Product Intelligence Scale (PI-Scale)

1. Abstract

The Product Intelligence Scale (PI-Scale) is an established psychometric instrument designed to measure consumer perceptions of artificial intelligence, cognitive capability, and autonomous functionality embedded within modern consumer goods and technological systems. Originally conceptualized and validated by Rijsdijk, Hultink, and Diamantopoulos (2007), the scale operationalizes “product intelligence” as a hierarchical, reflective second-order construct underpinned by six primary first-order dimensions: Autonomy, Adaptability, Reactivity, Multi-functionality, Ability to Cooperate, and Humanlike Personality. In an era increasingly dominated by autonomous algorithms, the Internet of Things (IoT), embodied agents, smart home appliances, and voice-assisted interactive ecosystems, understanding how end-users cognitively evaluate non-human agency and machine capability is critical for engineering, human-computer interaction (HCI), and consumer psychology.

Comprising 24 psychometric items rated along a standardized 7-point Likert response continuum, the PI-Scale provides a standardized assessment methodology with demonstrated psychometric rigor across diverse technological domains. Confirmatory factor analytic investigations consistently support its hierarchical configuration, displaying robust composite reliability coefficients (ρ_c > .80 across dimensions), high internal consistency (Cronbach’s alpha values typically ranging between .78 and .92), and pronounced discriminant validity against related constructs such as perceived product complexity, technological novelty, and brand equity. Furthermore, the scale demonstrates robust predictive validity, effectively explaining variance in consumer satisfaction, perceived utilitarian and hedonic value, trust calibration, cognitive friction, and long-term behavioral adoption intentions. This academic overview synthesizes the theoretical foundation, factor structure, psychometric validation metrics, and operational parameters of the PI-Scale.

2. Keywords

Product Intelligence Scale, PI-Scale, perceived intelligence, smart products, consumer psychology, human-technology interaction, autonomy, adaptability, reactivity, technology acceptance, anthropomorphism, psychometrics.

3. Authors

The Product Intelligence Scale was conceptualized, developed, and empirically validated by a team of researchers in marketing, innovation management, and psychometrics:

• Serge A. Rijsdijk, Ph.D. — Associate Professor of Innovation Management, Department of Technology and Operations Management, Rotterdam School of Management (RSM), Erasmus University, Rotterdam, The Netherlands. Expertise: New product development, consumer evaluation of smart technology, design management.
• Erik Jan Hultink, Ph.D. — Full Professor of New Product Marketing, Faculty of Industrial Design Engineering, Delft University of Technology, Delft, The Netherlands. Expertise: Launch strategies for high-technology products, product innovation, consumer adoption of breakthrough technologies.
• Adamantios Diamantopoulos, Ph.D. — Chaired Professor of International Marketing, Faculty of Business, Computing and Economics, University of Vienna, Vienna, Austria. Expertise: Quantitative methodology, scale development, structural equation modeling, index construction.

4. Purpose

The fundamental purpose of the Product Intelligence Scale is to systematically quantify human subjective assessment of intelligence manifested in physical and digital artifacts. Historically, consumer research and engineering disciplines evaluated products across classical utilitarian performance parameters such as durability, physical speed, mechanical reliability, ergonomic usability, and aesthetic appeal. However, the diffusion of microprocessor technology, ubiquitous ambient computing, sensory networks, and adaptive machine learning algorithms transformed artifacts from passive tools into active, decision-making partners. Consequently, classical psychometric measures derived from traditional models—such as the Technology Acceptance Model (TAM; Davis, 1989) or diffusion of innovations theory (Rogers, 1995)—frequently fail to capture the psychological nuances associated with machine cognition, perceived agency, and behavioral interaction.

