Competence with the AI Platform

1. Abstract

The Competence with the AI Platform scale is a specialized psychometric instrument developed to evaluate an individual’s perceived capability, domain knowledge, and subjective efficacy regarding the safeguarding of personal data privacy in interactions with artificial intelligence (AI) ecosystems. Formally validated by Ertugrul Uysal, Sascha Alavi, and Valéry Bezençon (2022) in their foundational investigation published in the Journal of the Academy of Marketing Science, the scale addresses the urgent psychosocial friction arising from the pervasive domestic deployment of conversational AI assistants, such as Amazon Alexa, Apple Siri, and Google Assistant. Comprising three parsimonious, self-report items administered on a 7-point Likert scale ranging from 1 (“strongly disagree”) to 7 (“strongly agree”), the instrument quantifies the psychological resource of perceived competence within human–computer interaction (HCI) and artificial intelligence marketing contexts.

Psychometrically, the instrument demonstrates robust unidimensionality, high internal consistency reliability (α = .92 to .94 across validation samples), and strong composite reliability (CR > .90). Confirmatory factor analysis confirms excellent construct validity, marked by substantial standardized factor loadings (λ > .85) and an average variance extracted (AVE) exceeding .80, well above standard psychometric benchmarks. By assessing the interplay between perceived algorithmic intrusion and individual coping agency, the scale functions as an essential metric for researchers in consumer psychology, privacy behavior, and technology governance, distinguishing itself as a vital tool for capturing user agency amidst the pervasive ubiquity of ambient computing systems.

2. Keywords

Competence with the AI Platform, artificial intelligence, privacy competence, self-efficacy, consumer data protection, human-AI interaction, conversational agents, perceived control, psychometrics, privacy paradox, algorithmic surveillance

3. Authors

The scale was developed and operationalized by a multidisciplinary team of scholars specializing in marketing, consumer behavior, and human-technology interactions:

• Ertugrul Uysal — Institute of Marketing and Communication Management, Faculty of Communication, Culture and Society, Università della Svizzera italiana (USI), Lugano, Switzerland. Research focus: Consumer-technology relationships, conversational artificial intelligence, and digital privacy dynamics.
• Sascha Alavi — Institute of Marketing, Faculty of Management and Economics, Ruhr-University Bochum, Bochum, Germany. Research focus: Sales management, digital transformation, customer relationship dynamics, and psychological reactions to emerging retail technologies.
• Valéry Bezençon — Enterprise Institute, Faculty of Economics and Business, University of Neuchâtel, Neuchâtel, Switzerland. Research focus: Ethical consumer behavior, technological sustainability, and decision-making mechanisms under information asymmetry.

4. Purpose

The primary purpose of the Competence with the AI Platform scale is to provide a reliable, theoretically grounded measurement of an individual’s perceived self-efficacy and procedural knowledge concerning data privacy protection within ambient, voice-enabled, or automated artificial intelligence environments. With the rapid expansion of voice user interfaces (VUIs) and ambient computing, devices such as smart speakers continuously listen for wake words and process domestic, intimate acoustic environments. This continuous sensory integration generates a profound psychological paradox: while consumers derive high functional utility and relational comfort from humanlike AI interactions, they simultaneously encounter heightened vulnerabilities regarding algorithmic data surveillance, secondary data harvesting, and potential breaches of informational privacy.

In academic research, this scale allows investigators to decouple general technological competence from domain-specific data privacy self-efficacy. Traditional measures of computer self-efficacy often focus broadly on operational skill (e.g., configuring software, browsing networks), failing to capture the subtle, cognitive, and regulatory agency required to govern personal data flows through proprietary AI algorithms. The Competence with the AI Platform scale explicitly captures whether users feel empowered with the concrete knowledge, operational capability, and emotional confidence needed to safeguard their privacy against specific ambient devices (such as Amazon Alexa).

