On this page

Technology Adoption Models — the family of theories that explain and predict why individuals and institutions accept, adopt and keep using new technologies, of which the Technology Acceptance Model (TAM) is the most widely applied. In AI in education they are the standard lens for asking why learners, teachers and organizations adopt generative AI tools, and they are best treated together because GenAI-adoption studies draw on several members of the family rather than one.

Questions to Consider

  • Think of a time you adopted a new app, tool, or AI service — and a time you abandoned one. What actually drove each decision: how useful it seemed, how easy it was, or what people around you were doing? Which factor do you suspect matters most, and can a survey really capture that?
  • A common belief is that 'if a tool is clearly useful, people will use it.' Have you seen situations where a genuinely useful technology still failed to catch on, or a clearly limited one spread anyway? What might explain the gap between objective usefulness and actual adoption?
  • Adoption frameworks like TAM were designed in the 1980s for fairly simple systems. If you've used generative AI, in what ways does it differ from a word processor or a learning management system — and why might a model built around 'ease of use' and 'usefulness' struggle to capture how people relate to something that talks back?
  • Researchers often say perceived risk and trust matter less for AI adoption than expected. Before reading further, what do you predict: do students adopt AI because they trust it, despite risks, or are those concerns actually minor next to convenience and social pressure?
  • Adoption frameworks treat using a tool as a one-time decision, but effective use may be an ongoing judgment. Where in your own or your students' practice does the line between 'choosing to use AI' and 'continually deciding how to use it well' seem to blur — and what would change if we measured that instead of mere uptake?
  • Some researchers cluster learners into different 'adoption personas' instead of assuming one model fits everyone. What differences do you see among your own learners, colleagues, or students that a single average model of adoption might hide — and how could those differences shape how you support them?

Introduction

Meta-analytic evidence on AI adoption. A meta-analysis of tertiary students' AI adoption (233 correlations, 32 studies, N = 16,977) finds moderate positive correlations for individual (r = 0.57), contextual (r = 0.53), and technological (r = 0.50) factors, with usage intentions the strongest predictor (r = 0.64) and perceived risks/trust weaker than expected. Its central critique is that the field over-relies on traditional TAM/UTAUT frameworks that predate modern intelligent systems and neglect AI-specific factors such as anthropomorphism and ethics — arguing these gaps matter for advancing theory and evidence-based policy.

  • Ease of use stopped predicting once the sample had already adopted. Wei, Zhang and Heo (2026) extend TAM on 526 design students with prior AIGC use and find perceived ease of use was the only technological perception not significantly associated with self-efficacy, while knowledge, trust, reliability, usefulness, intelligence and anxiety each were (Wei et al., 2026). Usage intention was strongly associated with reported dependence in cross-sectional data.

The model family

Technology Acceptance Model (TAM)

Proposed by Davis (1989), TAM explains adoption through two core beliefs — Perceived Usefulness (PU) and Perceived Ease of Use (PEOU) — which shape users' Attitude (ATT), then Behavioral Intention (BI), then actual usage. Grounded in information-systems theory (adapted from the Theory of Reasoned Action), TAM has become the dominant framework for GenAI adoption in education. In educational GenAI research it is frequently extended with AI literacy, trust, social influence, self-determination, and critical use to capture adoption complexity beyond simple uptake.

UTAUT / UTAUT2 / UTAUT3

The Unified Theory of Acceptance and Use of Technology consolidates TAM with eight prior models into four core determinants — performance expectancy, effort expectancy, social influence, and facilitating conditions (with UTAUT2 adding hedonic motivation, price value, and habit; UTAUT3 adding personal innovativeness). UTAUT is applied across teacher and student populations: Austrian secondary math teachers (UTAUT with 448 teachers), pre-service science teachers in Ghana (UTAUT + TPB), and students in Lesotho (UTAUT3 + ARCS, using PLS-SEM and fsQCA). Jácome-Vásconez et al. extend UTAUT2 with explainable AI (Random Forest, SHAP, NCA, IPMA, K-Means) for 522 university students, finding habit the strongest predictor of ChatGPT intention and showing that effort expectancy is a necessary condition rather than a linear driver — a result only visible when XAI complements the regression.

