Retrieving past experiences to inform novel decisions through a process of cascading episodic sampling

Abstract

We can guide our decisions in novel situations by drawing on our past experiences (episodic memories). While at times we can retrieve relevant episodes via cued recall, other situations may require a process of memory search. But what mechanisms underlie this search? In this work we synthesize six key principles concerning the storage and retrieval of episodic memories, and build on these principles to propose a cognitive mechanism which allows for the retrieval of relevant past experiences through a process of cascading recall. In this process, observing a stimulus triggers the cued recall of a past event. If this memory does not provide sufficient information to warrant a decision, then it next reinstates all the memory's constituent features. These features then form the inputs to sample an additional memory in a subsequent recall step, which in turn reinstates its own features and so forth. This process continues until a suitable past experience is retrieved. We provide empirical support for key predictions of this cascading process through three online experiments in which participants interacted with unfamiliar stimuli. The results indicate that participants rely on cued recall of similar past experiences (experiment 1), and on indirectly related experiences when cued recall is not informative (experiment 2). Additionally, participants were substantially more likely to retrieve a predicted memory, and did so faster, when relying on cued recall versus cascadizng memory search (experiment 3). We conclude by discussing how this cascading recall process bridges several influential models of memory-based decision-making, as well as offering promising directions for future research.