Clinically and practically, the scale has substantial application across consumer welfare advocacy, human-centered UI/UX design, and regulatory compliance evaluations. As regulatory bodies worldwide enforce mandates such as the European Union’s General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA), technology firms face rigorous demands to establish verifiable transparency and consumer autonomy. Applying this instrument enables organizations to assess whether interface alterations (such as physical mute switches, transparent voice-log management consoles, or proactive privacy reminders) meaningfully augment user competence or whether consumers remain functionally disempowered despite ostensibly transparent disclosures.

5. Psychological Construct

The psychological construct captured by this instrument is perceived privacy competence within an artificial intelligence ecosystem. Conceptually, this construct is rooted in the broader psychological framework of self-efficacy, initially formalized by Albert Bandura, which reflects an individual’s cognitive appraisal of their capacity to execute behaviors necessary to produce specific performance attainments. When transposed onto the digital landscape and algorithmic mediation, perceived competence encompasses three interlinked psychological dimensions:

1. Cognitive Competence (Declarative and Procedural Knowledge): This dimension addresses the user’s internal mental model of how the AI platform collects, stores, processes, and disseminates personal data. It measures subjective knowledge regarding the functional architecture of the platform, including settings, opt-out protocols, audio retention parameters, and third-party data-sharing permissions. Users with high cognitive competence believe they understand what happens to their data and the specific mechanisms required to manage that data pipeline.
2. Operational/Behavioral Capability: This dimension assesses the individual’s subjective appraisal of their ability to execute protective actions in practice. Even if an individual theoretically knows that a privacy dashboard exists, they may feel cognitively overwhelmed, fatigued by complex menus, or behaviorally incapable of executing defensive steps. High operational competence reflects the user’s perception that they can practically implement privacy safeguards without prohibitive friction.
3. Affective and Evaluative Confidence: Bandura emphasized that self-efficacy is fundamentally characterized by an emotional and evaluative certainty that one’s actions will successfully yield the anticipated outcome. In the context of AI data management, this involves feeling resilient against algorithmic ambiguity and confident that one’s defensive measures effectively prevent unauthorized exposure or surveillance.

Importantly, the construct operates as a psychological moderator in human–AI relationship trajectories. In the findings of Uysal et al. (2022), perceived competence dictates whether humanlike features in an AI assistant (such as anthropomorphic voice inflections, conversational turn-taking, and affective responsiveness) are perceived as a benevolent, useful helper or as an intrusive “Trojan Horse.” When users possess low competence with the AI platform, the assistant’s humanlike social cues trigger severe perceptions of manipulative intrusion, threatening psychological freedom and inducing protective resistance. Conversely, when users report high competence, they perceive themselves as possessing the requisite agency to mitigate vulnerability, permitting healthy relationship formation, relational trust, and uninhibited functional utilization.

6. Theoretical Framework

The Competence with the AI Platform scale is built upon several foundational psychological theories, integrating concepts from cognitive psychology, social-cognitive theory, and human-technology relational dynamics:

Social Cognitive Theory and Perceived Self-Efficacy

Bandura’s (1986, 1997) Social Cognitive Theory posits that human agency operates within a model of triadic reciprocal causation, wherein cognitive factors, environmental events, and behavior interact continuously. Within this paradigm, self-efficacy constitutes the primary cognitive mechanism governing motivation, thought patterns, emotional states, and behavior. In ambient computing environments, the platform represents a complex, semi-opaque socio-technical environment. The user’s perceived competence determines their coping effort: individuals who perceive themselves as efficacious in data protection engage in proactive privacy configurations, whereas those who feel inefficacious surrender to technological fatalism, experiencing learned helplessness regarding data harvesting.

The Protection Motivation Theory (PMT)

Originally formulated by R. W. Rogers (1975), Protection Motivation Theory explains how individuals appraise threats and coping options in high-risk contexts. PMT bifurcates cognitive processing into threat appraisal (perceived severity and vulnerability to an external hazard) and coping appraisal (response efficacy and self-efficacy). The Competence with the AI Platform scale captures the coping self-efficacy dimension of PMT. In AI contexts, privacy risks (e.g., constant recording of private domestic discourse) function as the threat, while competence measures the consumer’s subjective appraisal that they possess the internal resources to mitigate that threat through self-directed action.

