Decision-Making in Drosophila with Two Conflicting Cues

Experimental data has revealed that the decision behavior of Drosophila is almost linear when facing a single cue. However, when two conflicting cues are presented, the decision behavior of Drosophila becomes winner-takes-all. We propose a connectionist model to elucidate the underlying computational mechanism.We consider two neural states, representing different action choices, compete with each other through mutual inhibition. They receive inputs from Mushroom Body when conflicting information arise. The role of Mushroom Body is to average out temporal noises in external inputs, so that subtle differences between two conflicting cues can be identified, leading to higher discrimination accuracies. Our model successfully describes the experimental findings.