Generative AI summarization tools and deep reading comprehension among Chinese university students: the mediating role of elaborative processing and the moderating role of reading motivation
- 1Executive Office, Changchun Finance College , ,
Abstract
Generative AI tools that condense texts into summaries have become a routine part of student reading routines, and educators increasingly worry that convenient summaries displace the effortful processing on which deep comprehension depends. This study drew on cognitive offloading theory and the levels-of-processing framework to test a moderated mediation model in which elaborative processing statistically links generative AI summarization tool use to deep reading comprehension, and reading motivation moderates the negative association between tool use and processing. Undergraduate students (N = 608; 53.6% female; mean age 19.9 years) from three universities in Changchun, China, completed a six-item scale of generative AI summarization tool use, an eight-item elaborative processing scale adapted from the deep-processing subscale of the R-SPQ-2F, a seven-item reading motivation measure adapted from the Motivation for Reading Questionnaire, and a researcher-developed deep reading comprehension test. Scale reliabilities were satisfactory (Cronbach alpha = 0.88, 0.87, and 0.86). Tool use correlated negatively with elaborative processing (r = −0.41) and comprehension (r = −0.24); elaborative processing correlated positively with comprehension (r = 0.59). PROCESS analyses (Model 7) with 5,000 bootstrap resamples showed that tool use negatively predicted elaborative processing (b = −0.310, p < 0.001) and that this association weakened as reading motivation rose (interaction b = 0.433, p < 0.001). Elaborative processing, in turn, predicted comprehension (b = 10.097, p < 0.001), whereas the residual direct path of tool use did not reach significance. The indirect effect of tool use on comprehension through elaborative processing was significant and negative at low (−6.142, 95% BC CI [−7.353, −5.091]) and average (−3.130, [−3.851, −2.497]) reading motivation but nonsignificant at high reading motivation (−0.117, [−0.876, 0.629]); the index of moderated mediation was 4.374 [3.428, 5.402]. These results are consistent with an account in which elaborative processing statistically links summary reliance to lower deep reading, with reading motivation associated with a weaker negative indirect association; given the cross-sectional design, the modeled paths represent associations rather than established causal effects. They suggest that motivation-oriented reading instruction should accompany, rather than follow, the integration of AI tools into academic reading.
Related
- Multidimensional Assurance of an Electronic Health Record Embedded Generative Artificial Intelligence Summarization Tool
- Visual–Semantic Processing and Reading Comprehension in Chinese Deaf and Hard-of-Hearing Students: Roles of Updating, Semantic Radical Awareness, and Daily Sign Language Use
- A Pragmatic Randomized Trial of an EHR-Integrated Generative AI Chart Summarization Tool for Ambulatory Clinicians
- Electrophysiological Indices of Hierarchical Speech Processing Differentially Reflect the Comprehension of Speech in Noise
- Psychometric validation of the Chinese generative artificial intelligence acceptance scale among university students
BibTeX
@article{bai2026generative,
title = {{Generative AI summarization tools and deep reading comprehension among Chinese university students: the mediating role of elaborative processing and the moderating role of reading motivation}},
author = {Ruyu Bai},
journal = {Frontiers in psychology},
year = {2026},
month = oct,
doi = {10.3389/fpsyg.2026.2002898},
url = {https://europepmc.org/article/PMC/PMC13719582},
note = {CC BY},
}
RIS
TY - JOUR
TI - Generative AI summarization tools and deep reading comprehension among Chinese university students: the mediating role of elaborative processing and the moderating role of reading motivation
AU - Ruyu Bai
PY - 2026
DA - 2026/10/16
JO - Frontiers in psychology
DO - 10.3389/fpsyg.2026.2002898
UR - https://europepmc.org/article/PMC/PMC13719582
AB - Generative AI tools that condense texts into summaries have become a routine part of student reading routines, and educators increasingly worry that convenient summaries displace the effortful processing on which deep comprehension depends. This study drew on cognitive offloading theory and the levels-of-processing framework to test a moderated mediation model in which elaborative processing statistically links generative AI summarization tool use to deep reading comprehension, and reading motivation moderates the negative association between tool use and processing. Undergraduate students (N = 608; 53.6% female; mean age 19.9 years) from three universities in Changchun, China, completed a six-item scale of generative AI summarization tool use, an eight-item elaborative processing scale adapted from the deep-processing subscale of the R-SPQ-2F, a seven-item reading motivation measure adapted from the Motivation for Reading Questionnaire, and a researcher-developed deep reading comprehension test. Scale reliabilities were satisfactory (Cronbach alpha = 0.88, 0.87, and 0.86). Tool use correlated negatively with elaborative processing (r = −0.41) and comprehension (r = −0.24); elaborative processing correlated positively with comprehension (r = 0.59). PROCESS analyses (Model 7) with 5,000 bootstrap resamples showed that tool use negatively predicted elaborative processing (b = −0.310, p < 0.001) and that this association weakened as reading motivation rose (interaction b = 0.433, p < 0.001). Elaborative processing, in turn, predicted comprehension (b = 10.097, p < 0.001), whereas the residual direct path of tool use did not reach significance. The indirect effect of tool use on comprehension through elaborative processing was significant and negative at low (−6.142, 95% BC CI [−7.353, −5.091]) and average (−3.130, [−3.851, −2.497]) reading motivation but nonsignificant at high reading motivation (−0.117, [−0.876, 0.629]); the index of moderated mediation was 4.374 [3.428, 5.402]. These results are consistent with an account in which elaborative processing statistically links summary reliance to lower deep reading, with reading motivation associated with a weaker negative indirect association; given the cross-sectional design, the modeled paths represent associations rather than established causal effects. They suggest that motivation-oriented reading instruction should accompany, rather than follow, the integration of AI tools into academic reading.
SN - 1664-1078
ER -