Extended TAM2: Comprehensive Tech Acceptance

• Extended TAM2 is a comprehensive framework that integrates social influence, cognitive processes, and system attributes to predict user acceptance of new technologies.
• It employs multi-item scales and structural equation modeling to quantify relationships among constructs such as perceived usefulness, ease of use, and contextual moderators.
• Empirical studies show its robust predictive power in settings like education, mobile commerce, and AI-mediated learning, informing practical strategies for technology adoption.

The Extended Technology Acceptance Model (TAM2) is a comprehensive framework for predicting and explaining users’ acceptance and intended use of new technologies, especially in domains where sociocultural context, individual attitudes, and system attributes interact in complex ways. Developed as a substantial augmentation of the original Technology Acceptance Model (TAM), TAM2 incorporates additional determinants from social influence theory, cognitive instrumental processes, and—depending on context—further extensions such as cultural, demographic, and innovation-diffusion factors. Recent empirical studies have tested TAM2 across domains including educational technology, mobile commerce, and AI-mediated learning environments.

1. Definitions and Constructs in TAM2

TAM2 advances the original TAM by incorporating a broader set of latent constructs. Core constructs include:

• Perceived Usefulness (PU): The extent to which a user believes that a system will enhance task performance. In educational studies, PU has been operationalized as enhancement of investigation effectiveness and data accuracy for student-led science projects (Ga et al., 6 Nov 2025), shopping efficiency for mobile commerce (Nabot et al., 2021), and learning performance for AI integration in mathematics education (Setälä et al., 2 Jan 2025).
• Perceived Ease of Use (PEOU): The belief that system use is free of effort. In educational applications, PEOU is closely aligned with the cognitive load imposed by interfaces, such as block-coding tools for Arduino in Korean classrooms (Ga et al., 6 Nov 2025).
• Job Relevance (JOB): Originally described as applicability to one’s job, this has been adapted to future educational or career alignment in school contexts (Ga et al., 6 Nov 2025).
• Output Quality (OQ): The degree to which the system performs its intended tasks accurately (e.g., sensor data reliability in Arduino projects).
• Result Demonstrability (RD): The tangibility and communicability of technology-driven results, such as sharing time-series experimental data with peers (Ga et al., 6 Nov 2025).
• Subjective Norm (SN): Perceived social pressure from important others to use (or not use) a technology. SN is context-sensitive—for example, linked to prestige and university admissions in Korean education systems (Ga et al., 6 Nov 2025), or moderated by cultural dimensions such as Power Distance (Tarhini, 2016).
• Image (IMG): The perceived enhancement of one’s social status from technology adoption.

Recent extended models also integrate:

• Compatibility (CMP/COM): The perceived fit of a technology with a user’s values, experiences, and existing workflow, a construct drawn from Innovation Diffusion Theory and found strongly relevant in mobile commerce and AI adoption (Nabot et al., 2021, Setälä et al., 2 Jan 2025).
• User Experience (UX): Composite HCI metrics including efficiency, effectiveness, and subjective satisfaction; prominent in mobile commerce modeling (Nabot et al., 2021).
• Uncertainty Avoidance (UA): Derived from Hofstede’s cultural dimensions, representing risk aversion and privacy-security concerns in technology use (Nabot et al., 2021, Tarhini, 2016).
• Perceived Enjoyment (PE): The affective or hedonic dimension of technology use, important in settings such as generative AI adoption for mathematics education (Setälä et al., 2 Jan 2025).

2. Structural Relationships and Theoretical Pathways

TAM2 elaborates the original causal pathways of TAM with multiple antecedents and mediators. The structural equations typically take the form:

$$\mathrm{BI} = \beta_1\,\mathrm{PU} + \beta_2\,\mathrm{PEOU} + \beta_3\,\mathrm{SN} + \beta_4\,\mathrm{IMG} + \beta_5\,\mathrm{JOB} + \beta_6\,\mathrm{OQ} + \beta_7\,\mathrm{RD}$$

$$\mathrm{PU} = \gamma_1\,\mathrm{PEOU} + \gamma_2\,\mathrm{SN} + \gamma_3\,\mathrm{IMG} + \gamma_4\,\mathrm{JOB} + \gamma_5\,\mathrm{OQ} + \gamma_6\,\mathrm{RD}$$

Extensions often add additional predictors (UX, UA, CMP/COM) and moderators (cultural and demographic dimensions). Structural Equation Modeling (SEM), including both PLS-SEM and covariance-based approaches, is the standard analytic technique for estimating direct, indirect, and moderated effects. In mobile commerce, for instance, Behavioral Intention was most powerfully predicted by UX ($\beta=0.46$), Ease of Use ($\beta=0.35$), Uncertainty Avoidance ($\beta=0.33$), and Usefulness ($\beta=0.20$) (Nabot et al., 2021).

