Differential sensitivity to outcome valence reveals two classes of shift behaviour related to exploitation and exploration.

The number of available response options is a central design feature in any experimental psychology paradigm, with binary response paradigms restricting the range of behavioural responses into staying or shifting. Participants (n = 72) played nine blocks of a novel zero-sum game, Dice Dual, while we parametrically manipulated the different number of response options (2, 4, 6) against an unexploitable computerized opponent. When the number of responses exceeded two, this allowed us to identify two types of shift behaviours: between-category shifts relevant to mental model exploitation, and, within-category shifts reflective of random exploration. Between-category shifts were sensitive to outcome valence, exhibiting an increased shift rate when more response options were available after-win, while shift rates were independent of response availability after-loss. In contrast, within-category shifts were only sensitive to response availability, with increased shift rates when more response options were available, irrespective of outcome valence. Identifying multiple, functionally different classes of shift behaviour using non-binary response paradigms allows us to more easily separate attempts at exploitation and exploration. The data also offer more general models of performance regarding sub-optimal decision-making, where negative outcomes constrict the range of factors considered in future mental model updates.