Product Intelligence (Autonomy) (PI)

1. Abstract

The Product Intelligence (Autonomy) (PI) subscale is a specialized psychometric instrument designed to measure consumers’ subjective perceptions of a technological product’s capacity to operate in an independent, self-directed, and goal-oriented manner without necessitating ongoing human intervention. Developed and psychometrically validated by Rijsdijk, Hultink, and Diamantopoulos (2007), this subscale forms a vital operational pillar of the broader multi-dimensional Product Intelligence framework, which conceptually integrates cognitive and behavioral dimensions such as autonomy, the ability to learn, reactivity, adaptability, communication ability, and cooperation. The Autonomy dimension specifically consists of four positively keyed items evaluated on a standard 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree).

Extensive psychometric evaluations confirm the instrument’s robust measurement integrity. Across initial scale development and cross-validation samples comprising varied product categories (ranging from automated household appliances to smart consumer electronics), the subscale demonstrated strong internal consistency reliability, with Cronbach’s alpha coefficients routinely exceeding α = .85 and composite reliability (CR) surpassing .86. Confirmatory factor analysis (CFA) supports a unidimensional structure for this specific facet, exhibiting high standardized factor loadings (λ ≥ .75), satisfactory average variance extracted (AVE > .60), and rigorous discriminant validity against adjacent constructs such as reactivity, ease of use, and perceived technological complexity. By evaluating how autonomous operations systematically impact user expectations, perceived competence, usability trade-offs, and downstream consumer satisfaction, the Product Intelligence (Autonomy) subscale provides a theoretically grounded, empirically resilient diagnostic tool for human-computer interaction (HCI) researchers, consumer psychologists, and industrial product designers.

2. Keywords

product intelligence, autonomy, smart products, consumer behavior, technological perception, psychometrics, human-computer interaction, perceived autonomy, goal-directed behavior, consumer satisfaction

3. Authors

The Product Intelligence (Autonomy) subscale was conceptualized, operationalized, and validated by an interdisciplinary team of scholars specializing in marketing, new product development, and quantitative methodology:

• Serge A. Rijsdijk, Ph.D. — Associate Professor of Marketing and Innovation, Rotterdam School of Management, Erasmus University, Rotterdam, The Netherlands. Dr. Rijsdijk’s scholarship focuses on the organizational and consumer adoption dynamics of intelligent, smart, and autonomous products, as well as new product development processes.
• Erik Jan Hultink, Ph.D. — Professor of New Product Marketing, Faculty of Industrial Design Engineering, Delft University of Technology, Delft, The Netherlands. Professor Hultink has published extensively on competitive launch strategies, industrial design, consumer evaluation of innovation, and product portfolio management.
• Adamantios Diamantopoulos, Ph.D. — Professor of International Marketing and Chaired Professor, Department of International Business, Faculty of Business, Computing and Economics, University of Vienna, Vienna, Austria. Professor Diamantopoulos is an internationally renowned psychometrician and methodologist in business and consumer research, recognized for his foundational work on structural equation modeling, scale development procedures, and formative versus reflective index construction.

4. Purpose

The primary purpose of the Product Intelligence (Autonomy) subscale is to quantify the extent to which human users perceive an artifact or smart device as exhibiting autonomous functioning. In classical engineering and artificial intelligence literature, autonomy is often quantified via objective algorithmic capabilities, sensor bandwidth, or deterministic control loops. However, in the realm of consumer psychology, technology adoption, and human-machine interaction, the subjective perception of the end user represents the critical determinant of behavior, cognitive appraisal, and adoption attitudes. The scale addresses this theoretical gap by translating abstract computational autonomy into a standardized, consumer-facing psychological measurement model.

As consumer goods increasingly integrate sensors, microprocessors, machine-learning models, and cloud-based actuation, products have evolved from passive, user-directed tools into proactive, semi-autonomous agents. Devices such as robotic vacuum cleaners, autonomous lawn mowers, automated thermostats, and advanced driver assistance systems (ADAS) alter the fundamental nature of the user-artifact relationship. The scale serves several critical empirical and practical purposes:

