Observers exhibit greater caution and faster evidence accumulation when identifying joint versus individual actions

Coordination requires anticipating and adapting to others' actions in real time. While research has shown that observers predict observed solo actions, little is known about how predictive processes change when observing joint actions. We tested whether, for joint actions, observers form (1) stronger predictions because they expect others to act more reliably due to social considerations (e.g., commitment), or (2) weaker predictions because they expect coordination to introduce additional sources of error (e.g. due to miscommunication). To this end, we presented participants with image sequences depicting either one or two actors building a block tower. The solo actor added blocks, alternating between her left and right hands, while the joint actors took turns adding blocks with one hand each. Participants made speeded responses to identify the side from which the block came. Of interest was how accurately participants identified the block's side when it violated the alternating pattern, with errors indicating anticipation for the alternating pattern to continue. Participants were less likely to make anticipatory errors on violation trials for joint actions, indicating weaker action predictions (Study 1). Hierarchical drift-diffusion modelling (Study 2) revealed that this was due to participants accumulating sensory evidence more rapidly and responding more cautiously on joint action trials, relying less on their prior expectations about an action's identity and more on incoming sensory information. This strategy might serve to enable observers to flexibly update their predictions about unfolding actions in the face of coordination-induced errors.