Follow the signal: models of attention, reason, and belief

Attention, reasoning, and belief are central to how we perceive, decide, and collaborate. Though inherently abstract-with no direct physical manifestation these phenomena leave behind observable signals in subtle traces in gaze, language, timing, and interaction. These traces vary across individuals and contexts, yet they offer a window into the underlying cognitive processes. In this thesis, I model the behavioral and linguistic signals that reflect aspects of attentional shifts, expressions of reasoning, and evolving belief states, and investigate how machine learning can be used to detect and interpret them as they arise in everyday settings. First, I focus on moments of inward attention, identifying gaze patterns that predict when participants feel familiarity—even without conscious recall, using eye-tracking during immersive virtual tours. I then analyze written descriptions of three distinct internal attentional states: familiarity, unexpected thoughts, and involuntary memories. Then, I frame the link of probing questions i.e questions that explicitly elicit justifications or clarifications, and their causal utterances as traces of reason as they emerge in group dialogue Next, in the case of belief, I extract explicitly stated propositions from natural dialogue. These structured propositions reflect participants' evolving belief states during a collaborative task. I design and evaluate multiple extraction pipelines, demonstrating the feasibility of tracking belief expression in real time. Finally, I holistically examine how automated systems with noisy data shape downstream performance on collaborative problem-solving detection—a task that inherently reflects attention, belief, and reasoning. I show that, while performance remains comparable across systems, lower fidelity inputs reduce interpretive granularity. In combination, these contributions demonstrate how machine learning can detect the emergence of traces of these phenomena–—transforming these abstract states into observable patterns.

Name thesis:degree_name

Master of Science (M.S.)

Level thesis:degree_level

Masters

Discipline thesis:degree_discipline

Computer Science

Grantor dc:publisher

Colorado State University. Libraries

Year dc:date.issued

2025

Authors dc:creator

• Venkatesha, Videep, author
• Blanchard, Nathaniel, advisor
• Krishnaswamy, Nikhil, committee member
• Sreedharan, Sarath, committee member
• Cleary, Anne, committee member

Statement dc:rights

• Copyright and other restrictions may apply. User is responsible for compliance with all applicable laws. For information about copyright law, please see https://libguides.colostate.edu/copyright.

Language dc:language.iso

eng, English

Identifier URI

https://doi.org/10.25675/3.02152

OAI identifier oai:identifier

oai:mountainscholar.org:10217/241832