Transforming Curriculum Design with Generative AI: A Model for Assessing Teacher Digital Competence

Synthesis: This PLS-SEM study of 434 in-service teachers in the Dominican Republic designs and psychometrically validates an extended Technology Acceptance Model (TAM) instrument for assessing teacher digital competence in integrating generative AI tools into curriculum planning. All 12 hypothesized paths were supported, and behavioral intention emerged as the main predictor of digital competence (β = 0.479), with Self-Efficacy exerting the largest effect on perceived enjoyment (f² = 1.839). The model demonstrated strong reliability, convergent and discriminant validity, and good explanatory power (BI R² = 0.694; DC R² = 0.230), positioning behavioral intention as a strategic priority for Teacher AI Competency development.

Core Finding

The study confirms that an extended TAM framework — combining Self-Efficacy, perceived usefulness, perceived ease of use, perceived enjoyment, and attitude of use — predicts teachers' behavioral intention to use GenAI tools in curriculum design, and that this behavioral intention is the strongest direct predictor of teachers' pedagogical digital competence (H12: β = 0.479, t = 10.226, f² = 0.298). The IPMA analysis reinforces this: behavioral intention combined the highest importance (≈0.49) and performance (≈86) for predicting digital competence, identifying it as the strategic lever for training interventions. Notably, self-efficacy was the exogenous root of the model, feeding usefulness (β = 0.446), ease of use (β = 0.500), and especially enjoyment (β = 0.805, f² = 1.839 — the largest effect in the model).

The Extended TAM Model

The instrument (40 items, 7-point Likert) operationalized seven latent factors, all specified as reflective: Digital Competence (DC), Behavioral Intention (BI), Attitude of Use (AOU), Perceived Usefulness (PU), Perceived Ease of Use (PEOU), Perceived Enjoyment (PE), and Self-Efficacy (SE). Reliability was strong across all factors (Cronbach's α .952–.985; CR .961–.985), with convergent validity (AVE all > .50) and discriminant validity confirmed via both Fornell-Larcker and HTMT criteria.

All 12 hypotheses were supported:

• PU → AOU (β = 0.178) and PU → BI (β = 0.171) — perceived usefulness shapes both attitudes and intention.
• PEOU → AOU (β = 0.158) and PEOU → PU (β = 0.384) — ease of use feeds attitude and usefulness.
• PE → AOU (β = 0.530), PE → BI (β = 0.188), PE → PEOU (β = 0.323) — intrinsic enjoyment is the strongest antecedent of attitude.
• SE → PU (β = 0.446), SE → PEOU (β = 0.500), SE → PE (β = 0.805) — self-efficacy underpins the whole acceptance pathway.
• AOU → BI (β = 0.541) and BI → DC (β = 0.479) — the attitude–intention chain culminates in digital competence.

The model explained moderate-high variance in the perceptual factors (PU R² = 0.607, PEOU 0.614, PE 0.648, AOU 0.644, BI 0.694) but only weak variance in digital competence (DC R² = 0.230), signaling that unmeasured contextual factors also shape competence.

Self-Efficacy as the Root Driver

Self-efficacy is the sole exogenous variable and the most powerful effect in the model. Consistent with Bandura's Social Learning Theory, teachers confident in their ability to solve technical problems, learn autonomously, and innovate are substantially more likely to perceive GenAI tools as useful, easy, and enjoyable. The authors argue that a sense of effectiveness builds "technoptimism," reducing resistance to new technologies and accelerating implementation — a key consideration for Teacher AI Competency and Educational Development programs.

Behavioral Intention as a Competence Lever

A novel contribution is the explicit demonstration that behavioral intention predicts digital competence — a relationship not previously established in the traditional TAM. The authors interpret this through practice: as teachers declare intent to learn and use AI tools, they actively seek methodological knowledge, experiment (even through failures), and build skills through use, leading to real competence gains. Conversely, replicating familiar methods and standard edtech limits competence development in a rapidly evolving technological landscape.

Context: The Digital Divide

The study is grounded in the Dominican Republic's pronounced dual digital divide: in rural areas 69.7% of low-income households lack home internet access versus 34.6% in urban contexts, with even high-income rural households (42.4%) lagging metropolitan areas (10.6%). These structural constraints mean individual-level determinants coexist with — and are constrained by — infrastructure and connectivity gaps, shaping the real possibilities of GenAI integration.

Relevance to the Knowledge Base

This paper is a significant contribution to the Teacher AI Competency concept: it provides a validated, psychometric instrument and an extended-TAM model specifically for GenAI in curriculum planning, an area with limited empirical evidence in Latin American contexts. It connects individual acceptance psychology (Self-Efficacy, Motivation) to Learning Design practice and Educational Development/professional development policy. Its core actionable message — that behavioral intention, fueled by self-efficacy and intrinsic enjoyment, is the primary lever for building teacher digital competence — offers a testable model for AI in Education training design.

What this means for practice

• Instructors. Begin GenAI training with hands-on work in real curriculum-planning tools rather than theoretical explanation: the authors conclude that success depends on the practical (praxeological) dimension of GenAI use, easy access to simple edtech tools, and overcoming motivational barriers.
• Faculty developers. Build Self-Efficacy first, since it is the model's only exogenous root and its strongest effect (self-efficacy → perceived enjoyment β = 0.805, f² = 1.839): give teachers small, solvable technical problems so confidence, enjoyment, and ease of use rise together.
• Faculty developers. Treat behavioral intention as a training outcome in its own right, not a byproduct — it was the strongest direct predictor of digital competence (β = 0.479, t = 10.226) and combined the highest importance and performance in the IPMA.
• Administrators. Fund connectivity and device access alongside teacher training: in rural Dominican households 69.7% of low-income and 42.4% of high-income households lacked home internet, against 34.6% and 10.6% in urban areas.
• Researchers. Reuse the validated 40-item, 7-point instrument (Cronbach's α .952–.985) to test whether these acceptance pathways hold across teacher subgroups and other national contexts.

Limitations

• The 434 teachers were recruited by non-probability convenience sampling, and while the sample spanned educational levels and urban and rural contexts, subgroup differences were not formally examined, so the structural paths describe the overall sample rather than any subpopulation.
• All seven factors, digital competence included, were measured with self-report Likert items rather than observed practice or performance tasks.
• Digital competence had R² = 0.230, so the model accounts for only 23% of its variance and leaves most of the construct to unmeasured determinants.
• Contextual structural factors — connectivity, school infrastructure investment, software licensing, and national policy — were outside the model, and the authors present the findings as specific to the Dominican Republic.

Citation

Guillén-Gámez, F. D., Tomczyk, Ł., Habibi, A., & Díaz Vargas, B. L. (2026). Transforming curriculum design with generative AI: a model for assessing teacher digital competence. Educational Technology Research and Development.