How Motivation Relates to Generative AI Use: A Large-Scale Survey of Mexican High School Students

This study examined how high school students with different motivational profiles use generative AI tools in math and writing. Through K-means clustering analysis of survey data from 6,793 Mexican high school students, we identified three distinct motivational profiles based on self-concept and perceived subject value. Results revealed distinct domain-specific AI usage patterns across students with different motivational profiles. Our findings challenge one-size-fits-all AI integration approaches and advocate for motivationally-informed educational interventions.

💡 Research Summary

This paper investigates how high‑school students’ motivational profiles shape their use of generative artificial intelligence (AI) tools in mathematics and writing. Drawing on situated expectancy‑value theory, the authors operationalized motivation through two constructs: self‑concept (students’ beliefs about their competence) and perceived subject value (students’ beliefs about the importance and enjoyment of a subject). A large‑scale anonymous online survey was administered in collaboration with the Ministry of Education of Chihuahua, Mexico, reaching 7,739 students in their final two years of public high school. After excluding respondents who failed attention checks or reported no prior experience with generative AI, the analytical sample comprised 6,793 participants (61.5 % female).

Motivation was measured separately for math and writing using five items for self‑concept and three items for subject value, each on a 1‑to‑7 Likert scale. AI usage was captured with a 1‑to‑5 frequency scale for six distinct activities per domain (e.g., formula retrieval, step‑by‑step guides, answer copying in math; brainstorming, outline generation, language improvement in writing). All instruments were originally in Spanish and later translated into English for analysis.

The authors applied K‑means clustering to the two‑dimensional motivational space (self‑concept × subject value) for each domain. Three statistical criteria—elbow method, silhouette analysis, and the gap statistic—converged on a three‑cluster solution, which the researchers labeled:

1. Aspirational (high subject value, low self‑concept) – the largest group, comprising 2,809 students for math and 3,191 for writing.
2. Confident (high subject value, high self‑concept) – 2,207 math students and 1,915 writing students.
3. Disengaged (low subject value, low self‑concept) – 1,777 math students and 1,687 writing students.

AI usage patterns differed markedly across these profiles.

In Mathematics:

• Aspirational students used AI most frequently for step‑by‑step problem guides (M = 2.81) and for interpreting or re‑phrasing complex problems (M = 2.59), indicating a compensatory tutoring role.
• Confident students showed lower reliance on basic tasks such as formula retrieval (M = 2.50) and step‑by‑step guides (M = 2.60) but higher usage of AI to generate practice problems (M = 2.48), reflecting a partnership model where AI extends learning opportunities.
• Disengaged students matched the Aspirational group in formula retrieval and guide usage but reported the highest frequency of directly copying answers (M = 2.69), a behavior significantly above both other groups (p < .001). This pattern raises concerns about academic dishonesty and surface‑level learning.

In Writing:

• Aspirational learners again adopted a compensatory stance, favoring AI for brainstorming ideas (M = 2.83) and producing outlines (M = 2.72).
• Confident learners treated AI as a refinement partner, reporting the highest frequencies for language/grammar improvement (M = 2.79) and seeking AI‑generated feedback (M = 2.67).
• Disengaged learners exhibited the lowest overall AI usage across most writing tasks, suggesting that low motivation translates into limited technology engagement.

The authors argue that these domain‑specific, motivation‑driven usage patterns challenge “one‑size‑fits‑all” AI integration strategies. They advocate for interventions that are tailored to students’ motivational profiles:

• Aspirational students could benefit from AI‑driven scaffolding that strengthens conceptual understanding while gradually building self‑efficacy.
• Confident students might be offered AI tools that promote higher‑order tasks such as generating novel problems, peer‑review simulations, or advanced feedback loops.
• Disengaged students require motivational support (goal‑setting, relevance framing) alongside monitoring mechanisms to prevent misuse of AI for cheating.

The paper acknowledges several limitations. Reliance on self‑reported usage may introduce social desirability bias, and the cross‑sectional design precludes causal inference. Moreover, the sample is geographically and culturally specific to Mexican public high schools, limiting generalizability. Nonetheless, the study’s strengths include its large, representative sample, rigorous clustering validation, and nuanced examination of both math and writing contexts.

Future research directions proposed include: (1) linking AI interaction logs with academic performance to assess learning outcomes; (2) longitudinal designs to track how motivational profiles and AI usage evolve over time; and (3) experimental interventions that test tailored AI support against generic implementations.

In sum, the study provides compelling evidence that students’ motivational dispositions critically shape how they adopt generative AI tools, underscoring the need for pedagogical designs and policy frameworks that recognize and respond to these individual differences.