AI evidence extraction

Towards an AI Buddy for every University Student? Exploring Students' Experiences, Attitudes and Motivations towards AI and AI-based Study Companions

Judit Martínez Moreno, Markus Christen, Abraham Bernstein · arXiv (Cornell University) · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

8/10

Relevance

3/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

Methodology & findings

Study design

Cross-sectional survey design.

Sample

N = 926, 6 groups

Primary method

Quantitative data analyzed using Jamovi (version 2.3.21.0), a statistical software package based on R. Methods employed: (1) Descriptive statistics (means, standard deviations, frequency distributions) for digital literacy variables; (2) Inferential statistics: chi-square tests and analysis of variance (ANOVA) to examine differences between demographic variables (gender, faculty affiliation); (3) Pearson correlation analyses to explore relationships between key variables with significance threshold p < 0.05; (4) Effect sizes interpreted using Pearson's r. Qualitative data from open-ended responses analyzed using R text analysis capabilities to identify recurring themes and patterns.

Main result

The study found that "96.9% of participants indicated that they had already interacted with AI tools," with AI used primarily for study-related purposes (82.4%). Students "expressed strong enthusiasm for adopting an AI Buddy, valuing its potential for time efficiency, personalized academic support, and study organization, but expressed significant concerns about data privacy and overreliance." A "weak negative correlation emerged between AI Buddy adoption willingness and motivations for attending lectures or using library resources, while social and collaborative motivations remained unaffected."

Reports effect sizes and confidence intervals.

Research paradigm

positivist/empiricist - quantitative survey methodology with descriptive and inferential statistics

Author conclusions

"This study provides practical recommendations including the need for robust privacy protections and critical engagement strategies to ensure AI Buddies enhance, rather than undermine, the academic and communal value of higher education." More specifically, "Overall, these results suggest that students' willingness to adopt an AI Buddy is modestly related to a reduced inclination toward certain traditional academic activities, particularly lecture attendance and library use. However, it appears largely unrelated to motivations that reflect the social, collaborative, or digitally supported dimensions of university life. This implies that while the AI Buddy may serve as a partial substitute for formal academic structures, it does not replace, nor diminish, the value students place on interactive and communal aspects of the on-campus experience."

Risk of bias

Self-report bias (social desirability bias in reporting digital competence); Selection bias (3.3% of total university population; higher proportion from Arts/Social Sciences and Science faculties); Response time exclusion bias (responses completed in less than 5 minutes excluded); Institutional context bias (single university, limiting generalizability); Temporal bias (snapshot at one time point in April 2024; current AI capabilities); Social desirability bias in self-reported competence measures; Self-assessment bias regarding digital and AI competence (potential overestimation documented in literature); Selection bias: 926 out of 1310 responses retained after exclusion of incomplete/rapid responses; Faculty representation bias: higher proportion from Arts/Social Sciences and Science; lower from Medicine, Law, Vetsuisse; Single-institution sampling limiting generalizability; Correlational design preventing causal inference; Selection bias: 3.3% response rate from total student population; sample over-represented Faculty of Arts and Social Sciences (38.0% vs 34.2% institutional); under-represented Medicine (10.3% vs 14.6%), Law (10.5% vs 13.8%), and Vetsuisse (1.5% vs 2.7%); Self-report bias: All data self-reported, risk of social desirability bias and inaccurate self-assessment of competence; Attrition: 384 responses excluded as incomplete or completed in less than 5 minutes (reduction from 1310 to 926); Temporal bias: Cross-sectional design (single time point April 2024); responses shaped by current AI capabilities; Measurement bias: Hypothetical scenario (AI Buddy does not yet exist); students rating intentions not actual behavior; Gender bias in interpretation: Limited sample sizes for non-binary (n=20) and undisclosed (n=13) categories noted by authors

Limitations

"The reliance on self-reported data introduces risks of bias, such as social desirability or inaccurate self-assessment, particularly regarding students perceived digital competence." Additionally, "the correlational design further restricts causal inferences, and the modest effect sizes observed suggest that findings should be interpreted with caution, as they may not fully reflect the practical significance of AI Buddy adoption." Further, "the study's scope was limited to a single institutional context, which may constrain the generalizability of results across diverse academic or cultural settings." The authors also note that "the student's answers were clearly shaped by the capabilities of today's AI" and "our study design posed many questions in isolation and not in terms of trade-offs."

Open questions raised

Longitudinal studies examining long-term impacts of AI Buddy integration on academic performance, study behaviours, and campus engagement
Qualitative approaches (in-depth interviews, focus groups) to explore students' lived experiences with AI support
Cross-institutional and cross-cultural comparative studies to enhance generalizability
Investigation of sustained AI literacy training effectiveness across diverse student populations
Theory-based interventions grounded in Self-Determination Theory to foster intrinsic motivation in AI-supported learning environments
Limited empirical insight into students' experiences, competences, and readiness to adopt personalized AI companions

Data: "The survey instrument and the corresponding dataset generated and analysed during this study are available from the corresponding author upon reasonable request."; "The survey instrument and the corresponding dataset generated and analysed during this study are available from the corresponding author upon reasonable request." (Contact: martinez.moreno.judit@gmail.com)Code: Not mentionedExtracted from: pdfAgreement 53%

Explore related topics

Responsible AI Use Human-AI Collaboration Survey