Product Intelligence (Human-like Interaction)

1. Abstract

The Product Intelligence (Human-like Interaction) scale is a specialized psychometric instrument formulated by Serge A. Rijsdijk, Erik Jan Hultink, and Adamantios Diamantopoulos in their seminal 2007 study published in the Journal of the Academy of Marketing Science. Developed as a core sub-dimension of the overarching multidimensional construct of Product Intelligence (PI), the scale quantifies the degree to which consumers perceive an autonomous, smart, or technology-driven product to communicate, exchange information, and interact in a natural, anthropomorphic, and human-like manner. The complete Product Intelligence battery measures seven distinct cognitive and behavioral dimensions: autonomy, adaptability, reactivity, multi-functionality, ability to cooperate, human-like interaction, and learning ability. The Human-like Interaction subscale comprises five carefully refined items evaluated on a seven-point Likert scale ranging from 1 (“Strongly Disagree”) to 7 (“Strongly Agree”). Extensive empirical validation demonstrates robust psychometric properties, including high internal consistency (Cronbach’s alpha values regularly exceeding .85 and composite reliability indices surpassing .88), exceptional convergent validity with average variance extracted (AVE) exceeding recommended benchmarks, and established discriminant validity against adjacent dimensions of smart product functionality. In structural equation models, perceived human-like interaction significantly influences consumer satisfaction, perceived ease of use, product affection, and overall adoption intentions. This psychometric tool remains central to consumer psychology, human-computer interaction (HCI), artificial intelligence adoption studies, and industrial design engineering.

2. Keywords

Product Intelligence, Human-like Interaction, Anthropomorphism, Smart Products, Human-Computer Interaction, Consumer Satisfaction, Perceived Communicativeness, Technology Acceptance, Psychometrics, Rijsdijk and Hultink, Social Agency, Artificial Intelligence

3. Authors

The scale was developed and psychometrically validated by a team of leading scholars in marketing, innovation management, and psychometrics:

• Serge A. Rijsdijk, Ph.D.: Associate Professor of Innovation Management, Rotterdam School of Management, Erasmus University, Rotterdam, The Netherlands. His research specializes in product development, smart product adoption, autonomous technologies, and new product performance metrics.
• Erik Jan Hultink, Ph.D.: Full Professor of New Product Marketing, Faculty of Industrial Design Engineering, Delft University of Technology (TU Delft), Delft, The Netherlands. An authority on launch strategies, industrial design engineering, innovation commercialization, and consumer evaluations of technologically sophisticated consumer goods.
• Adamantios Diamantopoulos, Ph.D.: Chaired Professor of International Marketing, Faculty of Business, Computer Science and Statistics, University of Vienna, Austria. A prominent psychometrician and methodologist in marketing science, internationally renowned for his work on formative versus reflective measurement models, scale development procedures, and structural equation modeling.

4. Purpose

The advent of microprocessors, ubiquitous sensor networks, and natural language processing in everyday consumer goods has transformed static mechanical artifacts into dynamic, communicative agents. Historically, product evaluations focused on mechanical durability, core utility, and ergonomic usability. However, as appliances, automobiles, domestic robots, and wearable devices began providing voice feedback, diagnosing their own faults, and conversing with users, traditional psychometric instruments failed to capture the cognitive attribution of “intelligence” made by end-users. The primary purpose of the Human-like Interaction scale is to quantify consumer perceptions of a product’s social and linguistic communicative capacity—specifically, whether the artifact communicates in natural, bilateral, and human-like ways rather than through rigid, cryptic, or purely mechanical codes.

From a research perspective, the instrument enables scholars in human-computer interaction, consumer behavior, and artificial intelligence to isolate the communicative facet of smart devices from their technical autonomy or computational capacity. Researchers deploy the scale to understand how human-like communicative features affect consumers’ psychological comfort, reduction of cognitive load, feelings of technological anxiety, and anthropomorphic evaluations. The instrument provides a standardized, empirically validated operationalization that prevents ad-hoc, unvalidated question phrasing across empirical investigations.

In applied product design, clinical usability trials, and commercial market research, the scale serves as a critical diagnostic metric. Engineers and user experience (UX) designers utilize the instrument during iterative prototyping to assess whether conversational interfaces (such as speech agents, conversational displays, or domestic assistive robots) feel genuinely intuitive and natural or whether they induce dissonance. Furthermore, marketing analysts employ the scale to forecast consumer satisfaction and adoption willingness. Overly technical interfaces often alienate non-expert users; conversely, interfaces that score high on human-like interaction bridge the psychological divide between user and machine, transforming a utilitarian apparatus into an approachable partner.

