De-identified eye-tracking dataset: attention allocation and cognitive effort in GenAI-assisted course selection

Abstract

Derived, de-identified data supporting the manuscript Cognitive Patterns of Multi-Source Integration in GenAI-Assisted Course Selection: An Eye-Tracking and Stimulated Recall Study (Zhang, Wei & Hu, under review). Sixty undergraduates completed a simulated course-selection task on a split-screen interface combining a generative AI chatbot, a course-selection concept map, and a web search engine while eye movements were recorded at 100 Hz. The deposit contains, for pseudonymised participants P01–P60: area-of-interest (AOI) definitions; 11 eye-movement indicators for 8 AOIs (long format) and their participant-level means used as clustering input; K-Means cluster assignments and profile labels; the Eye-tracking Cognitive Effort Indicator (ECEI) with components and weight-sensitivity results; AOI-labelled fixation sequences (185,834 records, with the 39,537 artefact records flagged); transition entropy; all statistical outputs; pseudonymised interview code frequencies for 12 interviewees; analysis scripts and notebooks; and a verification script that reproduces the paper's Table 2, ECEI, sub-AOI and transition results from the deposited files. No names, student numbers, demographic records, raw gaze recordings, chat transcripts, or interview transcripts are included, in accordance with the ethics approval and informed-consent conditions. See README.md for the data dictionary.