In an already-adopted sample, habit carries the model and the expectancies go quiet. Pinedo et al. (2026) tested UTAUT2 on 310 marketing students at two Latin American universities: habit predicted intention at β = 0.426 and use at β = 0.412, and intention predicted use at 0.324. Performance expectancy (0.037), effort expectancy (0.054) and social influence (0.048) were not significant, in a model explaining R² = 0.665 of intention and 0.707 of use. The authors read the null paths as variance restriction in a population that had already adopted the tool; the outcome is self-reported use, not logged behavior.

Anxiety, not social influence, carried the affective path in a low-resource setting. A UTAUT extension with 350 Ghanaian humanities undergraduates found performance expectancy the strongest predictor of intention (β = 0.377) and effort expectancy second (β = 0.281), social influence null (β = −0.038), and digital learning anxiety at β = −0.281 (Davor et al., 2026). Effort expectancy also reduced anxiety (β = −0.126), so ease of use worked partly by lowering it.

Theory of Planned Behavior (TPB)

TPB explains intention through attitude, subjective norms, and perceived behavioral control. It is frequently paired with TAM/UTAUT in AI-adoption studies — e.g. Rizun et al. integrate TAM, TPB, UTAUT, and the FATE (Fairness, Accountability, Transparency, Ethics) framework to model the behavioral and ethical drivers of student ChatGPT adoption.

Adoption research usually stops at intention; a TPB extension that instead measured delay despite a genuine intention found perceived behavioral control the strongest negative predictor of GenAI-assisted learning procrastination (β = −0.331) across 1,243 undergraduates, with the intention–delay link significant only at moderate-to-high learning GenAI anxiety (Li et al. (2026)).

Diffusion of Innovation (DOI)

Rogers' DOI theory explains adoption as a social process in which innovations diffuse through populations over time, emphasizing innovation attributes (relative advantage, compatibility, complexity, trialability, observability) and adopter categories. It appears in institutional-level analyses, e.g. Al-Rahmi et al. combine the Technology–Organization–Environment (TOE) framework with DOI to model organizational AI adoption in Saudi universities.

At the student level, Gocyła and Strzelecki (2026) found relative advantage, compatibility, simplicity and observability predicting intention to use GenAI, explaining 48.7% of the variance, while trialability did not (f² = 0.001): free access may make trial a baseline expectation rather than a differentiator.

Technology–Organization–Environment (TOE)

TOE frames adoption as shaped by technological, organizational, and environmental contexts — a complement to individual-level TAM/UTAUT for studying institutional adoption (see Al-Rahmi (2026)).

Applications

Adoption models are applied to model student and teacher uptake of AI tools:

  • Critical-use extension: Nguyen et al. extended TAM with critical use for engineering/CS students, finding that attitudes and critical use directly predict intention — and that critical use safeguards against over-reliance.
  • Unified socio-cognitive model: Asag & Al Mamun integrated TAM with UTAUT for engineering students in Bangladesh (explaining 64% of usage variance).
  • Person-centered profiles: Rather than variable-centered models, Saihi & Ahmed cluster GenAI-adoption personas, and Chen et al. use latent profile analysis to identify four ChatGPT-acceptance profiles among pre-service teachers — showing the field's move beyond single-model, linear accounts.
  • Cross-cultural validation: Martínez-Moreno et al. cross-culturally validate adoption-related motivation constructs across Switzerland and China.
  • A knowledge mediator beside the expectancy mediators: İşbulan et al. (2026) place AI literacy between AI self-efficacy and acceptance in 622 undergraduates in Türkiye. The indirect path was significant (B = 0.15, 95% CI [0.12, 0.19]) and left a direct path intact (B = 0.43, R² = 0.29); acceptance also showed a weak negative association with AI anxiety (B = -0.34, R² = 0.05). Where the overtrust work below routes confidence through performance expectancy, this study routes it through what students know about the technology.
  • Psychological correlates: Wu et al. build on TAM to relate motivation, self-efficacy, anxiety, and risk perception to acceptance of AI-assisted English learning.
  • Regulatory competence critique: Kim argues that adoption-centered TAM models treat use as a stable decision, whereas effective AI use is an ongoing process of judgment, revision, and selective uptake — reframing AI literacy as regulatory competence and critical thinking rather than acceptance.

