Navigating decision space: Causal structure improves performance in a branching choice task

Abstract

Previous research has shown that the causal structure of events influences how well they are recalled in episodic memory later on. Here, we aimed to investigate whether these effects apply not only to events that are passively observed but also situations directly shaped by an individual’s decisions. We designed a task in which participants had to traverse decision trees of varying causal structure: ‘Coherent’ trees where each decision followed from the consequences of the preceding decision, and ‘fragmented’ trees where each subsequent decision was only statistically (but not causally) contingent on the preceding decision. In a between-subjects experiment, participants first completed an exploration phase in which they had to explore the decision trees without a specific goal; in a subsequent search phase, they had to reach a target outcome in as few attempts as possible. Analyses of participants’ performance showed that those in the coherent group required significantly fewer attempts to reach a correct outcome than those in the fragmented group. A follow-up experiment surprisingly found that the advantage of causal structure does not depend on episodic memory: Removing the exploration phase barely diminished the positive effect causal coherence had on participants’ performance. In further follow-up experiments without an exploration phase, neither the additional removal of ‘process images’ that show how a choice leads to an outcome, nor the removal of text labels describing decisions, was individually sufficient to equalize performances. Only when both were eliminated at once did participants perform equally well on coherent and fragmented trees. This indicates that cues relating to causal mechanisms (images) and predictive cues (text) each facilitate goal-directed decision making without relying on extensive learning, and that only the absence of both is sufficient to suppress the advantage causal structure provides.