Undergraduate Research Day 2024
Permanent URI for this collectionhttp://hdl.handle.net/1903/31825
Browse
Search Results
Item Using LLMs to score Hierarchical Organization of Narrative Recall(2024) You, Hannah; Worthy, Jaylen; Rickles, Ben; Bolger, DonaldResearch has suggested that the ability to comprehend narratives relies on hierarchical organization of memory. Disorders like dementia can be detected by measures of narrative recall, however these often rely on human graders. Here we explored how LLMs can be leveraged to measure hierarchical organization of narrative recall (HONR). We had participants (n=4) recall two, 45-50 sentence stories. We then submitted the recalls (n=8) to a two-part process where we first asked chatGPT to connect the sentences in the recall to the sentences in the actual stories. Next we took the output from ChatGPT and devised two measures of HONR: Event Bunching and Linearity as proxies for Local and Global Coherence, respectively. We compared a sample of human scored responses to those semi-automated responses and found a high interrater reliability, suggesting LLMs can reliably answer questions about narrative structure.Item Effects of Presentation Order and Delay on Recall of Narrative Details(2024) Worthy, Jaylen; Wokoma, Sotonye; Canter, Allison; Chavan, Sidhath; Harris, Noelle; Fatima, Noorain; You, Hannah; Rickles, Ben; Bolger, DonaldRecalling stories relies on episodic memory. According to the Process model of memory (Hasson, et al., 2015), there is no important difference between long term and short term episodic memory. This means that during narrative comprehension, online encoding and subsequent consolidation are part of the same process. To test this theory, we attempted to disrupt working memory (WM) after a story was heard and test whether this affected recall of that story. We hypothesized that in contrast to the Process theory, reconstructive processes relying on WM might be used to recall stories after the story is heard. Furthermore, if this were the case this would be more likely to occur when attempting to recall details of a scrambled story. We had participants listen to a series of brief narratives. Half of the stories are regular stories while the other half have a scrambled order. There were 6 blocks with different stories in each block. Narratives were all between 45-57 sentences. In a between-subjects design, one group (immediate, n=11), listened to stories and then went right into recall, while another group (delay, n=10) listened to the stories and then were asked to complete a complex spanWM task (strenuous delay) before completing recall. Preliminary results show that while encoding was made more difficult by scrambling the order of the story supported by worse recall scores, distracting subjects between the story and the recall did little to affect recall performance for both types of stories. This means that encoding of story information is likely happening online, and not using WM after the story is over.