Psychological Reactance Theory

Jack Brehm’s (1966) Psychological Reactance Theory suggests that when an individual’s behavioral freedoms are threatened or eliminated, a motivational state of reactance emerges, compelling the person to re-establish the compromised freedom. Uysal, Alavi, and Bezençon (2022) integrated reactance theory to demonstrate that AI anthropomorphism can be perceived as an invasive encroachment on personal autonomy if users lack personal competence. Competence with the AI platform acts as an essential psychological buffer: high competence preserves perceived autonomy, minimizing the cognitive perception that an ambient smart device is compromising personal boundary control.

7. Validity

The psychometric validity of the Competence with the AI Platform scale was rigorously established through the empirical procedures reported by Uysal, Alavi, and Bezençon (2022) across extensive field and experimental studies involving verified smart speaker users:

Construct and Convergent Validity

Construct validity assesses whether the operationalized scale genuinely captures the underlying conceptual construct. In confirmatory factor models, all three items demonstrated standardized factor loadings exceeding the conventional .70 threshold, with empirical loadings ranging from λ = .88 to .93. These strong loadings yielded an Average Variance Extracted (AVE) well above the conservative .50 benchmark, typically exceeding .80. This demonstrates that the shared variance between the latent competence construct and its indicators is substantially greater than the associated measurement error variance, verifying solid convergent validity.

Discriminant Validity

Discriminant validity was verified using both the classical Fornell-Larcker criterion and the modern Heterotrait-Monotrait (HTMT) ratio of correlations. In the focal research, the square root of the AVE for the competence scale was notably higher than its highest inter-construct correlations with related focal constructs, including:

• Perceived Privacy Risk (discriminating functional capability from externalized threat appraisal);
• Anthropomorphism (verifying that subjective competence is distinct from perceptions of the agent’s humanlike characteristics);
• Perceived Intrusion / Trojan Horse Beliefs (differentiating internal coping ability from external intrusive experiences);
• Relational Trust (ensuring that perceived competence is not merely an alternative manifestation of overall trust in the manufacturer).

Furthermore, all HTMT values involving the competence scale remained well below the conservative .85 threshold, establishing that the scale represents an empirically distinct cognitive dimension.

Nomological and Predictive Validity

Predictive and nomological validity were demonstrated by testing structural equation models and moderated mediation paths. The scale successfully predicted key relational and behavioral outcomes: higher scores on Competence with the AI Platform systematically attenuated consumer feelings of reactance and perceived invasion of privacy when exposed to highly humanlike AI conversational cues. Moreover, the scale displayed significant predictive capacity on sustained usage intensity, proactive privacy management actions, and reduced psychological friction during ambient home surveillance.

8. Reliability

The scale exhibits high internal consistency reliability across varied empirical samples and contextual testing conditions:

• Cronbach’s Alpha (α): In the primary validation samples reported by Uysal et al. (2022), Cronbach’s alpha coefficients for the 3-item measure consistently ranged between α = .92 and α = .94. These values reflect exemplary internal consistency without reaching the redundancy threshold (> .95) that can indicate excessive item duplication.
• Composite Reliability (CR): The composite reliability index was calculated to provide a more nuanced parameter of reliability that does not assume equal factor loadings across items. The CR values across empirical samples exceeded .92, well surpassing the standard .70 requirement for psychometric instruments in social and organizational research.
• Inter-Item Correlations: Item-to-total correlations for each of the three indicators consistently exceeded .80, indicating that each item contributes strongly and reliably to the focal construct of personal data privacy competence.

