform perception-action tasks effortless but may make mistakes. To better understand the interplay between attentional and action selection, we extended a neurobiologically plausible cognitive robotics model to simulate human behavior, to provide the underlying mechanisms of visually-guided action. Specifically, we studied the pertaining influence of the irrelevant features (IFs) on search performances, modelled the initiation latency (IL) and maximum deviation (MD) measures of a color-oddity choice reaching task. The size of the items as IF, irrelevant to the task-defining dimension, was found to be coinciding with multiple items, produced proximity and similarity grouping effects affecting the behaviors. Our model effectively explains the assertions that 1) IF distracted the search based on the proximity and similarity grouping effects, 2) proximity affected both IL and MD representing the brain’s early and late selection activities, respectively, and 3) similarity only affected the MD, reflecting late selection.