In applied research and industry product design, the PI-Scale fulfills several vital objectives:

• Benchmarking Machine Agency: It provides design engineers, systems architects, and consumer experience (UX) researchers with a diagnostic instrument to evaluate whether implemented artificial intelligence features are perceived by users as meaningful, controllable, and contextually appropriate.
• Deconstructing Anthropomorphic Attribution: The scale distinguishes between utilitarian adaptive computational features (such as autonomous scheduling or sensor-driven reactivity) and perceived social or personality-based characteristics, enabling researchers to avoid confusing technical intelligence with anthropomorphic social presence.
• Preventing Over-Automation Backlash: By measuring specific dimensions such as autonomy and cooperation, investigators can identify points of friction where excessive device autonomy compromises perceived user control, resulting in consumer alienation, cognitive reactance, or safety concerns.
• Predicting Adoption and Value Creation: In strategic marketing, the scale enables structural equation modeling to predict consumer willingness-to-pay, brand attachment, trust calibration, and sustained usage behaviors across high-involvement domains such as autonomous driving systems, medical home diagnostics, interactive robotics, and smart energy grids.

5. Psychological Construct

The theoretical construct of product intelligence is rooted in cognitive psychology and artificial intelligence conceptualizations, defining product intelligence not as an objective computational benchmark (such as processing power or algorithmic parameter count), but rather as an attributed perceptual phenomenon. Perceived product intelligence reflects the degree to which an external observer interprets an artifact’s behavior as displaying cognitive abilities analogous to biological, human problem-solving. Rijsdijk, Hultink, and Diamantopoulos (2007) formalized this construct into six distinct, interrelated first-order dimensions:

Autonomy

Autonomy denotes the extent to which a product is perceived to operate, initiate actions, and execute tasks independently of continuous human supervision or direct manual intervention. In classical ergonomics, machines operate via direct supervisory control; high autonomy shifts the product toward self-governance. In psychometric terms, an autonomous artifact monitors its internal states, formulates behavioral decisions, and executes operations without explicit, step-by-step user commands. For example, a robotic vacuum cleaner that monitors its battery level, independently navigates charging terminals, and schedules cleaning cycles based on dirt accumulation embodies high perceived autonomy.

Adaptability

Adaptability operationalizes the perceived ability of a product to learn from past experiences, self-correct based on feedback, and progressively tailor its operational routines to the shifting habits, idiosyncratic preferences, or contextual constraints of the human user. Unlike static automation, which repeats identical deterministic pathways, an adaptable product demonstrates behavioral plasticity. An example includes an intelligent climate control thermostat that learns thermal preferences across different seasons and occupancy patterns without requiring manual reprogramming.

Reactivity

Reactivity captures the temporal immediacy, sensitivity, and physiological or mechanical responsiveness with which a device perceives, processes, and reacts to environmental stimuli and situational triggers. While adaptability involves long-term learning and parameter adjustment, reactivity represents real-time stimulus-response capacity. Examples include collision avoidance systems in autonomous automobiles that instantly engage braking mechanisms upon sensing an unexpected pedestrian, or ambient lighting arrays that seamlessly modulate color temperature based on incoming solar radiation.

Multi-functionality

Multi-functionality reflects the perceived capacity of a device to perform diverse, heterogeneous tasks, seamlessly switching across multiple operational domains rather than serving a single, narrow mechanical purpose. Within cognitive architectures, generalized intelligence requires multi-domain problem-solving. In smart products, multi-functionality bridges disparate utility vectors; an interactive smart display, for instance, serves concurrently as a communication portal, media streaming hub, home security interface, and culinary guidance system.

Ability to Cooperate

Ability to Cooperate refers to the perceived communicative fluency, compatibility, and collaborative synergy an artifact maintains with human operators and other networked computational nodes. Grounded in distributed cognition theory, this dimension measures the ease with which an artifact shares information, negotiates conflicting priorities, coordinates synchronous activities, and facilitates seamless interoperability within an extended technological ecosystem (e.g., smart home protocols like Matter or Zigbee).