5. Psychological Construct

The psychological construct captured by this instrument is Human-like Interaction, situated as a foundational reflective dimension of the broader Product Intelligence paradigm. In psychological and marketing literature, product intelligence is defined as the degree to which a product exhibits behaviors and capabilities that are traditionally attributed to human cognitive faculties. Within this overarching framework, Rijsdijk, Hultink, and Diamantopoulos (2007) delineate Human-like Interaction as the specific capability of a product to communicate and interface with human operators using natural, intuitive, and social modalities comparable to interpersonal human communication.

Human-like Interaction involves several underlying psychological components:

• Linguistic Naturalness and Semantic Fluency: Rather than forcing users to decipher flashing binary indicator lamps, numeric error codes, or convoluted menu hierarchies, a product high in human-like interaction uses natural language, speech feedback, and human-compatible dialogue. It presents information in conversational terms that mimic interpersonal dialogue.
• Bilateral and Reciprocal Communication: The construct captures the perception of a true two-way exchange. The user perceives that the machine not only receives commands but also communicates back in an expressive, understandable, and context-aware manner, closing the interactive loop.
• Social and Anthropomorphic Attribution: The construct taps into the user’s subconscious or conscious tendency toward anthropomorphism. Consumers attribute intentionality, social presence, and conversational agency to products that score highly on this dimension, treating the interface as a social entity rather than a cold, inert assembly of plastic and circuitry.
• Intuitive Behavioral Alignment: High human-like interaction implies that the artifact adapts its interactive cues to normal human social norms. This includes timely verbal confirmations, polite query structures, and understandable warnings that do not require specialized technical expertise to decipher.

It is vital to distinguish Human-like Interaction from related constructs within the Product Intelligence framework. While Autonomy denotes a product’s capacity to act independently without human intervention, and Adaptability reflects the ability to alter operations based on environmental changes, Human-like Interaction is strictly relational and communicative. A product may possess massive computational autonomy (such as an algorithmic power grid optimizer) while having virtually zero human-like interaction. Conversely, an interactive personal assistant or voice-guided vehicle navigation device prioritizes the communicative bridge, rendering human-like interaction its defining psychological feature.

6. Theoretical Framework

The scale is anchored in several established paradigms across social psychology, cognitive science, and communication theory. The primary theoretical foundation is the Computers are Social Actors (CASA) paradigm advanced by Clifford Nass, Byron Reeves, and their colleagues at Stanford University (Reeves & Nass, 1996). The CASA paradigm posits that individuals automatically and mindlessly apply social rules, polite expectations, and interpersonal communicative heuristics to interactive media and computational devices whenever those entities display minimal human-like cues (such as voice, turn-taking, or reciprocal language). Rijsdijk and colleagues operationalized this principle within marketing by demonstrating that smart consumer products trigger similar social responses when equipped with human-compatible interaction systems.

A second foundational pillar is the Three-Factor Theory of Anthropomorphism formulated by Epley, Waytz, Akalis, and Cacioppo (2007). This framework explains that humans anthropomorphize non-human entities driven by three cognitive motives: elicited agent knowledge (applying accessible human schemas to non-human targets), effectance motivation (the desire to make sense of, master, and predict the behavior of complex technological systems), and social contact motivation (the innate desire for human-like social affiliation). When a smart product demonstrates human-like interaction, it satisfies the user’s effectance needs by rendering complex machine states interpretable through everyday conversational norms, thereby reducing perceived uncertainty and cognitive strain.

Finally, the scale integrates principles from the Technology Acceptance Model (TAM) (Davis, 1989) and Expectancy-Disconfirmation Theory (Oliver, 1980). Within modern technology adoption modeling, perceived human-like interaction operates as an antecedent to perceived ease of use and perceived enjoyment. When consumers interact with a product that communicates like a human, the cognitive friction associated with learning complex operational protocols drops significantly. This seamlessness leads to positive expectation confirmation, which directly fuels consumer product satisfaction and emotional attachment.

