Optimal allocation of attentional resource to multiple items with unequal relevance

Abstract: In natural perception, different items (objects) in a scene are rarely equally relevant to the observer. The brain improves performance by directing attention to the most relevant items, for example the ones most likely to be probed. For a general set of probing probabilities, it is not known how attentional resources should be allocated to maximize performance. Here, we investigate the optimal strategy for allocating a fixed resource budget E among N items when on each trial, only one item gets probed. We develop an efficient algorithm that, for any concave utility function, reduces the N-dimensional problem to a set of N one-dimensional problems that the brain could plausibly solve. We find that the intuitive strategy of allocating resource in proportion to the probing probabilities is in general not optimal. In particular, in some tasks, if resource is low, the optimal strategy involves allocating zero resource to items with a nonzero probability of being probed. Our work opens the door to normatively guided studies of attentional allocation.