• Diagnostic Research in Technology Adoption: It allows investigators to isolate the psychological mechanisms through which autonomous features influence user perceptions. While autonomy can heighten convenience and liberate user cognitive resources, it can simultaneously induce feelings of loss of control, technological anxiety, and apprehension regarding safety and privacy.
• Resolving Paradoxes of Modern Innovation: Technology paradoxes suggest that the same functional attributes that foster consumer empowerment can trigger feelings of vulnerability or frustration. The scale enables researchers to model non-linear relationships, such as threshold effects where autonomy positively influences satisfaction up to the point where it undermines user agency.
• Product Design and Industrial Ergonomics: For industrial engineers and interface designers, the subscale serves as a benchmark metric during iterative prototyping. By administering the scale across different user testing phases, design teams can determine whether algorithmic updates or automated feedback mechanisms actually communicate perceived autonomy to consumers without inducing cognitive overload.
• Cross-Category Comparative Analytics: The generic, artifact-neutral wording of the items enables cross-comparative studies across heterogeneous product domains, spanning wearable computing, smart home environments, smart appliances, and autonomous vehicles.

5. Psychological Construct

The psychological construct captured by this instrument is Perceived Product Autonomy, defined conceptually as the degree to which an individual believes that a product possesses the agency, internal governance, and capability to execute operations, make decisions, and achieve targeted outcomes in an independent, goal-directed manner without user intervention (Rijsdijk et al., 2007).

Within the broader multi-dimensional paradigm of Product Intelligence, autonomy occupies a central position. Rijsdijk and colleagues delineated that intelligent behavior in consumer goods is not merely a reflection of sheer processing speed or network connectivity, but rather mirrors core constructs historically explored in cognitive psychology and artificial intelligence. To understand perceived product autonomy, it must be decomposed into its theoretical components:

• Self-Initiation and Independent Action: Traditional tools rely entirely on external mechanical initiation; a mechanical vacuum or manual iron executes zero processes without continuous muscular and cognitive steering. In contrast, an autonomous product initiates actions based on internal states, programmed schedules, or environmental triggers. Item 1 (“This product can take independent action“) captures this baseline perception of self-propulsion and agency.
• Independent Decision-Making: Autonomy requires an algorithmic arbitration mechanism. When faced with environmental variations (e.g., encountering an unexpected obstacle or detecting fluctuations in ambient room temperature), the product does not freeze or revert to passive states; it selects among alternative behavioral pathways. Item 2 (“This product can make decisions by itself“) gauges whether users attribute an internal decision-making locus to the technological entity.
• Autonomous Execution: The third facet pertains to operational independence during execution. While a user may provide an initial high-level intent (e.g., “clean the floor” or “maintain comfortable temperature”), the sequence of micro-operations, trajectory planning, power management, and error correction is handled entirely by the artifact. Item 3 (“This product can execute commands independently“) operationalizes this autonomous execution capability.
• Goal-Directed Behavior (Teleology): True psychological attribution of autonomy relies on teleological perception—the belief that the artifact’s actions are organized toward achieving a stable end-state. Item 4 (“This product can pursue a goal on its own“) reflects whether users perceive the machine’s actions as coherent, deliberate striving toward a functional objective rather than random or purely reflexive motion.

Crucially, perceived autonomy is an attributional phenomenon. When individuals interact with systems possessing opaque internal logic, they tend to anthropomorphize or adopt an “intentional stance” (Dennett, 1987), treating the machine as an intentional agent endowed with goals and beliefs. Consequently, perceived autonomy operates as a psychological bridge between objective software logic and user affective responses.

6. Theoretical Framework

The Product Intelligence (Autonomy) instrument is grounded at the intersection of several prominent theoretical paradigms within cognitive psychology, human-computer interaction, and consumer behavior:

1. The Intentional Stance and Theory of Mind

Daniel Dennett’s (1987) philosophical theory of the Intentional Stance posits that humans predict and interpret the behavior of complex entities (both biological organisms and complex computational systems) by assuming they possess rationality, beliefs, desires, and goals. When smart products display complex responsive behaviors, users naturally step beyond the “physical stance” (governed by physics) and the “design stance” (governed by engineered functionality) into the intentional stance. The Autonomy subscale measures constructs directly aligned with this stance: attributing independent decision-making and teleological pursuit to an inanimate object reflects the human propensity to ascribe proto-cognitive agency to machines.

2. Parasocial Interaction and the Computers Are Social Actors (CASA) Paradigm

The foundational work of Byron Reeves and Clifford Nass (Reeves & Nass, 1996) demonstrated that individuals mindlessly apply human social rules and heuristics to interactive technologies. When a product behaves autonomously, it transitions from being an extended physical tool to functioning as a quasi-social partner or teammate. The perception of autonomy measured by Rijsdijk et al.’s instrument captures this relational shift, providing an empirical basis for investigating how perceived agency alters expectations of responsiveness, accountability, and partnership in human-machine dyads.

