Proceedings of the Annual Meeting of the Cognitive Science Society

Humans must often make decisions in temporally autoregressive environments (e.g., weather, stock market). Here, current states of the environment regress on their previous states (either across consecutive timesteps or from several timesteps back in a patterned fashion). The current work investigates people’s abilities to utilize previous states of autoregressive sequences as cues to its current state. In Experiment 1 we determine whether utilization of autoregressions reduces as the temporal distance of the predictive timestep increases; and in Experiment 2 we explore whether participants’ utilization of previous timesteps in predictions compete such that they reduce utilization of one timestep when increasing utilization of another timestep. We also fit data from both experiments with a trial-by-trial decision model. Overall, we find that participants significantly reduced utilization of a cue with its increased temporal distance. However, we obtained less conclusive results on competition among timestep cues. These results can explain people’s predictions in sequential decision tasks (e.g., their tendencies to perceive clumpiness in random environments).