Adding a TPACK-family capability alongside belief frameworks such as the Theory of Planned Behavior as parallel predictors invites construct overlap and leaves the causal ordering of capability, attitude and control undecided; the Capability–Decision Model instead places AI-TPACK capability upstream of attitude and perceived behavioral control (Mnguni (2026)).

  • ML determinants of ChatGPT adoption: An exploratory ML approach examined how students' perceptions and demographics relate to intended academic ChatGPT use, using SHAP analysis to identify key learning-related constructs — prioritizing educational meaning over maximizing algorithmic performance (Verma et al., 2026).

  • Explainability and domain relevance as acceptance levers for teachers: Adapting trust-in-automation theory to teacher acceptance of AI recommendations, Feldman-Maggor et al. (2025) found that understandability (raised by explainable AI) correlated positively with both trust and acceptance of an AI grouping tool, and that domain-driven explanations in curricular language outperformed data-driven feature-importance ones on all three. Acceptance was additionally driven by pedagogical alignment and workload-reduction potential — situational factors beyond trust that standard TAM/UTAUT constructs rarely capture, reinforcing the case for extending adoption models with context and explainability.

  • Acceptance measured at the level of instructional materials (2026). Wang, Chuang and Wu (2026) operationalized Performance Expectancy, Effort Expectancy, Perceived Playfulness, and Behavioral Intention for two age-tiered AI literacy guidebooks used by 794 K-12 students and 37 teachers, validating the four-factor structure with split-sample EFA and CFA and supporting measurement invariance across the 9-12 and 13-18 editions. Younger learners reported higher levels on all four constructs, and playfulness was the strongest correlate of intention in both cohorts — with the standardized playfulness-to-intention coefficient exceeding 1.00 in the younger group and an HTMT of .950 between the two constructs. The authors read that as construct overlap and possible suppression rather than a giant effect, a useful reminder that acceptance instruments applied to materials can produce empirically entangled factors that complicate structural interpretation.

In a two-wave interview study of 22 faculty, learning more about AI made some more cautious rather than more accepting, reversing the model's assumed direction from experience to ease of use (Stalboerger et al., 2026).

Even a clean interface win need not move intention: when adults aged 18 to 25 looked for university program information through a grounded chatbot or the program's website. Every perceptual measure favored the chatbot while behavioral intention to use it again did not differ, and technology readiness moderated the size of the advantage, with the least comfortable users gaining least (Goebert et al., 2026).

Limits and extensions

TAM's cognitive focus also under-captures emotional and relational dimensions of AI use. A qualitative study of nursing students and faculty found that, alongside the four classic TAM constructs, participants -- mainly students -- described AI as a source of psychosocial support: emotional comfort during stress and a confidential space for reflection. This extends perceived usefulness beyond performance gains to well-being in high-stress professional training, suggesting adoption models should attend to affective and psychosocial factors, not only instrumental utility.

