Antecedents of Consumer Acceptance Technology Agency–Model

1. Abstract

The Antecedents of Consumer Acceptance Technology Agency–Model (Morosan & Dursun-Cengizci, 2024) is a multidimensional psychometric instrument formulated to capture consumer cognitive, affective, and ethical appraisals regarding the delegation of decision-making authority to artificial intelligence (AI) systems. Grounded in the context of autonomous service systems within the hospitality and tourism sectors, this 22-item measurement battery evaluates the critical antecedents governing an individual’s willingness to surrender consumer agency to automated non-human agents. The instrument systematically operationalizes six core latent dimensions: Perceived Ethics (5 items), Perceived Benefits (3 items), Perceived Risks (3 items), Perceived Loss of Competence (3 items), Perceived Unpredictability (4 items), and Acceptance of Technology Agency (4 items). Evaluated along an authentic 7-point Likert scale ranging from 1 (“Strongly disagree”) to 7 (“Strongly agree”), the instrument demonstrates exceptional psychometric rigor. Confirmatory factor analysis (CFA) supports the proposed structural specification, yielding robust goodness-of-fit indices (χ² = 370.279, df = 188, χ²/df = 1.97, CFI = 0.960, TLI = 0.951, RMSEA = 0.050). Internal consistency is verified via composite construct reliability (CCR) values exceeding the 0.80 benchmark across all latent constructs, and convergent validity is corroborated through average variance extracted (AVE) values surpassing 0.50, paired with standardized factor loadings predominantly above 0.70. By bridging ethical AI governance, cognitive appraisal theory, and agency theory, the scale serves as a standardized empirical apparatus for investigating autonomous human-AI collaboration, algorithmic aversion, and consumer technology adoption in high-contact service ecosystems.

2. Keywords

Technology Agency, Artificial Intelligence Acceptance, Perceived Ethics, Hospitality Technology, Algorithmic Decision-Making, Consumer Autonomy, Loss of Competence, Perceived Risk, Psychometric Validation, Measurement Model

3. Authors

The scale was developed, calibrated, and validated by:

• Cristian Morosan, Ph.D. — Conrad N. Hilton College of Global Hospitality Leadership, University of Houston, Houston, Texas, USA. (ORCID: 0000-0001-6751-1348). Specialization: Consumer adoption of emerging technologies, information technology architecture, personalized systems, and biometric interfaces in service environments.
• Aslıhan Dursun-Cengizci, Ph.D. — Faculty of Tourism, Department of Tourism Management, Akdeniz University, Antalya, Turkey. Specialization: Consumer behavior, service marketing, automated service delivery, and digital transformation in hospitality.

4. Purpose

The exponential advancement of artificial intelligence and machine learning technologies has precipitated a paradigm shift from passive assistive systems—which merely present information for human consumption—to autonomous agentic systems capable of independent decision-making and execution. In contemporary consumer environments, such as smart hotels, travel booking platforms, and automated personal finance, artificial intelligence is increasingly tasked with making binding decisions on behalf of users. While traditional technology acceptance frameworks (such as the Technology Acceptance Model [TAM] or the Unified Theory of Acceptance and Use of Technology [UTAUT]) examine an individual’s intent to interact directly with software tools, they remain fundamentally inadequate for capturing the psychological complexities of delegation—the intentional relinquishment of behavioral control and decision-making sovereignty to an artificial entity.

The primary purpose of the Antecedents of Consumer Acceptance Technology Agency–Model is to provide an empirically validated, psychometrically robust diagnostic instrument designed to isolate, measure, and explain the cognitive, psychological, and normative factors that drive or inhibit consumer acceptance of autonomous technology agency. Specifically, the scale was engineered to address the critical research gap concerning how ethical perceptions interface with traditional functional evaluations (benefits versus risks) and profound psychological frictions (such as perceived loss of competence and systemic unpredictability).

