AGAWA Scale: A Psychometric Tool for GenAI Attitudes

• AGAWA scale is a four-item instrument that measures perceptions of GenAI as both a productive and morally acceptable work collaborator.
• It integrates core constructs from TAM and UTAUT, assessing productivity, performance expectancy, social influence, and ethical comfort on a 7-point Likert scale.
• Empirical validation confirms its robust reliability and validity, making it a valuable tool for diagnosing and evaluating GenAI acceptance in various settings.

The AGAWA scale (Attitudes to Generative AI as Work Associate) is a succinct, theory-driven, four-item psychometric instrument designed to quantify the extent to which individuals—specifically knowledge workers or university students—perceive generative artificial intelligence (GenAI) as a productive and morally acceptable collaborator. Developed to address the rapid integration of GenAI into professional environments, the scale synthesizes established models of technology acceptance with explicit attention to the affective and ethical dimensions that shape real-world adoption. The AGAWA scale aids in identifying psychological and moral barriers to GenAI integration and supports rigorous, efficient measurement for research, diagnostics, and intervention evaluation (Sikorski et al., 29 Dec 2025).

1. Theoretical Basis and Construct Framework

The AGAWA scale is grounded in the Technology Acceptance Model (TAM) and the Unified Theory of Acceptance and Use of Technology (UTAUT), both widely used for predicting and explaining technology uptake. Each AGAWA item targets core constructs from these frameworks:

• Perceived Usefulness (PU, TAM): Operationalized in items emphasizing productivity and performance gains attributable to GenAI.
• Performance Expectancy (PE, UTAUT): Implicit in statements about job performance improvements.
• Social Influence (SI, UTAUT): Embedded via normative language, suggesting competitive and organizational imperatives.
• Facilitating Conditions (FC, UTAUT): Addressed in allusions to infrastructural support and legality.
• Effort Expectancy (EE, UTAUT): Incorporated in workplace support statements.

Importantly, AGAWA advances beyond existing models by introducing an explicit moral/trust dimension: one item directly assesses whether respondents experience ethical or affective discomfort regarding GenAI as a colleague. This dual focus captures both pragmatic (productivity, performance) and affective/moral (emotional comfort, ethical acceptability) facets of GenAI acceptance (Sikorski et al., 29 Dec 2025).

2. Item Content, Response Format, and Administration

The finalized AGAWA scale comprises four items (S1, S2, S4, S7) selected via empirical analysis. Items are presented to respondents in either fixed or randomized order. Each item is rated on a seven-point Likert scale, with anchors:

• 1 = "Strongly disagree"
• 7 = "Strongly agree"

All items are positively keyed; higher values denote greater acceptance of GenAI as a work collaborator.

Item Code	Item Content	Primary Construct(s)
S1	Companies that utilize generative artificial intelligence will gain a competitive advantage over those that do not in the near future.	PU, PE, SI
S2	Companies should use generative artificial intelligence as extensively as possible—of course, only within legal bounds—to achieve maximum efficiency.	PU, PE, SI, FC
S4	Generative artificial intelligence will prove to be a support in my professional work.	PU, PE, EE, FC
S7	Using generative artificial intelligence as an assistance at professional work does not raise my moral objection.	Moral/affective trust

Administration requires approximately one minute and is suitable for large-scale, longitudinal, or diagnostic applications (Sikorski et al., 29 Dec 2025).

3. Empirical Scale Structure and Dimensionality

The scale’s dimensionality was empirically established through principal component analysis (PCA) on an initial item pool (S1–S8). Only S1, S2, S4, and S7 co-loaded strongly (λ₁ = 0.770, λ₂ = 0.796, λ₃ = 0.776, λ₄ = 0.769) on the primary component. Other candidates failed to meet the inclusion threshold (loadings < 0.63) and were excluded. Confirmatory factor analysis (CFA) of the four-item solution demonstrated good fit metrics: χ²(2) = 38.1, p = 0.004; Comparative Fit Index (CFI) = 0.959; Root Mean Square Error of Approximation (RMSEA) = 0.0614; Standardized Root Mean Square Residual (SRMR) = 0.0543.