3. Methodological Operationalization

Measurement of TAM2 constructs typically relies on multi-item scales with demonstrated psychometric validity (Cronbach's $\alpha>0.7$; AVE, composite reliability), often using 5-point Likert ratings. Qualitative adaptations use structured coding of interview statements and observational data to assign utterances or actions to constructs (PU, PEOU, SN, IMG, etc.) (Ga et al., 6 Nov 2025). Triangulation via independent coders and multiple data sources enhances credibility.

Indicator examples by domain:

Construct	Sample Item (paraphrased from studies)	Domain
PU	“Using MC on my smartphone would improve my shopping efficiency.”	Mobile Commerce (Nabot et al., 2021)
PEOU	“It was not difficult for me to use it… I just follow what I’ve learned.”	Arduino in Education (Ga et al., 6 Nov 2025)
CMP/COM	“MC fits well with the way I like to shop.”	Mobile Commerce (Nabot et al., 2021)
PE	“The activity of using the system is enjoyable in its own right.”	GenAI in Math Ed. (Setälä et al., 2 Jan 2025)

4. Domain-Specific Adaptations and Extensions

TAM2 has been rigorously adapted to fit contextual specificities of various technology domains:

• Educational Technology: Constructs like Job Relevance and Image are operative as alignment with academic/career aspirations and perceived social prestige (Ga et al., 6 Nov 2025). The presence of visual block-coding interfaces substantively elevated PEOU by reducing cognitive load, contrary to earlier literature that emphasized programming barriers.
• Mobile Commerce: User Experience and Uncertainty Avoidance are prominent, with UX emerging as the strongest predictor of intention, and UA significantly shaping both PU and CI in high-risk environments (Nabot et al., 2021).
• Generative AI in Education: Compatibility with existing digital workflows significantly improves the explanatory power of the model, particularly in accounting for Perceived Usefulness. Ease of Use and Enjoyment, while positively associated with PU, exhibit negligible direct effects on intention in exam-oriented educational settings (Setälä et al., 2 Jan 2025).
• Cross-Cultural E-Learning Acceptance: Individual-level measures of Hofstede’s cultural dimensions (PD, UA, MF, IC) partly moderate core TAM2 path strengths. Demographics (age, gender, experience, education level) also moderate both direct and indirect pathways to Behavioral Intention and Actual Usage (Tarhini, 2016).

5. Empirical Findings, Moderation, and Implications

Recent studies yield convergent results on TAM2’s empirical utility:

• Explanatory Power: Extended TAM2 models frequently explain a large proportion of the variance in Behavioral Intention ($R^2$.68–.80) (Nabot et al., 2021, Setälä et al., 2 Jan 2025, Tarhini, 2016). The inclusion of compatibility raised $R^2(\text{PU})$ from 0.609 to 0.732 in generative AI adoption (Setälä et al., 2 Jan 2025).
• Direct Effects: PU remains the dominant driver of BI, with notable augmentation from UX and UA in mobile commerce. In educational contexts, multifactorial influences—social reputation (SN, IMG), technical facilitation (PEOU), career relevance (JOB)—jointly shape acceptance.
• Moderators: Cultural and demographic factors demonstrate significant, though context-specific, moderation. For example, Power Distance strengthens the influence of SN on BI, and gender moderates PU→BI and QWL→BI path strengths (Tarhini, 2016).
• Practical Recommendations: Technology adoption interventions should address multiple TAM2 dimensions—optimizing interface usability, aligning with users’ goals and workflows, supporting social prestige needs, and mitigating risk aversion via clear privacy/security assurances. Curriculum design for educational integration should explicitly link technology to diverse career pathways and provide pedagogical scaffolding.

6. Methodological Innovations and Comparative Context

Methodological advancement within TAM2 research includes:

• Qualitative Approaches: Through semi-structured interviews and real-time observational coding, the underlying rationales behind construct formation and interplay are made explicit, extending TAM2’s explanatory reach into practice-based settings (Ga et al., 6 Nov 2025).
• Construct Expansion: The introduction of novel predictors—hedonic motivation (PE), compatibility (COM), UX metrics—responds to the limitations of classic TAM2 in specialized contexts.
• Comparative Assessment: Empirical effect sizes for canonical pathways (e.g., PEOU→BI, PEOU→PU) are consistently smaller in some domains (e.g., Finnish secondary mathematics), while compatibility and enjoyment play elevated roles. Ease of use may be less critical where baseline digital competence is high and assessment outcomes predominate (Setälä et al., 2 Jan 2025).

7. Limitations and Directions for Future Research

Empirical generalizability of TAM2 extensions is conditional on methodological design (quantitative survey vs. qualitative coding), sample context, and technological affordances. Future work may further refine the operationalization of constructs like compatibility and image, systematize multi-level moderation (demographic, cultural, organizational), and explore cross-domain adoption dynamics via comparative modeling. Integration of TAM2 with adjacent models (e.g., UTAUT2, Innovation Diffusion Theory) is suggested for domains with multifaceted social and psychological determinants.

A plausible implication is that researchers and practitioners should not rely on generic TAM2 coefficients but rather calibrate construct weights and pathway specifications to the specific sociotechnical environment and adoption target under paper, using both theoretical and empirical insights for maximum explanatory power.