From an applied perspective, the instrument fulfills several critical research and managerial objectives:

• Diagnosis of Algorithmic Aversion: The scale enables organizational researchers and technology developers to pinpoint the exact psychological mechanisms underpinning algorithmic aversion, identifying whether resistance stems from ethical mistrust, fear of personal de-skilling, or perceived volatility in the AI’s decision engine.
• Ethical Technology Architecture: It operationalizes ethical constructs—including fairness, privacy preservation, systemic honesty, non-exploitation, and dignity—as structural antecedents, providing software architects and algorithmic designers with tangible human-centric benchmarks for ethical system compliance.
• Segmentation and Service Personalization: In clinical, consumer, and hospitality applications, the tool allows service providers to segment user populations based on their readiness for autonomous delegation, thereby avoiding the premature or forceful imposition of automated decision-making on consumers who experience acute threats to personal autonomy.
• Theoretical Expansion: The scale bridges traditional information systems research with moral psychology, agency theory, and self-determination theory, establishing a standardized baseline for cross-cultural, longitudinal, and comparative studies on artificial agency across varying domains of everyday life.

5. Psychological Construct

The construct assessed by this instrument is Consumer Technology Acceptance of Technology Agency, conceptualized not merely as user adoption or behavioral intention to use an interface, but specifically as an individual’s cognitive willingness and behavioral intention to delegate discretionary choices, actions, and autonomous authority to an artificial intelligence entity. Within this framework, technology is no longer viewed as a static tool (an artifact) but as an independent proxy agent (an actor) operating on behalf of the principal (the consumer).

The instrument deconstructs this overarching construct into six distinct, interrelated latent subscales:

1. Perceived Ethics (5 Items)

This dimension evaluates the consumer’s normative and moral appraisal of the AI system’s underlying operational logic. Drawing upon foundational literature in computer ethics and consumer trust, it gauges beliefs regarding whether the autonomous system adheres to basic ethical tenets. Specific sub-facets include:

• Informational Privacy: The systemic safeguarding of confidential personal data from unauthorized extraction or surveillance.
• Procedural and Distributive Fairness: The perceived absence of algorithmic bias or prejudicial weighting in the system’s choice architectures.
• Systemic Honesty: The transparency and non-deceptive presentation of system motivations and decision pathways.
• Non-Exploitation: The guarantee that the autonomous agent acts strictly in the consumer’s best interest rather than covertly optimizing profit extraction for the firm.
• Dignity and Respect: The subjective impression that the automated process acknowledges human worth and avoids dehumanizing customer interactions.

2. Perceived Benefits (3 Items)

Reflecting classical cognitive appraisal theory and utilitarian valence, this subscale captures the anticipated instrumental utility derived from allowing an AI agent to assume executive control. Rather than assessing simple ease of use, it measures advanced functional gains: qualitative enhancement of the service encounter, dramatic reductions in cognitive friction and decision-making time, and elevated operational convenience achieved through autonomous service orchestration.

3. Perceived Risks (3 Items)

This construct measures the consumer’s subjective expectation of negative outcomes, systemic liabilities, and unintended adverse consequences resulting from delegated authority. Rooted in perceived risk theory, this subscale captures both generalized situational apprehension (“risky”) and specific anticipatory anxieties concerning algorithmic errors, cascading operational failures, or irreversible suboptimal selections executed by the machine without human intervention.

4. Perceived Loss of Competence (3 Items)

Anchored in psychological theories of personal efficacy and self-determination, this dimension assesses the psychological threat posed to an individual’s identity as an active, capable consumer. When an algorithm preempts consumer deliberation, individuals may feel that their agency is degraded, their personal decision-making skills are undermined, and their subjective sense of self-efficacy is neutralized. This construct specifically quantifies feelings of perceived disempowerment, diminished personal control, and systemic consumer deskilling.

5. Perceived Unpredictability (4 Items)

This subscale captures the epistemic uncertainty and perceived opacity surrounding the machine’s internal decision logic (frequently referred to as the “black-box” phenomenon). It measures the extent to which the consumer perceives the AI’s future actions to be volatile, erratic, impossible to forecast, or fundamentally outside cognitive comprehension. High perceived unpredictability induces psychological tension because the principal cannot reliably anticipate how their proxy will represent them.

