ORCID

https://orcid.org/0009-0009-5127-9324

Keywords

Artificial Intelligence, Comprehension Monitoring, Inconsistency Detection Paradigm, Error Detection in AI-Generated Text, Perceived Authorship, Presentation Speed

Subject Categories

Cognition and Perception | Cognitive Psychology | Psychology

Abstract

The increased us of artificial intelligent (AI) chatbots powered by Generative Pre-trained Transformers has raised concerns about users’ ability to detect errors of AI-generated text. This dissertation examined how readers identify inconsistencies in AI-generated expository passages using the inconsistency detection paradigm from comprehension monitoring research. Across two experiments, it investigated whether perceived authorship and word presentation speed influenced inconsistency detection. Experiment 1 used a 3 (author: human, AI, no author) × 2 (inconsistency type: internal, external) between-subjects design. Perceived authorship did not significantly affect hit rates, sensitivity, or response bias. Exploratory analyses indicated that reading comprehension and working memory predicted error detection. Experiment 2 used a 2 (inconsistency type: internal, external) × 3 (word presentation speed: 100, 300, 600 words per minute [WPM]) between-subjects design to examine effects on inconsistency detection and attitudes toward AI. Competing predictions were derived from system attitudes research, which suggests that faster systems are evaluated positively, and cognitive psychology, which predicts that presentation may impair comprehension. Participants in the 100 WPM condition had significantly higher hit rates and a more liberal response bias than those in the 600 WPM condition, suggesting that slower presentation enhanced inconsistency detection. Presentation speed did not influence sensitivity or attitudes toward AI. Overall, the findings supported the cognitive perspective but not the system attitudes perspective. These results extend comprehension monitoring research by showing how external factors such as speed may shape inconsistency detection and inform chatbot design by demonstrating that speed can affect whether users notice and evaluate errors.

Completion Date

2026

Semester

Summer

Committee Chair

Sims, Valerie

Degree

Doctor of Philosophy (Ph.D.)

College

College of Sciences

Department

Psychology

Format

PDF

Document Type

Dissertation

Language

English

Download

Share

COinS
 

Accessibility Statement

This item was created or digitized prior to April 24, 2027, or is a reproduction of legacy media created before that date. It is preserved in its original, unmodified state specifically for research, reference, or historical recordkeeping. In accordance with the ADA Title II Final Rule, the University Libraries provides accessible versions of archival materials upon request. To request an accommodation for this item, please submit an accessibility request form.