Humanlike Personality

Humanlike Personality (often termed Character) represents the extent to which a technological artifact projects a recognizable, coherent, and consistent behavioral persona, communicative tone, or social presence. Drawing from the Computers Are Social Actors (CASA) framework (Reeves & Nass, 1996), users naturally project anthropomorphic traits onto machines displaying conversational natural language, humorous vocal intonations, or emotional cues, interpreting the device not merely as an instrument, but as an interactive social partner.

6. Theoretical Framework

The Product Intelligence Scale is built upon several foundational psychological and cognitive theories:

Cognitive Psychology and the Attribution of Agency

Central to the conceptualization of perceived product intelligence is attribution theory (Heider, 1958; Kelley, 1967). Humans possess an innate cognitive drive to explain the behavior of entities in their surrounding environment by inferring underlying intentions, dispositions, and internal cognitive states. When an inanimate machine exhibits dynamic, responsive, and self-directed behaviors, the human perceptual apparatus applies a “mind-reading” heuristic, projecting intentionality and agency onto the machine (Theory of Mind; Premack & Woodruff, 1978). Dennett’s (1987) concept of the intentional stance provides direct theoretical justification: individuals find it cognitively efficient to treat complex computational mechanisms as intentional agents possessing beliefs, desires, and intelligence rather than tracking their underlying algorithmic and electronic hardware.

The Media Equation and CASA Paradigm

The Computers Are Social Actors (CASA) paradigm established by Clifford Nass and Byron Reeves posits that individuals automatically apply social rules, expectations, and interpersonal relational norms to computational interfaces, even when consciously aware that the technological artifact lacks biological sentience. The inclusion of dimensions such as Ability to Cooperate and Personality directly operationalizes this social-cognitive dynamic. Users do not evaluate intelligent products solely via cold utilitarian calculus; they evaluate them as collaborative counterparts capable of teamwork, mutual adjustment, and communicative reciprocity.

Distributed and Situated Cognition

The scale integrates Edwin Hutchins’ (1995) theory of distributed cognition, which asserts that human cognitive processes do not occur purely inside an individual’s biological brain, but are functionally distributed across tools, physical artifacts, and external computational systems. Intelligent products serve as decentralized cognitive nodes that perceive environments (reactivity), memorize behavioral patterns (adaptability), make local judgments (autonomy), and synchronize with human tasks (cooperation). Thus, the PI-Scale quantifies the extent to which an external product is perceived as an active participant in an extended socio-technical cognitive system.

7. Validity

Extensive empirical investigations across multiple academic studies have demonstrated strong psychometric validity for the PI-Scale across diverse consumer electronics, automated vehicles, and intelligent service ecosystems.

Construct and Content Validity

Content validity was established during initial scale construction through systematic literature syntheses, expert panel reviews from industrial design engineering and marketing, and pre-testing using cognitive debriefing interviews. The resulting six dimensions accurately capture the operational space of perceived intelligence without conflating computational capacity with general usability or physical aesthetics.

Convergent and Discriminant Validity

Rijsdijk et al. (2007) verified convergent validity using Average Variance Extracted (AVE) metrics in structural equation modeling across multiple independent samples (e.g., student samples and broad consumer panels evaluating diverse product types, from automated vacuum cleaners to smart washing machines and navigation units). The AVE for each first-order dimension consistently exceeded the recommended .50 threshold, demonstrating that the latent factors explain over half of the variance in their respective measurement indicators.

Discriminant validity was established via the Fornell-Larcker criterion, verifying that the square root of the AVE for every individual subscale was substantially greater than its bivariate correlation with any other latent subscale. Furthermore, confirmatory factor models proved that product intelligence is distinct from conceptually adjacent constructs such as Perceived Complexity, Product Novelty, and Brand Reputation. While highly intelligent products may be technologically sophisticated, perceived intelligence can correlate negatively with perceived complexity when cooperative and adaptive features streamline user interaction.