3. Cognitive Appraisals and Perceived Control Theory

Within cognitive psychology, psychological well-being and satisfaction are heavily determined by feelings of personal agency and perceived locus of control (Rotter, 1966; Deci & Ryan, 2000). When an external product exercises high autonomy, it introduces a dynamic tension: it can relieve the user of cognitive burdens (enhancing perceived competence and freedom from mundane chores), but it can also threaten the user’s need for personal control. The theoretical model established by Rijsdijk, Hultink, and Diamantopoulos (2007) posits that perceived product autonomy influences customer satisfaction through dual mediating mechanisms: directly by enhancing perceived product performance and functional benefit, and indirectly (sometimes negatively) through perceived complexity and loss of user sovereignty.

7. Validity

The psychometric validity of the Product Intelligence (Autonomy) subscale was comprehensively established through rigorous developmental procedures and empirical validation protocols across multiple independent samples in Rijsdijk et al. (2007).

Content and Face Validity

The initial generation of items was grounded in an extensive review of literature spanning cognitive psychology, distributed artificial intelligence, robotics, and marketing science. A pool of candidate items was subjected to critical evaluation by panels of expert judges comprising industrial designers, consumer behavior specialists, and methodologists. Items were refined, pruned, and pre-tested to eliminate linguistic ambiguity, double-barreled phrasing, and technical jargon, ensuring that typical consumers could intuitively evaluate the statements without formal training in computer science.

Convergent and Discriminant Validity

Convergent validity was substantiated through confirmatory factor analysis (CFA), where all four items demonstrated large, statistically significant standardized factor loadings (λ ≥ .75, p < .001) on the latent autonomy dimension. The Average Variance Extracted (AVE) substantially exceeded the recommended threshold of .50, registering at approximately .62 to .68 across distinct validation datasets, indicating that the latent construct explains far more variance in the indicators than does measurement error.

Discriminant validity was established utilizing the rigorous Fornell-Larcker criterion (Fornell & Larcker, 1981) as well as cross-construct correlation matrices. The square root of the AVE for the Autonomy subscale exceeded all inter-construct correlations between autonomy and the adjacent dimensions of Product Intelligence (including reactivity, ability to learn, ability to communicate, adaptivity, and cooperation). Furthermore, the scale demonstrated clear discriminant validity against external technological constructs such as perceived product complexity, ease of use (derived from the Technology Acceptance Model), and general product aesthetics.

Predictive and Nomological Validity

Nomological validity was demonstrated within structural equation models examining the downstream behavioral and cognitive effects of product intelligence dimensions. Perceived autonomy exhibited robust predictive relationships with perceived relative advantage and customer satisfaction. The researchers demonstrated that autonomy interacts with user expertise: consumers with high technological self-efficacy derived positive utility and satisfaction from high product autonomy, whereas novices occasionally experienced heightened perceptions of complexity, validating the theoretical nomological network postulated by the authors.

8. Reliability

The Product Intelligence (Autonomy) subscale consistently exhibits high internal consistency and measurement precision across empirical investigations:

• Cronbach’s Alpha (α): In the initial validation study across multiple product categories (e.g., smart vacuum cleaners, automated heating systems, advanced digital cameras), Cronbach’s alpha coefficients consistently ranged from α = .84 to α = .89, well above the conventional benchmark of .70 recommended for academic research.
• Composite Reliability (CR): Structural equation modeling revealed composite reliability values ranging between CR = .85 and CR = .88, indicating that the four reflective indicators represent a cohesive, unidimensional latent measurement model without redundant or problematic items.
• Item-to-Total Correlations: Corrected item-total correlations for each of the four items routinely exceeded .65, reflecting strong individual contribution to the measurement of the underlying autonomy construct.
• Cross-Sample Stability: Subsequent empirical investigations in consumer robotics and intelligent vehicle interfaces have replicated these reliability metrics, demonstrating consistency across different national samples and technological categories.

9. Factor Analysis

The underlying dimensionality and factorial integrity of the Product Intelligence (Autonomy) scale were confirmed using both Exploratory Factor Analysis (EFA) during preliminary item purification and Confirmatory Factor Analysis (CFA) during final validation (Rijsdijk et al., 2007).

