Attitude Toward Use (AT) in Tech Adoption

• Attitude Toward Use (AT) is an individual’s overall affective evaluation of technology, incorporating both pragmatic benefits and moral considerations.
• It is measured through multi-item Likert scales and validated via factor analysis, linking constructs like perceived usefulness and ease of use to behavioral intention.
• AT plays a central mediating role in technology acceptance models, with empirical research supporting its impact on user behavior and system adoption.

Attitude Toward Use (AT) refers to an individual’s overall affective evaluation—favorable or unfavorable—of using a technology or system in a specified context. Established as a central component in the Technology Acceptance Model (TAM) and regularly refined in subsequent theoretical frameworks, AT both mediates core belief constructs such as perceived usefulness and ease of use, and serves as a proximal determinant of behavioral intention. Across domains ranging from generative AI adoption and online assessment, to smartphone use among older adults and metaverse education, AT has been operationalized to capture both pragmatic and moral responses to technology-facilitated transformation.

1. Theoretical Foundations

In the original TAM, AT is modeled as “the user’s overall affective disposition or desire to use” a system. Building on attitudes as theorized in the Theory of Reasoned Action (TRA) and Theory of Planned Behavior (TPB), Davis et al. (1989) structured AT to mediate between beliefs (such as Perceived Usefulness, PU, and Perceived Ease of Use, PEOU) and Behavioral Intention (BI) to use a technology (Garcia et al., 2021, Acosta-Gonzaga et al., 2016, Karkonasasi et al., 2018). Subsequent models, including UTAUT and context-specific extensions, variably maintain, replace, or supplement the AT construct to account for evolving user beliefs and social or moral dimensions (such as moral objection or professional identity) (Sikorski et al., 29 Dec 2025, Wolfe et al., 16 Oct 2025, Pucer et al., 2024).

2. Operationalization and Measurement

AT is commonly operationalized using multi-item Likert-type scales. Items target affective (enjoyment, liking), evaluative (wise, good idea), and pragmatic (utility, benefit) dimensions, tailored to the context and technology in question:

Study Context	Example AT Items (paraphrased/quoted)	Scale/Reliability
GenAI at work (Sikorski et al., 29 Dec 2025)	“Will give firms a competitive advantage”; “Should use extensively”; “Will support my work”; “No moral objection”	4 items, 7-point, α = 0.804
LMS (Canvas) (Garcia et al., 2021)	“I would like coursework more if I used Canvas”; “Pleasant experience”; “Good idea to use Canvas”	3 items, 5-point, α = 0.83
AI at work (Wolfe et al., 16 Oct 2025)	“Using AI is a good idea”; “Makes work more interesting”; “I like working with AI”	3 items, 7-point, α = 0.86
Smartphone (older adults) (Pucer et al., 2024)	“Using a smartphone is beneficial”; “Good idea”; “I like the idea”	3 items, 5-point, α = 0.929
Metaverse in education (Misirlis et al., 2023)	“Using metaverse is a good idea”; “I like the thought of learning through metaverse”	2 items, 5-point, λ > 0.7
VHC in health (Karkonasasi et al., 2018)	“Using VHC is a wise idea”; “Using VHC is a good idea”	2 items, 5-point, α = 0.50

Measurement commonly involves factor analysis: PCA or CFA confirms dimensionality; standardized loadings usually exceed 0.70, except where few items are used. Cronbach’s alpha is reported for scale reliability, generally above 0.80, though shortened two-item measures may have lower reliability.

3. Structural Role in Acceptance Models

AT plays a pivotal mediating role in TAM and its extensions. Typically, AT is predicted by cognitive (PU, PEOU), affective (Enjoyment, Hedonic Motivation), social (Subjective Norms), and moral (lack of objection) variables, and in turn predicts BI or actual use.

A standard structural model takes the form: $$\begin{align*} AT &= \beta_1\,PU + \beta_2\,PEOU + \varepsilon \ BI &= \beta_3\,AT + \beta_4\,PU + \zeta \end{align*}$$ For instance, in an LMS setting, PU strongly predicts AT (β = 0.53, p < 0.001), which then predicts BI (β = 0.25, p < 0.05) (Garcia et al., 2021). In online math assessment, the path AT → BI is especially strong (β = 0.64, p < 0.001) (Acosta-Gonzaga et al., 2016). The mediating role of AT is empirically confirmed in bootstrapping analyses and path models, supporting its centrality for explaining technology adoption (Karkonasasi et al., 2018, Pucer et al., 2024).