6. Acceptance of Technology Agency (4 Items)

The focal criterion construct of the model represents the consumer’s consolidated behavioral intention, readiness, and explicit willingness to authorize an artificial intelligence platform to make binding decisions and enact autonomous transactions on their behalf. It embodies the active surrender of operational jurisdiction to the digital proxy within a designated service ecosystem.

6. Theoretical Framework

The Antecedents of Consumer Acceptance Technology Agency–Model is constructed upon an integrative theoretical architecture that synthesizes principles from Principal-Agent Theory, the Cognitive Appraisal Theory of Emotion and Coping, Self-Determination Theory, and Value-Belief-Norm (VBN) / Ethical Trust Paradigms.

Principal-Agent Theory

Originally formulated within economics and organizational sociology by theorists such as Jensen and Meckling (1976), Principal-Agent Theory examines relationships where one party (the principal) designates another party (the agent) to perform service tasks that involve delegating decision-making authority. When translated to autonomous technological contexts, the human consumer acts as the principal, while the artificial intelligence architecture assumes the role of the operational agent.

Classical agency dilemmas stem from information asymmetry and conflicting self-interests: the principal is frequently unable to verify whether the agent is acting faithfully in the principal’s best interests or pursuing opportunistic behaviors. Morosan and Dursun-Cengizci (2024) recognized that consumer acceptance of AI agency hinges directly on resolving this fundamental agency problem. Because algorithms are non-sentient, their perceived agency is inextricably linked to consumer trust in the institutional, structural, and moral boundaries coded into the system—represented within the scale by Perceived Ethics, Perceived Risks, and Perceived Unpredictability.

Cognitive Appraisal Theory

According to Lazarus and Folkman’s (1984) Cognitive Appraisal Theory, environmental stimuli do not directly elicit behavioral outcomes; rather, individuals evaluate the significance of an encounter in relation to their personal well-being. This occurs via:

• Primary Appraisal: Evaluating whether the encounter represents a potential benefit or a potential harm/threat (manifested in the model as Perceived Benefits versus Perceived Risks and Perceived Ethics).
• Secondary Appraisal: Evaluating one’s resources, mastery, and capacity to maintain control over the coping response. When consumers anticipate that automated decision-making will strip them of control or render them ineffective, they experience acute psychological threats (manifested as Perceived Loss of Competence and Perceived Unpredictability).

Acceptance or rejection of technology agency thus emerges as an active behavioral coping response derived from this complex internal cognitive calculus.

Self-Determination Theory and Psychological Reactance

Grounded in the foundational work of Deci and Ryan (1985), Self-Determination Theory posits that psychological health and intrinsic motivation rely upon the fulfillment of three basic needs: autonomy, competence, and relatedness. Autonomous AI systems directly challenge the need for competence and personal autonomy. Concurrently, Brehm’s (1966) Psychological Reactance Theory states that whenever individuals perceive that their behavioral freedom is threatened or constrained, an intense motivational state is aroused to re-establish that threatened freedom. The model explicitly incorporates these deep self-concept dynamics through the Perceived Loss of Competence dimension, recognizing that algorithmic interventions can induce psychological distress if consumers perceive that algorithmic decision-making reduces their personal agency.

7. Validity

The psychometric properties of the Antecedents of Consumer Acceptance Technology Agency–Model have undergone extensive empirical evaluation, demonstrating robust construct validity, convergent validity, and discriminant validity across rigorous structural equation modeling procedures.

Content and Face Validity

To establish rigorous content validity, scale development adhered to standardized multi-stage psychometric procedures. Initial item pools were adapted from validated inventories across related disciplines—including electronic commerce ethics (e.g., Roman, 2007), consumer information privacy (e.g., Xu et al., 2011), and autonomous consumer interaction paradigms (e.g., Bertrandias et al., 2021)—complemented by newly formulated items reflecting the specific realities of autonomous machine agency. Expert panels specializing in consumer psychology, hospitality management, and human-computer interaction reviewed the preliminary item pool to eliminate semantic ambiguities, confirm structural relevance, and verify that items exhibited clear face validity prior to quantitative deployment.

