We compare storytelling in GPT-3.5, a recent large language model, with human storytelling. We hypothesized that GPT differs from humans in the kind of memories it possesses, and thus could perform differently on tasks influenced by memory, such as storytelling. We used an existing dataset of human stories, either recalled or imagined (Sap et al., 2022), and generated GPT stories with prompts designed to align with human instructions. We found that GPT's stories followed a common narrative flow of the story prompt (analogous to semantic memory in humans) more than details occurring in the specific context of the event (analogous to episodic memory in humans). Furthermore, despite lacking episodic details, GPT-generated stories exhibited language with greater word affect (valence, arousal, and dominance). When provided with examples of human stories (through few-shot prompting), GPT was unable to match its stories' narrative flow or affective aspects with human stories.

Contrary to common intuition, groups of people recalling information together remember less than the same number of individuals recalling alone (i.e., the collaborative inhibition effect). To understand this effect in a free recall task, we build a computational model of collaborative recall in groups, extended from the Context Maintenance and Retrieval (CMR) model which captures how individuals recall information alone (Polyn, Norman, & Kahana, 2009). We propose that in collaborative recall, one not only uses their previous recall as an internal retrieval cue, but also listens to someone else's recall and uses it as an external retrieval cue. Attending to this cue updates the listener's context to be more similar to the context of someone else's recall. Over an existing dataset (Gates, Suchow, & Griffiths, 2022), we show that our model successfully captures the collaborative inhibition effects, as well as additional recall patterns such as recency and semantic clustering effects.