Exploratory K-means clustering identifies four heterogeneous student AI-user profiles differing in experience, competence, usefulness, trust, and control, but the solution's optimality and stability are not established¶

Claim · All claims
Evidence · 1 study · associational r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? A four-cluster K-means solution on n = 961 students showed heterogeneity: cluster sizes were 221, 111, 378, and 251, ranging from low-experience/low-trust to high-experience/high-trust profiles. → Braun 2026

Evidence¶

Braun 2026¶

Braun, Y. S., & Khafizov, S. M. (2026). The AI Adaptation Gap in Higher Education: Students, Faculty, and Administrative Staff: AI Use, Trust, and Institutional Policy at a University Specializing in Teacher Education. https://arxiv.org/abs/2608.25063 doi:10.48550/arxiv.2608.25063

q2 · i? · associational · r2

Exploratory K-means clustering among students only (n = 961, nine features, fixed seed, ten initializations) identified four clusters, with Cluster 3 (n = 251) showing the highest experience, competence, usefulness, and trust and Cluster 1 (n = 111) the lowest.

"The primary student-only K-means specification identified four clusters in the complete-case subsample of n = 961. With the fixed random seed and ten initializations described above, cluster sizes were 221, 111, 378, and 251."

Discussion¶

Related Claims¶

• Teachers' post-workshop platform engagement fell into three behavioral profiles, with most teachers showing minimal creation and browsing — related
• Latent profile analysis of pre-service teachers' CT skills identifies three profiles (Novice, Developing, Proficient), revealing heterogeneous, non-linear skill acquisition — related
• Three distinct learner profiles emerge from function art activities: Repetitivists, Simplists, and Multifunctionists — related
• Hierarchical clustering of CDM attainment probabilities identifies five learner groups with distinct patterns of professional standards attainment — related
• K-means clustering of K–8 students' platform trace data yields seven distinct engagement profiles — related
• Two distinct group deliberation patterns emerge in response to regulation triggers: the Plan and Implementation Approach (PIA) and the Trials and Failure Approach (TFA) — related
• Clustering yields a small number of behaviorally distinct team clusters that instructors can use for cluster-differentiated feedback — related
• Gaussian Mixture Modeling identifies two latent clusters of delayed start behavior, outperforming three-cluster and unimodal models — related
• Cluster analysis of instructors identifies three user groups that differ in several AI literacy dimensions, with the ethics scale providing a clear distinction — related
• Learners' generative-AI use profiles are heterogeneous, with extractive delegation more common than potentially collaborative use — a narrower finding that bears on this claim
• Cluster analysis identifies three stable learner profiles: Engaged and Literate, Uncritical Reliant, and Skeptical Minimal users — related
• K-Means++ clustering of self-reported behavioral proxies yields six distinct, stable learner profiles (k = 6, Silhouette = 0.62) — related
• Cluster analysis of self-reported GenAI use identifies five profiles, including a strategic high-usage profile low on code generation but high on debugging and explanations — a narrower finding that bears on this claim
• A four-cluster solution is optimal for segmenting higher education users' GenAI chatbot adoption perceptions — related
• Four distinct GenAI chatbot adoption personas emerge: Cautious Achievers, Skeptical Utilitarians, Disengaged Doubters, and Engaged Enthusiasts — related
• Cluster analysis identifies four tentative moral response profiles: Comfortable Adopters, Guilty Non-Users, Cautious Users, and Morally Distressed Avoiders — related
• Four distinct LLM learner profiles emerge from latent class analysis of learning contexts, tasks, and devices — related
• Exploratory heat maps reveal three student groups, including disengaged-but-successful students more prevalent in the Python courses, showing low content completion can link to both positive and negative outcomes — related
• Latent Class Analysis identifies six provisional manager typologies with heterogeneous readiness and barrier profiles — related
• A two-dimensional profiling method based on initial artifact quality and criterion-aligned revision categorizes student groups into four analytic profiles in an AI-supported RQ/hypothesis revision activity — related
• Cluster analysis identifies two antagonistic student profiles differing in level of technology-supported self-regulated learning — related
• A cluster of highly active learners driving the learning communities included equally many low-hierarchy as high-hierarchy members — related
• SRL writing-stage clusters differed in production and revision but not in pausing — related
• Three learning profiles emerge among university students: organized-deep, unorganized-deep, and dissonant — related
• Cluster analysis across both studies identifies three meta-attention knowledge profiles, including an overconfident group with a self-knowledge/strategy mismatch — related
• Affinity propagation clustering of gameplay attempts yielded an interpretable six-cluster solution (part.6)
• Latent class analysis of 641 higher education students' survey responses identifies four distinct classes of feedback experience — related
• Secondary students in AI learning fall into three distinct SDT-based motivational profiles — Disengaged, Developing, and Self-Determined — at both pre- and post-test — related
• K-means clustering of standardized prior knowledge and final grades yields four performance profiles: Consistent Achievers, Growing Achievers, Declining Performers, and Struggling Performers — related
• Post-training GenAI use patterns were unevenly distributed: efficient collaborators were the largest profile (47.5%), followed by over-revisers (26.2%) and mixed users (21.2%) — related
• Latent profile analysis of six AI-attitude indicators yields four profiles: Positive Empowerment, Anxious Acceptance, Low-Perception Detached, and High-Perception High-Vigilance — related
• K-means clustering of stage-level CACI observations yields four student-AI interaction patterns: Intensive Engagers, Fast Scanners, Traditionalists, and Frequent Switchers — related
• Students with moderate initial CT levels show substantially greater self-efficacy gains than high-CT and low-CT peers, an 'Optimal Development Zone' effect — related
• Four distinct action-control belief profiles (agentic, moderate, avoidant, and mixed) emerge among Finnish ninth graders — related