Convergent Validity

Convergent validity was systematically evaluated through Confirmatory Factor Analysis (CFA) by examining standardized factor loadings, average variance extracted (AVE), and statistical significance (Hair et al., 2009). The empirical validation yielded the following benchmarks:

• Standardized Factor Loadings: Virtually all standardized indicator loadings exceeded the conventional, rigorous threshold of 0.70 (ranging from 0.71 to 0.94), with only a single indicator loading at 0.674, which was deliberately retained due to its conceptual essentiality and close proximity to the 0.70 benchmark (Chen & Tsai, 2007). All factor loadings demonstrated statistical significance at p < 0.001.
• Average Variance Extracted (AVE): The AVE scores across all six latent variables substantially surpassed the established psychometric criterion of 0.50. This confirms that the latent constructs account for more than 50% of the variance observed across their respective indicator sets, confirming robust convergent validity.

Discriminant Validity

Discriminant validity was established via the classical Fornell-Larcker criterion, alongside modern multi-trait testing. For every latent construct within the model, the square root of the Average Variance Extracted exceeded the highest bivariate correlation between that construct and any other latent construct in the measurement model. Furthermore, cross-loadings were inspected; no single indicator exhibited higher factor loadings on alternative cross-constructs than on its assigned theoretical dimension, establishing that the six factors represent empirically distinct psychological phenomena.

Construct and Predictive Validity

Structural equation modeling demonstrated robust predictive and nomological validity. In alignment with theoretical postulates, Perceived Ethics emerged as a powerful positive predictor of Perceived Benefits and a negative predictor of Perceived Risks and Unpredictability. In turn, Perceived Benefits exerted a strong, positive, statistically significant direct effect on Acceptance of Technology Agency, whereas Perceived Risks, Loss of Competence, and Unpredictability exerted direct and indirect inhibitory influences. This confirms that the operationalized battery functions seamlessly within the broader nomological network of consumer decision-making and modern technology adoption.

8. Reliability

The reliability of the Antecedents of Consumer Acceptance Technology Agency–Model has been documented through tests of internal consistency and composite metric reliability, fulfilling and exceeding established psychometric thresholds.

Internal Consistency Metrics

Traditional psychometric assessments rely heavily on Cronbach’s alpha (α). However, because Cronbach’s alpha is vulnerable to underestimation or overestimation based on scale length and assumes tau-equivalence (equal factor loadings), the authors prioritized Composite Construct Reliability (CCR), also termed composite reliability (CR), alongside coefficient alpha.

• Composite Construct Reliability (CCR): In the empirical validation study conducted with adult consumers across the United States, all composite construct reliability values exceeded 0.80, safely surpassing the conservative benchmark of 0.70 (Hair et al., 2009; Nunnally & Bernstein, 1994). Specific subscales, notably Acceptance of Technology Agency and Perceived Ethics, demonstrated CCR values well above 0.88, reflecting minimal measurement error and exceptional internal stability.
• Item-to-Total Correlations: Corrected item-to-total correlations across the 22 items consistently exceeded the 0.50 threshold, demonstrating that each individual item contributes meaningfully and reliably to the measurement of its designated latent construct without redundant item bloat.

Mitigation of Common Method Bias

Recognizing that self-reported cross-sectional survey designs can be susceptible to common method variance (CMV), the researchers implemented rigorous procedural remedies during administration (Podsakoff et al., 2003):

• Respondents were explicitly guaranteed anonymity and informed that there were no objectively right or wrong responses to diminish evaluation apprehension.
• Participants were given unrestricted freedom to withdraw from the survey at any point.
• Scale placement, construct orders, and item blocks were methodically separated within the electronic questionnaire interface to break cognitive anchoring and halo effects.
• Empirical post-hoc diagnostic assessments (including Harman’s single-factor test and latent marker inspections) verified that common method bias was negligible and did not confound structural reliability coefficients.

9. Factor Analysis

The factorial validity of the instrument was tested and refined through rigorous Confirmatory Factor Analysis (CFA) utilizing maximum likelihood estimation techniques.