9. Factor Analysis

The structural dimensionality of the Competence with the AI Platform scale was confirmed via both exploratory and confirmatory analytic frameworks:

Exploratory Factor Analysis (EFA)

During initial structural testing, maximum likelihood exploratory factor analysis with oblique rotation (promax) extracted a single clean factor with an eigenvalue substantially exceeding the Kaiser-Guttman criterion threshold of 1.0 (eigenvalues > 2.50). The primary single factor accounted for more than 82% of the total variance across the three items, with no cross-loadings or secondary dimensions identified. Scree plot analyses consistently supported this strict unidimensional structure.

Confirmatory Factor Analysis (CFA)

When evaluated within complete multi-construct structural measurement models, the three-item unidimensional specification exhibited outstanding model fit parameters. Because a saturated three-item single-factor model exhibits zero degrees of freedom when evaluated in total isolation, fit statistics were formally derived within larger measurement frameworks comprising the nomological network of AI interaction variables. Representative model fit indices across studies met or exceeded stringent psychometric cutoff criteria:

• Comparative Fit Index (CFI): > .98
• Tucker-Lewis Index (TLI): > .97
• Root Mean Square Error of Approximation (RMSEA): < .05 (with 90% confidence intervals spanning .00 to .07)
• Standardized Root Mean Square Residual (SRMR): < .03

Standardized item factor loadings across the three items were consistently high: Item 1 (λ ≈ .89), Item 2 (λ ≈ .93), and Item 3 (λ ≈ .92). These empirical metrics confirm that the scale operates as an exceptionally coherent, unidimensional measurement model.

10. Instrument / Measurement Tool

The instrument is specified below in structured psychometric format:

• Test Type: Self-report psychometric rating scale.
• Construct Measured: Perceived competence and self-efficacy regarding the protection of personal data privacy in an AI ecosystem.
• Target Platform/Context: Originally developed for Amazon Alexa; readily adaptable to alternative AI ecosystems (e.g., Apple Siri, Google Assistant, OpenAI ChatGPT, Microsoft Copilot) by substituting the device/agent name.
• Number of Items: 3 items.
• Item Phrasing / Structure: Direct, first-person declarative statements assessing knowledge, behavioral capability, and internal confidence.
• Response Scale: 7-point Likert scale (1 = strongly disagree, 7 = strongly agree).
• Scoring Rules: No reverse-coded items are present. Items are averaged to create an overall competence score, yielding an aggregate continuous index ranging from 1.00 to 7.00. Higher mean scores indicate greater perceived competence and autonomy in managing personal data privacy within the AI platform.

11. Permissions & Fee and Test Year

The Competence with the AI Platform scale was published in 2022 in the Journal of the Academy of Marketing Science. Under international fair use conventions for non-commercial academic and clinical research, the three-item instrument can be utilized without direct financial fees or explicit copyright licensing applications, provided appropriate bibliographic attribution is accorded to the original authors (Uysal, Alavi, & Bezençon, 2022).

Researchers intending to modify the target agent from “Alexa” to other technological platforms or conversational agents (e.g., smart home hubs, generative AI chatbots, healthcare diagnostics systems) are advised to report the specific platform substitution in their empirical methodology to ensure transparency and reproducibility.

12. References

Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Prentice-Hall.

Bandura, A. (1997). Self-efficacy: The exercise of control. W. H. Freeman.

Brehm, J. W. (1966). A theory of psychological reactance. Academic Press.

Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18(1), 39–50. https://doi.org/10.1177/002224378101800104

Rogers, R. W. (1975). A protection motivation theory of fear appeals and attitude change. The Journal of Psychology, 91(1), 93–114. https://doi.org/10.1080/00223980.1975.9915803

Uysal, E., Alavi, S., & Bezençon, V. (2022). Trojan horse or useful helper? A relationship perspective on artificial intelligence assistants with humanlike features. Journal of the Academy of Marketing Science, 50(6), 1153–1175. https://doi.org/10.1007/s11747-022-00881-8

13. Items of the Scale