Uncovering adoption personas for generative AI in higher education: a clustering-based segmentation approach

Synthesis: Saihi and Ahmed (2026) apply a person-centered, clustering-based approach to uncover distinct adoption personas for GenAI chatbots in higher education, moving beyond the aggregate, average-effect assumptions of traditional TAM/UTAUT models. Using hierarchical clustering followed by k-means on 192 validated observations (107 students, 85 educators), they identify a four-persona solution — Cautious Achievers, Skeptical Utilitarians, Disengaged Doubters, and Engaged Enthusiasts — that reflects diverse configurations of trust, usefulness, efficiency, ethical comfort, and satisfaction. The study argues that understanding this latent user heterogeneity is essential for designing inclusive, context-aware AI interventions rather than one-size-fits-all deployments.

Core Finding

GenAI chatbot users in higher education are not homogeneous; they cluster into four distinct adoption personas — Cautious Achievers, Skeptical Utilitarians, Disengaged Doubters, and Engaged Enthusiasts — that traditional variable-centered models obscure. While TAM and UTAUT estimate average structural relationships across an entire population (assuming user homogeneity), the person-centered analysis reveals that perceptual, experiential, contextual, and demographic indicators combine differently across user subgroups. A four-cluster solution was selected as optimal using internal validation indices (Silhouette, Calinski-Harabasz, Davies-Bouldin). These personas matter because attitudes like Trust, perceived usefulness, ethical comfort, and satisfaction configure differently across roles and contexts — for instance, an educator concerned with Academic Integrity may assess chatbots very differently than a student focused on task efficiency.

From Variable-Centered to Person-Centered Adoption

The paper's central methodological contribution is shifting from variable-centered technology-adoption modeling to a person-centered lens. Traditional frameworks like the Technology Acceptance Model (TAM) and UTAUT identify average predictors of adoption intention but cannot discover naturally occurring user profiles. The study repurposes a validated survey dataset (previously analyzed with structural equation modeling) and applies unsupervised machine learning — hierarchical clustering followed by k-means — to perceptual and experiential indicators (ease of navigation, trust in AI, data-Privacy concerns, perceived efficiency, user satisfaction, perceived learning performance). This complements rather than replaces the earlier SEM model, offering a richer depiction of user diversity within the same theoretical framework.

The Four Personas and What Drives Them

The identified personas reflect distinct configurations of adoption-related perceptions:

Engaged Enthusiasts — high trust, usefulness, and satisfaction; active, exploratory engagement with Conversational AI.
Cautious Achievers — reasonably positive but guarded; achievement-oriented with moderated trust.
Skeptical Utilitarians — value functional utility/efficiency but hold skeptical or ethical reservations.
Disengaged Doubters — low trust, low satisfaction, limited engagement.

The personas are profiled using demographic and contextual characteristics (role, tech-savviness, study field, gender, age, education level). Role distinctions matter: students tend to prioritize immediate Feedback and user-friendly interfaces, while educators are more concerned with content accuracy, instructional alignment, and academic integrity. Technological proficiency and disciplinary culture (e.g., humanities vs. STEM) further shape perceptions and engagement.

Implications for Design and Deployment

The study offers actionable guidance for tailoring chatbot onboarding, training, and support to specific user needs rather than assuming uniform adoption. By equipping institutions with a data-driven segmentation tool, it supports more inclusive, human-centered AI Education integration strategies. The authors emphasize that without understanding user diversity, even advanced AI technologies may fail to deliver value or exacerbate existing inequalities in digital readiness and inclusion — a point that ties the work to AI Literacy, Student Experience, and Ethics in institutional AI deployment.

Relevance to the Wiki

This article provides an empirical, person-centered complement to the wiki's coverage of AI adoption and acceptance in Higher Ed. It directly engages the Technology Acceptance Model concept and connects it to Trust, Conversational AI, and Student AI Interaction. It also speaks to Faculty Development and Teacher Role by highlighting how educators and students differ in their concerns and support needs, and to Governance by informing differentiated deployment strategies.

Connected Concepts

Connected Articles

Citation

Saihi, A., & Ahmed, V. (2026). Uncovering adoption personas for generative AI in higher education: a clustering-based segmentation approach. International Journal of Educational Technology in Higher Education.

View Source View Changes Edit