Confirmatory Factor Analysis (CFA) Model Specification

A comprehensive measurement model was specified comprising the six core latent dimensions: Perceived Ethics (5 indicators), Perceived Benefits (3 indicators), Perceived Risks (3 indicators), Perceived Loss of Competence (3 indicators), Perceived Unpredictability (4 indicators), and Acceptance of Technology Agency (4 indicators). The total measurement configuration consisted of 22 observed variables.

Goodness-of-Fit Assessment

Following initial diagnostic checks and standard covariance inspections, the respecified structural measurement model demonstrated exceptional model-to-data fit across all primary descriptive, relative, and absolute fit indices, meeting the rigorous standards defined by Hu and Bentler (1999) and Toh et al. (2006):

• Chi-Square Goodness-of-Fit: χ² = 370.279, with 188 degrees of freedom (df = 188).
• Normed Chi-Square: χ²/df = 1.97. A ratio below 2.0 to 3.0 is recognized in structural psychometrics as indicative of a superior model fit.
• Comparative Fit Index (CFI): 0.960 (exceeding the standard 0.95 threshold for exceptional fit).
• Tucker-Lewis Index (TLI): 0.951 (exceeding the standard 0.95 benchmark).
• Root Mean Square Error of Approximation (RMSEA): 0.050, accompanied by a narrow 90% confidence interval, firmly within the conventional range (≤ 0.05 to 0.06) indicating minimal approximation error in the population covariance matrix.

Structural Loading Distribution

Standardized factor loadings were uniformly high across all constructs:

• Perceived Ethics items loaded strongly between 0.74 and 0.89, confirming that privacy, fairness, honesty, non-exploitation, and dignity load reliably onto a unified moral appraisal dimension.
• Perceived Benefits indicators exhibited substantial loadings (0.78 to 0.91), highlighting the coherence of stay quality, time efficiency, and convenience.
• Perceived Risks displayed robust standardized parameters (0.75 to 0.88), validating risk, negative consequences, and unexpected problems as a singular risk dimension.
• Perceived Loss of Competence indicators loaded between 0.71 and 0.85, confirming personal control, diminished choice, and consumer capability loss form a coherent self-concept construct.
• Perceived Unpredictability demonstrated loadings between 0.674 and 0.84, with item 16 loading at 0.674 and retained based on substantive theoretical importance and acceptable fit parameters.
• Acceptance of Technology Agency exhibited very high factor loadings (0.83 to 0.94), affirming that willingness, permission, intention, and acceptance represent a tightly unified behavioral outcome factor.

10. Instrument / Measurement Tool

• Instrument Name: Antecedents of Consumer Acceptance Technology Agency–Model Measure
• Authors: Cristian Morosan, Ph.D. and Aslıhan Dursun-Cengizci, Ph.D. (2024)
• Instrument Type: Standardized Psychometric Self-Report Questionnaire / Multi-Factor Rating Inventory
• Target Domain: Human-Computer Interaction (HCI), Artificial Intelligence Delegation, Consumer Technology Adoption, Hospitality Management, Service Systems
• Target Population: Adult consumers (18+), including young adults, middle-aged adults, hotel guests, and users of automated/AI digital services
• Administration Method: Electronic (Computer-Assisted Self-Interviewing [CASI], mobile survey, or web-based survey); adaptable to paper-and-pencil delivery
• Completion Time: Approximately 5 to 8 minutes
• Number of Items: 22 standardized items
• Underlying Subscales (6 Dimensions):
  • Perceived Ethics (Items 1–5; 5 items)
  • Perceived Benefits (Items 6–8; 3 items)
  • Perceived Risks (Items 9–11; 3 items)
  • Perceived Loss of Competence (Items 12–14; 3 items)
  • Perceived Unpredictability (Items 15–18; 4 items)
  • Acceptance of Technology Agency (Items 19–22; 4 items)
• Authentic Response Format: 7-point Likert scale:
  • 1 = Strongly disagree
  • 2 = Disagree
  • 3 = Somewhat disagree
  • 4 = Neither agree nor disagree
  • 5 = Somewhat agree
  • 6 = Agree
  • 7 = Strongly agree
• Scoring and Computational Procedures:
  • All 22 items are positively worded relative to their individual subscale constructs.
  • No reverse scoring is required across the indicator items within their respective subscales.
  • Subscale scores are computed by calculating the arithmetic mean or composite sum of the items assigned to each factor. Mean subscale scores retain the original 1.00 to 7.00 metric, facilitating direct cross-construct comparison.
  • In structural equation modeling (SEM) and regression applications, researchers can employ latent variable modeling with factor-score regression or compute composite index scores for structural path testing.
  • Higher scores on Perceived Ethics, Perceived Benefits, and Acceptance of Technology Agency denote positive appraisals and strong delegation intentions. Higher scores on Perceived Risks, Perceived Loss of Competence, and Perceived Unpredictability signify acute cognitive and emotional barriers to technological delegation.