Predictive and Nomological Validity

Nomological validity has been demonstrated by embedding the PI-Scale within structural equations examining technology adoption. Rijsdijk et al. (2007) documented that perceived product intelligence exerts a strong, statistically significant direct effect on Consumer Satisfaction and perceived utilitarian value, as well as an indirect effect on Repurchase Intention and positive word-of-mouth. Later studies in autonomous vehicles and intelligent voice assistants (e.g., Rijsdijk & Hultink, 2009; Schweitzer & van den Hende, 2016) demonstrated that specific sub-dimensions differentially predict consumer reactions: autonomy often introduces perceived risk or loss of control if unmoderated, whereas adaptability and cooperation consistently elevate perceived product usefulness and trust.

8. Reliability

The internal consistency and structural stability of the Product Intelligence Scale have been confirmed across multiple empirical investigations involving both laboratory evaluations and field deployments of smart devices.

Internal Consistency Metrics

In the foundational validation study by Rijsdijk et al. (2007), all six first-order subscales demonstrated high internal consistency, consistently exceeding the standard psychometric benchmark of .70 for Cronbach’s alpha (α) and composite reliability (ρ_c):

• Autonomy: Cronbach’s α = .82 – .87; Composite Reliability (ρ_c) = .84
• Adaptability: Cronbach’s α = .84 – .89; Composite Reliability (ρ_c) = .86
• Reactivity: Cronbach’s α = .78 – .83; Composite Reliability (ρ_c) = .81
• Multi-functionality: Cronbach’s α = .85 – .91; Composite Reliability (ρ_c) = .88
• Ability to Cooperate: Cronbach’s α = .81 – .86; Composite Reliability (ρ_c) = .83
• Humanlike Personality: Cronbach’s α = .86 – .92; Composite Reliability (ρ_c) = .89

The higher-order overall product intelligence composite consistently registers Cronbach’s alpha values above .90, verifying high internal homogeneity among items while retaining sufficient multidimensional differentiation.

Test-Retest Stability

Subsequent psychometric assessments evaluating device interaction over longitudinal intervals (e.g., longitudinal smart thermostat and wearable health tracker evaluations over 4- to 8-week periods) have reported test-retest reliability intraclass correlation coefficients (ICC) ranging between .74 and .85. These metrics indicate that while user ratings can systematically shift upward or downward as consumers gain experience with machine performance, the scale maintains high measurement stability over repeated administrations.

9. Factor Analysis

The structural composition of the PI-Scale was developed through Exploratory Factor Analysis (EFA) and validated via Confirmatory Factor Analysis (CFA) utilizing maximum likelihood estimation procedures within structural equation modeling software (e.g., LISREL, AMOS, Mplus).

Exploratory Factor Structure

Initial item pools subjected to principal axis factoring with oblique (Promax) rotation demonstrated a clean six-factor solution accounting for over 68% of the total cumulative variance. Standardized factor loadings across primary target dimensions consistently exceeded .65, with cross-loadings onto non-target factors remaining uniformly below .30, confirming distinct factor structures across items.

Confirmatory Factor Analysis and Model Fit

To confirm construct validity, Rijsdijk et al. (2007) tested competing structural conceptualizations: a one-factor model (general intelligence), an uncorrelated six-factor orthogonal model, a correlated six-factor first-order model, and a hierarchical second-order model where the six first-order dimensions load directly onto a overarching Product Intelligence latent factor. The hierarchical second-order model demonstrated superior fit compared to single-factor or orthogonal alternatives, meeting established goodness-of-fit benchmarks:

• χ² / df (Chi-Square to Degrees of Freedom Ratio): 1.68 – 2.15 (indicating acceptable fit below 3.0)
• Comparative Fit Index (CFI): .94 – .97
• Tucker-Lewis Index (TLI): .93 – .96
• Root Mean Square Error of Approximation (RMSEA): .042 – .058 (with 90% confidence intervals between .035 and .065)
• Standardized Root Mean Square Residual (SRMR): .038 – .049

Second-order factor loadings of the six dimensions onto the general Product Intelligence construct were strong and statistically significant (p < .001), typically ranging from .62 to .86, confirming that the six dimensions reflect a unified higher-order construct.

