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

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

Conceptual priorities shape individual gaze patterns during naturalistic visual attention

Caroline E. Robertson¹, Katherine Packard, Amanda J Haskins²; ¹Dartmouth College, ²University of California, San Diego

Presenter: Caroline E. Robertson

In daily life, we readily recognize people, places, and objects (e.g., “soldier,” “stadium,” “flag”), as well as the conceptual links between them (e.g., “patriotism”). Here, we show that conceptual-level information shapes individuals’ gaze patterns in a naturalistic eye tracking paradigm. Participants (N = 61) explored a large set of real-world photospheres (N = 100) in headmounted VR while their gaze was continuously monitored using in-headset eye tracking. To assess the informational priorities guiding individual differences in gaze, we leveraged the embedding spaces of large vision and language models. We found that individually-specific gaze patterns across diverse real-world photospheres can be captured by a large language model (LLM) that encodes abstract relationships beyond the visual image content. We demonstrate that the embedding spaces of language and vision models explain unique variance in gaze behavior, and that LLM-based models capture individually specific attentional priorities. These results highlight a new dimension of human selective attention: namely, the influence of individuals’ unique conceptual-level information seeking priorities.

Topic Area: Object Recognition & Visual Attention

Extended Abstract: Full Text PDF