4. Empirical Correlates and Antecedents

Empirical findings consistently document strong positive links between AT and both intention and actual use. Notably:

• In GenAI at work, positive AT is correlated with higher use frequency (r = 0.568, p < .001). AT is suppressed by anxiety/discomfort (r = –0.562), concerns about human-like characteristics (r = –0.435), and belief in human uniqueness (r = –0.225) (Sikorski et al., 29 Dec 2025).
• For smartphone adoption in older adults, Hedonic Motivation is the dominant antecedent of AT (β = 0.454, p < 0.001), followed by Ease of Use (β = 0.227). Technology anxiety reduces AT (β = –0.188), with subgroup variation: for non-users, Ease of Use is more influential, for users, anxiety dominates (Pucer et al., 2024).
• In metaverse education, Subjective Norms show a uniquely large effect (β = +2.43) while Self-Efficacy surprisingly predicts lower AT (β = –1.97) (Misirlis et al., 2023).

A recurring empirical result is that AT captures both personal affective response and the perceived social or organizational expectation to use a technology, with variation in which antecedent dominates by technological context and user cohort.

5. Contextual Sensitivity and Multidimensionality

The construct of AT exhibits strong contextual sensitivity. For generative AI in higher education, AT is not monolithic; students display support for their own use but frequently reject faculty adoption, with 37.2% in a sample rejecting both and 30.8% supporting both (Gao et al., 26 Mar 2026). Themes such as GenAI output validity and pedagogical integrity emerge as central to negative AT toward faculty use. This underscores that AT is not only a generic evaluation but can be role- and context-specific, tracking perceived appropriateness, procedural justice, and broader moral or professional concerns.

Extensions to the base construct are often needed: for GenAI at work, the AGAWA scale integrates a moral-acceptance and social-influence dimension alongside traditional affective and instrumental components (Sikorski et al., 29 Dec 2025). The omission of dimensions such as Perceived Ease of Use in some AT measures is flagged as a limitation when nontrivial usability issues may still shape affective response.

6. Psychometric Properties and Methodological Considerations

Most studies report high internal consistency for AT measures, but brevity (e.g., two-item scales) can reduce reliability (e.g., α = 0.50 (Karkonasasi et al., 2018)). Factor analyses demonstrate that AT forms a robust single dimension in most samples. Confirmatory factor indices (CFI, RMSEA, SRMR) typically meet conventional thresholds (e.g., CFI > 0.95, RMSEA < 0.08 (Sikorski et al., 29 Dec 2025)).

Some contexts introduce additional methodological challenges:

• The use of forced-choice (dichotomous) AT measurement limits aggregation and reliability assessment (Gao et al., 26 Mar 2026).
• Cultural and sample constraints (e.g., single-country student cohorts) limit generalizability of factor structure and observed links (Sikorski et al., 29 Dec 2025, Misirlis et al., 2023).
• Later-generation acceptance models (such as UTAUT and its AI-focused extensions) variably include or exclude AT, affecting direct comparability of results (Wolfe et al., 16 Oct 2025).

7. Implications, Limitations, and Future Research

The predictive utility of AT is robust across domains: it reliably accounts for intention and actual system use, especially when affect, utility, and social/moral resonance are integrated in its operationalization. However, several limitations recur: sample specificity (student or professional), restricted item pools (under-representing cognitive or conative facets), and methodological constraints (cross-sectional design, lack of behavioral validation).

Priorities for future research include:

• Broadening cultural and professional validation of AT scales (e.g., AGAWA in cross-industry settings (Sikorski et al., 29 Dec 2025)).
• Expanding AT measurement beyond simple “no moral objection” toward richer moral and identity-relevant appraisals.
• Embedding AT in full structural models and multi-group SEM frameworks to rigorously establish its mediation and moderation effects.
• Incorporating behavioral or system-log triangulation to bolster the validity of self-reported attitude measures (Wolfe et al., 16 Oct 2025).
• Integrating cybersecurity education or awareness interventions in adoption campaigns for populations sensitive to threat perceptions (e.g., older adults and smartphone adoption (Pucer et al., 2024)).

In sum, Attitude Toward Use remains an indispensable construct for understanding and modeling technology acceptance. Its theoretical scope, empirical measurement, and contextual sensitivity have continued to evolve as novel technologies and adoption contexts challenge and expand its explanatory power.