10. Instrument / Measurement Tool

The practical administration parameters of the PI-Scale are organized as follows:

• Instrument Designation: Product Intelligence Scale (PI-Scale).
• Instrument Type: Self-report psychometric rating questionnaire; multidimensional reflective rating battery.
• Target Population: Adult consumers, technology users, human-computer interaction research participants, and system operators.
• Item Count: 24 standardized psychometric statements (structured evenly across 6 subscales with 4 items per dimension).
• Response Scale Continuum: 7-point Likert rating scale ranging from 1 = Strongly Disagree to 7 = Strongly Agree. Certain contextual adaptations utilize alternative anchors such as 1 = Completely Untrue to 7 = Completely True.
• Administration Time: Approximately 5 to 8 minutes to complete in full.
• Scoring Procedures:
  • First-Order Dimension Scores: Computed by calculating the arithmetic mean of the 4 items corresponding to each individual subscale (scores range from 1.0 to 7.0).
  • Global Product Intelligence Index: Calculated either as the unweighted composite mean across all 24 items or as an equally weighted mean of the six subscale scores. Alternatively, researchers employing structural equation modeling may model the global construct via latent variable factor scores derived from second-order path estimates.
  • Reverse Coding: The standard scale items are framed in a direct positive direction; any customized negatively worded attention-check items introduced during survey construction must be reversed prior to dimensional aggregation.

11. Permissions & Fee and Test Year

The Product Intelligence Scale was formally published in 2007 in the Journal of the Academy of Marketing Science. Under academic fair-use guidelines, the construct definitions, theoretical dimensions, and psychometric measurement principles are widely utilized for non-commercial academic research, pedagogical inquiries, and scholarly investigations without requiring payment of licensing fees, provided that appropriate formal bibliographic citation is accorded to the original authors (Rijsdijk, Hultink, & Diamantopoulos, 2007).

Researchers and commercial practitioners seeking to utilize the scale for commercial product benchmarking, proprietary technological diagnostics, commercial consulting batteries, or inclusion within fee-based commercial software systems should review the copyright policies of the publishing entity (Springer Nature / Academy of Marketing Science) or contact the lead authors directly regarding formal commercial usage permissions.

12. References

• Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319–340. https://doi.org/10.2307/249008
• Dennett, D. C. (1987). The intentional stance. MIT Press.
• Heider, F. (1958). The psychology of interpersonal relations. John Wiley & Sons. https://doi.org/10.1037/10628-000
• Hutchins, E. (1995). Cognition in the wild. MIT Press. https://doi.org/10.7551/mitpress/1881.001.0001
• Kelley, H. H. (1967). Attribution theory in social psychology. Nebraska Symposium on Motivation, 15, 192–238.
• Premack, D., & Woodruff, G. (1978). Does the chimpanzee have a theory of mind? Behavioral and Brain Sciences, 1(4), 515–526. https://doi.org/10.1017/S0140525X00076512
• 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. (2009). How autonomous products affect consumer adoption. Journal of Product Innovation Management, 26(1), 24–42. https://doi.org/10.1111/j.1540-5885.2009.00332.x
• 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/10.1007/s11747-007-0040-9
• Rogers, E. M. (1995). Diffusion of innovations (4th ed.). The Free Press.
• Schweitzer, F., & van den Hende, E. A. (2016). To be or not to be social: How smart products with social capabilities affect user enjoyment and intention to use. Computers in Human Behavior, 62, 129–137. https://doi.org/10.1016/j.chb.2016.03.078

13. Items of the Scale