Self-Regulated Reading with AI Support: An Eight-Week Study with Students

TL;DR

This study analyzes how college students use AI chatbots to support self-regulated reading over eight weeks, revealing a strong preference for comprehension-oriented prompts, with visible cognitive progression within sessions that is often truncated by the assignment's minimum interaction. The authors develop a four-theme coding schema (Decoding, Comprehension, Reasoning, Metacognition) to analyze 838 prompts from 239 sessions, finding an intention-behavior gap where students recognize effective prompting but rarely implement it, and an emergent pattern of reading through AI rather than with it. Across eight weeks, engagement patterns remained stable, with substantial individual differences in how students used AI and how much they wrote per prompt, highlighting the influence of efficiency and time pressures on engagement. The paper discusses design implications for AI reading tools, suggesting scaffolds that promote sustained cognitive progression, surfaced unexplored content, and metacognitive reflection, along with pedagogical strategies to counteract epistemic risks and overreliance on AI.

Abstract

College students increasingly use AI chatbots to support academic reading, yet we lack granular understanding of how these interactions shape their reading experience and cognitive engagement. We conducted an eight-week longitudinal study with 15 undergraduates who used AI to support assigned readings in a course. We collected 838 prompts across 239 reading sessions and developed a coding schema categorizing prompts into four cognitive themes: Decoding, Comprehension, Reasoning, and Metacognition. Comprehension prompts dominated (59.6%), with Reasoning (29.8%), Metacognition (8.5%), and Decoding (2.1%) less frequent. Most sessions (72%) contained exactly three prompts, the required minimum of the reading assignment. Within sessions, students showed natural cognitive progression from comprehension toward reasoning, but this progression was truncated. Across eight weeks, students' engagement patterns remained stable, with substantial individual differences persisting throughout. Qualitative analysis revealed an intention-behavior gap: students recognized that effective prompting required effort but rarely applied this knowledge, with efficiency emerging as the primary driver. Students also strategically triaged their engagement based on interest and academic pressures, exhibiting a novel pattern of reading through AI rather than with it: using AI-generated summaries as primary material to filter which sections merited deeper attention. We discuss design implications for AI reading systems that scaffold sustained cognitive engagement.

Figures (8)

• Figure 1: Distribution of prompts per reading session ($N$ = 239 sessions). The assignment required a minimum of three conversational turns per session. A large majority of sessions (72.0%) contained exactly three prompts, while 28.0% exceeded this minimum. The orange bar indicates the minimum requirement.
• Figure 2: Distribution of prompt codes across 838 prompts. Summary (30.3%), Inference (23.2%), and Explanation (21.9%) were the most frequent codes, together accounting for 75.4% of all prompts. Colors indicate higher-level cognitive themes.
• Figure 3: Distribution of prompts by cognitive theme ($n$ = 822 prompts). Comprehension-focused prompts were most prevalent (59.6%), followed by Reasoning (29.8%), Metacognition (8.5%), and Decoding (2.1%). Themes are ordered from lower-order (left) to higher-order (right) cognitive engagement.
• Figure 4: Distribution of prompt length (in words) across four cognitive themes. Data were trimmed at the 99th percentile to exclude outliers. Significance brackets indicate pairwise comparisons from a linear mixed-effects model. Metacognitive prompts were significantly longer than Comprehension, Reasoning, and Decoding prompts. **p < .01.
• Figure 5: Theme distribution by prompt position within reading sessions. The first prompt was predominantly comprehension-focused, with reasoning prompts increasing substantially by third prompt. Significance brackets indicate pairwise chi-square comparisons with Bonferroni correction; ** $p$ < .01, *** $p$ < .001.