Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session B: Wednesday, August 13, 1:00 – 4:00 pm, de Brug & E‑Hall

Human predictions of event sequences for novel videos

Caroline Lee¹, Dana Lauren Caroline Villamin¹, Paulina Margarita Bertilsson¹, Mariam Aly², Christopher Baldassano¹; ¹Columbia University, ²University of California, Berkeley

Presenter: Caroline Lee

Prior knowledge facilitates predictions in novel situations. Schematic templates for typical sequences of events, called event scripts, support event segmentation and linking of information across related contexts. Here, we present preliminary data comparing recall based on episodic memory and predictions based on event scripts. Participants viewed the first segment of instructional videos, some of which they had previously viewed, and then verbally recalled or predicted subsequent events. We used large language models (LLMs) to quantify the semantic content of the responses, enabling comparison between episodic memory and prediction-based knowledge.

Topic Area: Memory, Spatial Cognition & Skill Learning

Extended Abstract: Full Text PDF