Autonomy perception and risk aversion as tool-specific extensions. A mixed-methods survey of 287 GenAI-experienced teachers across 27 countries and regions (Dai et al. (2026)) extends TAM in a direction the model family has barely touched: the perceived autonomy of the tool itself. Adding perceived artificial autonomy (how independently teachers believe GenAI can carry out an assigned instructional task) and risk aversion to the classic constructs, their refined structural model explained 79.8% of the variance in behavioral intention (χ²(97) = 202.597, CFI =.963, TLI =.954, RMSEA =.062, SRMR =.041), with perceived usefulness again dominant (PU → BI β =.832, p <.001) and ease of use feeding usefulness (PEU → PU β =.479). The instructive results are the detours. Artificial autonomy did not act directly on intention (β =.026, ns) but acted through usefulness (AA → PU β =.188, p <.001; indirect AA → PU → BI =.156, 95% CI [.082,.240]) — teachers who saw GenAI as more capable found it more useful, not more immediately adoptable. Risk aversion lowered intention (RA → BI β = −.163, p <.001) without lowering perceived usefulness (RA → PU ns), because teachers' concerns centered on students' use — cheating and integrity, weakened foundational knowledge and higher-order thinking, hallucinations, reduced human interaction, and ethical, legal and equity risks — rather than on their own. Ease of use was non-significant overall (PEU → BI β =.068) but significant among the 225 teachers who had actually used AI in teaching (β =.138, p <.05), the familiar pattern that effort registers only after real encounter; attitude had to be dropped when its correlation with intention (r =.929) breached discriminant validity. Teachers placed GenAI at semi-autonomous levels (130 at Level 2, "teacher assistance"; 94 at Level 3, "partial automation"; one at Level 6) and insisted adoption decisions be made case by case. TAM still predicts intention well, but for teacher-facing AI the constructs that move the needle are tool-specific: how autonomous the tool seems, and whose use the risk attaches to. The authors' "frenemy" label — valued as a support tool, distrusted as an autonomous agent — is the teacher-side counterpart to the student-facing extensions catalogued above.

Risk perception as a dimension-specific extension. A survey of 814 Chinese university students (Du, Ning, Shi & Chen (2026)) folds Cognitive Appraisal Theory and Protection Motivation Theory into TAM/UTAUT2 to ask not whether students adopt generative AI but what they gain from it. The model family's assumption that risk uniformly suppresses adoption does not survive: risk perception splits into information, security, technical, ethical, and legal dimensions that move benefits in opposite directions. Security risk — a threat students believe they can manage through privacy practices — was positively associated with perceived academic assistance and skill development, consistent with problem-focused coping, whereas information risk, which learners cannot easily verify, eroded psychological and emotional support, daily-life, and leisure benefits through avoidance. Threshold analyses further showed effects that change sign beyond dimension-specific cut-points, and usage experience mattered independently: students with more than a year of GenAI use reported higher benefits across four domains. The practical implication for acceptance research is that "perceived risk" is too coarse a construct to model — controllability, not the mere presence of risk, is what determines whether students engage or withdraw.

A recent extension inverts the usual direction, placing subject-specific motivational beliefs upstream of the TAM perceptions rather than beside them. Xue, Xu, Qi & Lu (2026) report utility value predicting perceived usefulness at β = 0.43 and expectancy for success predicting ease of use at β = 0.39, while core TAM paths attenuate but stay significant.

Usefulness can belong to the task set rather than the tool. In four focus groups with 26 STEM undergraduates, the same discussion board was worth their time when the prompt asked for application or showed how classmates reasoned, and busywork when it asked for restatement (Reding, Gartner and Nugent, 2026). The authors state the conclusion as a TAM claim: perceived usefulness attaches to how the prompt is built rather than to the presence of the board.

Task–technology fit as a parallel path, and necessity as a separate question. Most extensions above add constructs inside the belief model. One sibling framework that TAM studies rarely carry is Task–Technology Fit, which asks whether a tool's capabilities match the demands of the specific task. Chu and Liu (2026) joined the two in 405 design graduate students at Chinese universities, and fit acted as a second route into attitude rather than a determinant of usefulness.

