Perceived Usefulness in Tech Acceptance
- Perceived Usefulness is defined as the belief that utilizing a system enhances performance and task success.
- It is operationalized with multi-item Likert scales, revealing strong reliability and validity across domains like education and software engineering.
- PU robustly predicts behavioral intention by being influenced by ease of use, system quality, and social factors within the Technology Acceptance Model.
Perceived Usefulness (PU) is a cornerstone construct within technology acceptance and user evaluation research, particularly as operationalized in the Technology Acceptance Model (TAM) and its derivatives. PU captures the cognitive judgment that using a given system, tool, or information source will enhance performance, learning, or task success. Although its canonical definition and measurement are rooted in the TAM literature, recent research has expanded, contextualized, and empirically tested PU across a wide array of domains—from interactive information retrieval to AI-assisted programming and autonomous vehicles.
1. Theoretical Definition and Operationalization
The canonical definition of Perceived Usefulness, as introduced by Davis (1989), is "the degree to which a person believes that using a particular system would enhance his or her job performance." This conceptualization has been universally adopted and slightly adapted depending on context:
- Educational Systems: PU becomes “the extent to which a student believes that employing [system/tool] will improve learning outcomes or study efficiency” (Garcia et al., 2021, Setälä et al., 2 Jan 2025, Hasan et al., 2023).
- Information Retrieval and Digital Humanities: PU is operationalized as “the extent to which a retrieved resource or document helps the user accomplish a simulated or real-world writing or research task” (Kettunen et al., 2022, Kettunen et al., 2022).
- AI Tool Adoption in Software Engineering: PU is framed as “the belief that a system enhances performance,” specifically, that an AI-powered tool improves coding work effectiveness and efficiency (Zakharov et al., 29 Apr 2025).
- Consumer Technology: In smartphone adoption research, PU is “the degree to which an individual believes that using a smartphone will enhance their daily activities” (Pucer et al., 2024).
These domain-specific operationalizations retain the original TAM focus on performance enhancement as the evaluative core of usefulness.
2. Measurement and Psychometric Properties
Measurement of PU nearly always relies on multi-item Likert scales, with items directly adapted from TAM or contextualized for domain-specific activities.
PU Item Formulation and Scale Examples
| Domain | Sample PU Items | Scale | Cronbach's α / CR |
|---|---|---|---|
| AI in software engineering (Zakharov et al., 29 Apr 2025) | "Using the AI tool would improve my coding performance," "I would find the AI tool useful..." | 7-point Likert | α ≥ 0.90 (previous) |
| E-learning/Canvas (Garcia et al., 2021) | "Canvas would improve my ability to learn," "Canvas is useful for learning" | 5-point Likert | α = .92, CR = .915 |
| Autonomous Vehicles (Shen et al., 2024) | "Using an AV is very useful in life and work," "can improve travelling comfort" | 5-point Likert | α = 0.776, CR = 0.776 |
| Blockchains (Shrestha et al., 2019) | “This system would increase my productivity,” "Would make it easier to share the data" | 7-point Likert | α = 0.89, CR = 0.919 |
| AI in EFL (Yao et al., 23 Mar 2025) | “AI tools shortens the time I need to complete English-learning tasks... overall, raises performance” | 7-point Likert | α = 0.934, CR = 0.951 |
Psychometric results consistently demonstrate strong internal consistency (Cronbach’s α > 0.80) and satisfactory composite reliability and convergent validity (AVE > 0.5) (Pucer et al., 2024, Setälä et al., 2 Jan 2025, Jalali, 2021). Discriminant validity is routinely assessed via AVE and HTMT criteria, and multi-group invariance for cross-cultural applications is established at least at the metric level (Tarhini et al., 2015).
3. Determinants and Structural Role in Acceptance Models
PU is universally modeled as an endogenous latent variable influenced by various antecedents—most notably Perceived Ease of Use (PEOU), but also by domain-specific constructs, social factors, affective states, and system qualities.
Antecedents of PU
- Perceived Ease of Use (PEOU): Positive, direct, and often strong effect on PU, though the strength can vary by context; dominant in consumer technology and LMS adoption (Garcia et al., 2021, Pucer et al., 2024, Shen et al., 2024).
- System Quality/Output Quality: Robust predictor in platform and IS research (e.g., blockchain, search engines) (Shrestha et al., 2019, Saravanos et al., 2022).
- Perceived Enjoyment & Compatibility: Significant in educational and GenAI studies; for Finnish students, perceived enjoyment (PE) and tool compatibility are stronger predictors of PU than ease of use (Setälä et al., 2 Jan 2025).
- Social Influence and Self-Efficacy: Subjective norm and self-efficacy consistently boost PU, particularly in novel, collaborative, or metaverse contexts (Misirlis et al., 2023, Pucer et al., 2024).
- Affective and Motivational States: Optimism and innovativeness (from Technology Readiness Index), as well as extrinsic “warm-glow,” yield significant positive effects on PU (Hasan et al., 2023, Saravanos et al., 2022).
- Negative Predictors: Technology anxiety, insecurity, rejection of benefits can lower PU (Pucer et al., 2024, Hasan et al., 2023).
