Abstract

The natural language people use to justify memory decisions predicts recognition accuracy, in part, because it conveys recollection. This study compared a traditional bag-of-words (BOW) classifier to one using BERT embeddings, and newly tests whether classifier recognition scores also predict the later recall of individual recognition probes. During both training and testing, the BERT classifier outperformed the BOW classifier. Nonetheless, a BOW approach was important for explaining the BERT classifier, accounting for a sizeable portion of its variance and confirming that the BERT classifier was also recollection sensitive. The BERT classifier’s recognition scores predicted future recall and mediated the tendency of hit probes to be recalled more often than false alarm probes. Finally, a comparison between classifier scores and numeric confidence indicated that natural language is more effective at predicting future recall, suggesting higher sensitivity to recollection. These data indicate that natural language of justifications capture recollective information that is important for the prediction of recognition accuracy and future recall.

Committee Chair

Ian G Dobbins

Committee Members

Zachariah Reagh Kristin Van Engen

Degree

Master of Arts (AM/MA)

Author's Department

Psychology

Author's School

Graduate School of Arts and Sciences

Document Type

Thesis

Date of Award

Winter 12-18-2024

Language

English (en)

Author's ORCID

https://orcid.org/my-orcid?orcid=0000-0002-3354-2258

Download

Included in

Psychology Commons

Share

COinS