Task characteristics fed fit at β = 0.322 and technology characteristics at β = 0.266. Fit then fed attitude at β = 0.224, beside perceived usefulness (0.376) and perceived ease of use (0.185), and attitude fed intention at β = 0.388. Self-efficacy carried the largest upstream coefficients, 0.458 on usefulness and 0.394 on ease of use. Perceived enjoyment raised ease of use (0.220) and left usefulness untouched (−0.033, p = 0.483) — an instrumental/affective split the classic model does not draw.

The methodological addition is the more transferable part. A multilayer-perceptron network reproduced the PLS-SEM importance ranking in all four models. Necessary condition analysis then found only two floors: self-efficacy for usefulness (d = 0.125, p = 0.015) and attitude for intention (d = 0.132, p < 0.001). Attitude at 22.1%, 34.3% and 46.4% of its scale was required for intention at 80%, 90% and 100%, so attitude behaves as a gate rather than a weight, which is a different claim from a path coefficient of 0.388. The caution sits in the outcome: R² = 0.150 and Q² = 0.011 for intention say the ten hypothesized paths leave most of the variance elsewhere.

Design features upstream of the beliefs, and importance that is not necessity. Zhang, Xin and Qing (2026) put six system features in front of the two TAM beliefs for 301 Chinese university students, borrowing a stimulus–organism–response frame. Personalization and interactivity moved perceived usefulness alone, while intelligence and content completeness moved ease of use alone. Responsiveness moved both; anthropomorphism moved neither. The model explained R² = 0.244 of intention (SRMR = 0.033), with ease of use ahead of usefulness on intention (β = 0.340 against 0.265). Ten-fold cross-validated network importance agreed, ranking ease of use at 100% against 84.70% for usefulness.

Necessary condition analysis then separated influence from requirement. Ease of use, the top-ranked predictor, failed outright (d = 0, p = 0.592). Usefulness was necessary but small (d = 0.027, p = 0.011). The authors read the split as a caution against treating a path coefficient as a floor. A strong coefficient says a unit of ease buys intention on average; only necessity says whether the other belief can be absent.

Design features routed through trust, with DOI alongside TAM. Wang et al. (2026) surveyed 653 learners about AI avatars and found the strongest path in their combined TAM–DOI model ran from perceived transparency to trust (β = 0.542). Relative advantage (β = 0.267) and perceived usefulness (β = 0.247) carried intention more than attitude did (β = 0.126).

Risk's link to use also depends on the system: across the 49 TALIS 2024 systems that administered the items, perceived AI utility was positively associated with AI use in all 49, while perceived risk was negatively associated in 40 and positively in nine (Fang & Jin (2026)).

Performance expectancy can carry a relationship that acceptance research reads as intention to use. In Kim, Kim and Zhang (2026) performance expectancy predicted AI overtrust at B = 0.6107 and was the first mediator through which self-efficacy, intrinsic motivation and extrinsic motivation reached overtrust. The expectancy-only indirect path was positive for all three predictors, 0.1355 (95% CI [0.0542, 0.2330]) for self-efficacy, so expectancy was not merely predicting whether students would use the tool.

Common Misconceptions

  • "TAM's constructs explain adoption." Asag and Al Mamun integrated TAM with UTAUT for engineering students in Bangladesh and explained 64% of the variance. Substantial and incomplete at once: roughly a third of what drives adoption sits outside the classic acceptance constructs.
  • "Acceptance models need no domain modification." Adding perceived artificial autonomy — how independently teachers believe GenAI can carry out an assigned instructional task — and risk aversion to the classic constructs changed the model materially. The original constructs were built for tools without agency, and GenAI has perceived agency.
  • "A meta-analytic adoption estimate describes any population." The meta-analysis of tertiary students is the population-level estimate; the engineering result explaining 64% is a specific population. Importing one onto the other assumes a stability the evidence does not establish.

Connected Concepts

Connected Articles

Embed this page

Copy the code below to embed this page in a learning management system or other website. The embedded version shows only the page content — it hides the site header, navigation, and footer.