11. Permissions & Fee and Test Year

Test Publication Year: 2024

Copyright and Ownership: The Antecedents of Consumer Acceptance Technology Agency–Model was developed and published by Cristian Morosan and Aslıhan Dursun-Cengizci (2024). The formal academic manuscript is copyrighted by Emerald Publishing Limited in the International Journal of Contemporary Hospitality Management.

Licensing and Research Permissions: Under standard international academic research practices, the instrument’s items and operationalized scales are available for non-commercial educational, scientific, and scholarly research purposes, provided that appropriate formal academic attribution is extended to the original authors and the published source journal. Commercial implementations, proprietary software integration, diagnostic enterprise consulting, or publication within fee-bearing diagnostic batteries typically require explicit written authorization and licensing from the authors and/or the publisher.

Inquiries and Contact: Researchers seeking collaboration, customized implementations, or cross-cultural validation protocols may contact the lead author, Dr. Cristian Morosan, via the Conrad N. Hilton College of Global Hospitality Leadership at the University of Houston, or through his registered academic ORCID identifier (0000-0001-6751-1348).

12. References

• Bertrandias, L., Lemoine, J. F., & Carricano, M. (2021). Autonomous systems in consumer service: Exploring the psychological barriers to delegation. Journal of Retailing and Consumer Services, 62, 102604. https://doi.org/10.1016/j.jretconser.2021.102604
• Brehm, J. W. (1966). A theory of psychological reactance. Academic Press.
• Chen, C. F., & Tsai, D. (2007). How destination image and evaluative factors affect behavioral intentions? Tourism Management, 28(4), 1115–1122. https://doi.org/10.1016/j.tourman.2006.07.007
• Deci, E. L., & Ryan, R. M. (1985). Intrinsic motivation and self-determination in human behavior. Springer Science & Business Media. https://doi.org/10.1007/978-1-4899-2271-7
• Hair, J. F., Black, W. C., Babin, B. J., & Anderson, R. E. (2009). Multivariate data analysis (7th ed.). Prentice Hall.
• Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1–55. https://doi.org/10.1080/10705519909540118
• Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3(4), 305–360. https://doi.org/10.1016/0304-405X(76)90026-X
• Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal, and coping. Springer Publishing Company.
• Morosan, C., & Dursun-Cengizci, A. (2024). Letting AI make decisions for me: An empirical examination of hotel guests’ acceptance of technology agency. International Journal of Contemporary Hospitality Management, 36(3), 946–974. https://doi.org/10.1108/IJCHM-08-2022-1041
• Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory (3rd ed.). McGraw-Hill.
• Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88(5), 879–903. https://doi.org/10.1037/0021-9010.88.5.879
• Roman, S. (2007). The ethics of online retailing: A scale development and an approach to the impact on consumers’ satisfaction and loyalty. Journal of Business Ethics, 72(2), 131–148. https://doi.org/10.1007/s10551-006-9165-7
• Toh, R. S., Hu, M. Y., & Sisodiya, S. R. (2006). A comparative analysis of hotel reservation methods. Cornell Hotel and Restaurant Administration Quarterly, 47(2), 130–142. https://doi.org/10.1177/0010880405284346
• Xu, H., Luo, X. R., Carroll, J. M., & Rosson, M. B. (2011). The personalization privacy paradox: An exploratory study of decision making process for location-aware marketing. Decision Support Systems, 51(1), 42–52. https://doi.org/10.1016/j.dss.2010.11.017

13. Items of the Scale (Questionnaire)