Context-Dependent Predictive Patterns
Certain studies report context-driven attenuation or reversal of the canonical PEOU→PU link (e.g., weak or non-significant in blockchain (Shrestha et al., 2019), metaverse education (Misirlis et al., 2023), or Finnish GenAI adoption (Setälä et al., 2 Jan 2025))—often when system novelty or enjoyment dominates utility judgments.
4. Predictive Power in Behavioral Intention and Attitude
PU is, in canonical TAM, the principal cognitive antecedent of behavioral intention (BI) to use a system, typically modeled as:
where AT represents attitude. PU generally explains substantial portions of variance in intention (e.g., β = 0.506, p < 0.001 for smartphones (Pucer et al., 2024); β ≈ 0.7 for AI in mathematics (Setälä et al., 2 Jan 2025)) and attitude toward use (Garcia et al., 2021, Tarhini et al., 2015). However, empirical potency can be context-dependent. For example, in MetaEducation, PU's effect on intention is negligible, likely due to limited student familiarity (Misirlis et al., 2023).
5. Domain-Specific Implementations and Findings
Information Retrieval and Digital Humanities
In interactive IR with historical newspaper archives, PU is operationalized as graded task-relevance. Upgrading OCR from ~68% to ~83% word recognition accuracy caused a statistically significant 7.93% increase in mean PU (Wilcoxon p = 0.002, pre-formulated queries), with less pronounced effects in exploratory, user-formulated tasks (Kettunen et al., 2022, Kettunen et al., 2022). This demonstrates that perceived utility is strongly sensitive to backend data quality, confirming that subjective user judgments track objective processing improvements.
Software Engineering and AI
In AI-powered software development, PU increases as developers assign multiple distinct “roles” to the AI (e.g., assistant, advisor, problem solver). The number of roles attributed to an AI tool exhibits a strong positive correlation with PU (r = 0.59, p < 0.001), suggesting that multi-faceted user mental models foster higher adoption potential (Zakharov et al., 29 Apr 2025).
Programming Education
Survey and hierarchical clustering with first-year programming students reveals that high system usage does not guarantee high PU; perceived trust, infrastructural support, and moderate engagement levels are crucial for the actualization of usefulness beliefs (Hernandez et al., 30 Nov 2025).
Consumer/Communications Technology
In cross-cultural TAM studies (e.g., RSS feed adoption in the UK and Lebanon), PU consistently exhibits metric invariance across cultural contexts and remains a significant, direct driver of intention (UK: β = 0.210, Lebanon: β = 0.203), despite latent differences in other TAM factors (Tarhini et al., 2015).
6. Structural Equation Modeling and Comparative Effect Strengths
Empirical TAM and extended TAM (ETAM, TAM3, etc.) research generally confirms the structural primacy of PU, though effect sizes and mediation patterns vary:
| Model / Domain | β (PU → BI/ITU) | R² (PU, explained) | Context-Note |
|---|---|---|---|
| AV adoption in Shanghai (Shen et al., 2024) | 1.677, p=0.012 | 0.776 | Exceptional effect (field deployment) |
| AI in mathematics education (Setälä et al., 2 Jan 2025) | ~0.7, p<.001 | 0.609–0.732 | Compatibility amplifies effect |
| Smartphones, older adults (Pucer et al., 2024) | 0.506, p<.001 | 0.914 | EoU is dominant predictor of PU |
| Canvas LMS adoption (Garcia et al., 2021) | 0.49, p<.001 | 0.17 (PU) | PU fully mediates EoU → AT |
| Blockchain research-sharing (Shrestha et al., 2019) | 0.50, p<.01 | 0.89 | System Quality > PEOU as PU antecedent |
| MetaEducation (Misirlis et al., 2023) | n.s. | n/a | Limited effect; unfamiliar context |
When new antecedents such as “warm-glow” are introduced, extrinsic warm-glow (e.g., status/satisfaction) positively predicts PU (β = 0.195, p < 0.01), whereas intrinsic (altruistic) warm-glow does not (Saravanos et al., 2022).
7. Practical and Theoretical Implications
PU remains the linchpin of technology acceptance across virtually all empirical contexts. Its maximization is more reliably achieved through demonstrable performance gains, robust system quality, and contextual alignment with user needs than through superficial ease-of-use enhancements. When properly measured, PU is both psychometrically robust and structurally central.
Design and deployment strategies aimed at enhancing PU include:
- Transparent demonstration of concrete performance improvements (e.g., dashboards, before/after comparisons) (Yao et al., 23 Mar 2025).
- Social- and context-driven onboarding (e.g., peer influence, community recognition) (Misirlis et al., 2023, Saravanos et al., 2022).
- Alignment of tool features with domain-specific workflows and established user habits (Setälä et al., 2 Jan 2025).
- Targeted training to address technological anxiety and bridge awareness-action gaps (Pucer et al., 2024).
Research further recommends infrastructure investment (especially in developing contexts), explicit framing of multi-role AI capabilities, and continuous backend data quality improvements for IR systems.
Conclusion
Perceived Usefulness exhibits exceptional theoretical stability, psychometric reliability, and structural centrality across domains. Its predictive effect on technology acceptance is robust but context-sensitive. Proper operationalization—grounded in domain-relevant performance criteria, aligned with user expectations, and validated through rigorous SEM—ensures that PU continues to function as an essential metric for system evaluation, adoption modeling, and innovation diffusion in both academic and applied technology research.