Exploratory Factor Analysis (EFA)

During initial scale development across an exploratory sample, principal component analysis with varimax rotation was executed on the broad pool of candidate product intelligence items. The four autonomy items cleanly loaded onto a single distinct factor with an eigenvalue well above 1.0, accounting for substantial variance. Cross-loadings onto adjacent factors (such as product reactivity or ability to learn) were minimal (typically < .25), demonstrating a crisp factor separation.

Confirmatory Factor Analysis (CFA)

The definitive psychometric structure was tested via CFA using maximum likelihood estimation in structural equation modeling software. The four items were specified as reflective indicators of a single first-order latent variable (Perceived Autonomy). Key parameters and fit indices obtained across validation models demonstrated exceptional model-to-data fit:

• Standardized Factor Loadings (λ):
  • Item 1 (Independent action): λ ≈ .77 – .82
  • Item 2 (Decisions by itself): λ ≈ .80 – .85
  • Item 3 (Execute commands independently): λ ≈ .75 – .79
  • Item 4 (Pursue a goal on its own): λ ≈ .76 – .81
• Model Fit Indices: The first-order measurement model, whether estimated individually or embedded within the full multi-dimensional Product Intelligence CFA model, satisfied conventional standards of structural adequacy:
  • Comparative Fit Index (CFI) ≥ .96
  • Tucker-Lewis Index (TLI) ≥ .95
  • Root Mean Square Error of Approximation (RMSEA) ≤ .055 (with 90% confidence intervals spanning .000 to .075)
  • Standardized Root Mean Square Residual (SRMR) ≤ .040
  • χ²/df ratio < 2.5

These findings demonstrate that the four items form a statistically parsimonious, unifactorial measure that operates effectively as an independent subscale or as a sub-component within higher-order hierarchical models of product intelligence.

10. Instrument / Measurement Tool

The Product Intelligence (Autonomy) subscale is structured as follows:

• Test Type: Self-report psychological measurement scale / psychometric questionnaire.
• Format: Standardized self-administered survey administered either via paper-and-pencil questionnaires, digital web-based surveying platforms, or post-interaction usability lab evaluations.
• Number of Items: 4 items.
• Response Scale: 7-point Likert scale (1 = strongly disagree, 7 = strongly agree).
• Scoring Rules:
  • All 4 items are positively keyed (worded in the direction of higher perceived autonomy).
  • No reverse scoring is required.
  • Subscale Score Calculation: An overall autonomy score is derived by computing the arithmetic mean of the 4 items (yielding a continuous score from 1.00 to 7.00), or alternatively, by summing all items (yielding a raw sum score from 4 to 28). Higher scores reflect a stronger perception that the product operates autonomously, independently, and in a self-directed manner.
  • In structural equation modeling (SEM), the items are modeled as reflective continuous indicators loading onto a single latent construct.

11. Permissions & Fee and Test Year

• Publication Year: 2007.
• Original Publication: Journal of the Academy of Marketing Science (Volume 35, Issue 3, pp. 340–356).
• Intellectual Property & Permissions: The scale was developed for academic, scientific, and empirical research purposes. In accordance with standard fair use conventions in behavioral research, academic scholars and non-commercial researchers may utilize the four scale items in non-commercial scientific studies without paying licensing fees, provided that appropriate scholarly attribution is accorded by citing Rijsdijk, Hultink, and Diamantopoulos (2007). Commercial organizations, market research firms, or proprietary product testing entities seeking to embed the scale into commercial diagnostic software suites should verify terms with the publisher (Springer Science+Business Media / Academy of Marketing Science).

12. References

• Deci, E. L., & Ryan, R. M. (2000). The “what” and “why” of goal pursuits: Human needs and the self-determination of behavior. Psychological Inquiry, 11(4), 227–268. https://doi.org/10.1207/S15327965PLI1104_01
• Dennett, D. C. (1987). The intentional stance. MIT 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
• Reeves, B., & Nass, C. (1996). The media equation: How people treat computers, television, and new media like real people and places. Cambridge University Press.
• Rijsdijk, S. A., Hultink, E. J., & Diamantopoulos, A. (2007). Product intelligence: Its conceptualization, measurement and impact on consumer satisfaction. Journal of the Academy of Marketing Science, 35(3), 340–356. https://doi.org/10.1177/0092070307303001
• Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs: General and Applied, 80(1), 1–28. https://doi.org/10.1037/h0092976

13. Items of the Scale