AGAWA should be interpreted as a single-factor measure, though the items can be conceptually partitioned into pragmatic (“competitive advantage,” “maximum efficiency,” “support at work”) and moral (“moral objection”) dimensions (Sikorski et al., 29 Dec 2025).

4. Psychometric Properties and Correlate Evidence

AGAWA demonstrates robust internal consistency (Cronbach’s α = 0.804). Its convergent and discriminant validity was evaluated relative to three barrier constructs, each assessed using adapted subscales:

• NATIR-AI-PL: Fear of interaction with GenAI (α = 0.83)
• NARHT-AI-PL: Fear of human-like features in GenAI (α = 0.73)
• BHNU-AI-PL: Belief in human nature’s uniqueness/superiority over GenAI (α = 0.83)

Inter-factor correlations among the barrier constructs were high:

• r(NATIR, NARHT) = +0.661, p < .001
• r(NATIR, BHNU) = +0.495, p < .001
• r(NARHT, BHNU) = +0.500, p < .001

Correlations between AGAWA and the barrier scales were strongly negative:

• r(AGAWA, NATIR) = –0.562, p < .001
• r(AGAWA, NARHT) = –0.435, p < .001
• r(AGAWA, BHNU) = –0.225, p < .001

AGAWA scores were strongly and positively associated with self-reported GenAI usage frequency (r = +0.568, p < .001), underscoring criterion validity. Content validity was established by expert panel review (relevance ratings: 78%, 78%, 100%, 89%, threshold = 70%) (Sikorski et al., 29 Dec 2025).

5. Scoring Procedures and Interpretation

Each respondent’s AGAWA summary score is the arithmetic mean of the four items:

$AGAWA = \frac{1}{4}\sum_{i=1}^4 x_i$

where $x_i$ denotes the response to item $i$ (S1, S2, S4, S7).

Interpretive thresholds:

• AGAWA < 4.0: Negative or reserved attitude
• AGAWA ≈ 4.0: Neutral attitude
• AGAWA > 4.0: Positive attitude toward GenAI as a collaborator

Dimensional subscores (optional):

• Pragmatic dimension: $(x_1 + x_2 + x_3)/3$
• Moral dimension: $x_4$

For example, responses [6, 5, 7, 4] yield AGAWA = 5.50 (positive), pragmatic = 6.0, moral = 4 (residual moral reservation) (Sikorski et al., 29 Dec 2025).

6. Applications, Utility, and Future Directions

AGAWA’s brevity permits rapid deployment in diverse research and organizational settings. Applications include:

• Embedding within employee‐experience or change-management assessments to identify groups with low GenAI readiness
• Correlating with objective behavioral data (e.g., GenAI log-ins) for predictive validation
• Pre-/post-intervention evaluation (e.g., training, ethics briefings) to track attitudinal change
• Translation and cultural validation for cross-linguistic or comparative studies, employing standard forward/backward procedures
• Modular extension with perceived ease-of-use (TAM) or fine-grained trust metrics (e.g., cognitive vs. affective trust) for enhanced diagnostic granularity

A plausible implication is that AGAWA can serve both as a standalone monitor of GenAI attitudes and as a diagnostic tool to flag populations at risk for resistance due to pragmatic or moral misgivings (Sikorski et al., 29 Dec 2025).

7. Companion Scales and Broader Context

AGAWA’s correlational validation rests on three companion barrier subscales (NATIR-AI-PL, NARHT-AI-PL, BHNU-AI-PL), enabling detailed analysis of resistance sources:

Scale	Construct	α
NATIR-AI-PL	Fear of interaction	0.83
NARHT-AI-PL	Fear of human-like AI	0.73
BHNU-AI-PL	Human uniqueness/superiority	0.83

These relationships confirm the entwined affective and moral components underlying GenAI adoption. Research using AGAWA can be situated within the broader literature on technology acceptance, human-automation trust, and workplace AI ethics, while offering rapid, theoretically coherent, empirically validated measurement for emerging workplace challenges (Sikorski et al., 29 Dec 2025).