7. Validity

The psychometric validity of the Human-like Interaction scale was rigorously evaluated by Rijsdijk, Hultink, and Diamantopoulos (2007) across multiple methodological phases, adhering to contemporary psychometric standards for scale purification and construct validation:

• Content and Face Validity: Initial item pools were generated through qualitative consumer depth interviews, focus group discussions, and thorough literature reviews in robotics, consumer electronics, and cognitive psychology. A panel of international experts in marketing, product design, and psychometrics reviewed the items to ensure semantic clarity, conciseness, and unambiguous representation of human-like communication. Iterative expert rating rounds eliminated redundant or confusing indicators, resulting in the refined five-item set.
• Convergent Validity: In confirmatory factor analysis (CFA), the five items loaded strongly and significantly on the latent Human-like Interaction dimension. All standardized factor loadings ($lambda$) exceeded the recommended .70 threshold (with loadings ranging between .74 and .89, all statistically significant at $p < .001$). The Average Variance Extracted (AVE) substantially exceeded the recognized .50 benchmark (reaching .65 to .72 across independent product category samples), demonstrating that the majority of variance captured by the indicators is shared with the underlying theoretical construct rather than measurement error.
• Discriminant Validity: Discriminant validity was empirically established using the Fornell-Larcker criterion. The square root of the AVE for the Human-like Interaction construct was systematically higher than any pairwise correlation between this dimension and any other Product Intelligence dimension (autonomy, adaptability, reactivity, multi-functionality, cooperation, and learning ability). Additionally, nested model $\chi^2$ difference tests confirmed that constraining the correlation between Human-like Interaction and adjacent constructs to unity (1.0) led to a significant deterioration in model fit ($\Delta\chi^2, p < .001$), confirming empirical distinctiveness.
• Nomological and Criterion-Related Validity: In structural equation modeling (SEM), Human-like Interaction exhibited predictable, theoretically sound paths to downstream dependent constructs. The construct exhibited strong positive relationships with perceived product usability, perceived product value, and overall consumer satisfaction. Further cross-sectional investigations in smart home ecosystems and autonomous vehicles have corroborated its predictive validity regarding consumer trust, adoption intention, and willingness to pay price premiums.

8. Reliability

The Human-like Interaction subscale has demonstrated high internal consistency and stability across diverse empirical samples and product categories. Psychometric evaluations yield the following reliability parameters:

• Cronbach’s Alpha ($lpha$): In the initial scale purification and validation studies across diverse product categories (such as intelligent washing machines, robot vacuum cleaners, and satellite navigation systems), Cronbach’s alpha values for the five-item subscale ranged consistently from .88 to .91, well above the standard academic acceptability benchmark of .70 and the stringent .80 threshold required for basic research.
• Composite Reliability (CR): Because Cronbach’s alpha assumes tau-equivalence (equal factor loadings across items) and can underestimate reliability in congeneric measurement models, composite reliability ($
  ho_c$) was calculated. The CR values for the Human-like Interaction dimension routinely surpass .89, providing strong evidence of composite internal consistency.
• Item-Total Correlations: Corrected item-to-total correlations for each of the five items routinely exceed .65, with none falling below the .50 cutoff, confirming that all items contribute meaningfully to the measurement of the underlying construct without introducing multidimensional noise.
• Cross-Category Replicability: Subsequent replication studies evaluating interactive conversational assistants, in-car voice systems, and smart health devices have repeatedly demonstrated alpha coefficients between .86 and .93, confirming that the scale’s internal consistency remains robust across diverse technological formats and respondent demographics.

9. Factor Analysis

The factorial validity of the scale was established through sequential exploratory and confirmatory factor analytic procedures:

Exploratory Factor Analysis (EFA)

During initial exploratory analyses, principal axis factoring with oblique rotation (promax) was conducted on the full pool of product intelligence indicators. The items designed to measure Human-like Interaction loaded onto a dedicated, distinct factor with high eigenvalues (> 1.0) and high factor loadings ranging from .72 to .88. No cross-loadings above .25 onto other intelligence dimensions (such as reactivity or autonomy) were observed, confirming clear factor separation.

Confirmatory Factor Analysis (CFA)

A multi-sample confirmatory factor analysis was conducted using structural equation modeling (maximum likelihood estimation) to test both first-order independent factor structures and higher-order Product Intelligence models. The measurement model demonstrated excellent fit to the empirical data across multiple validation samples, satisfying standard cutoff criteria:

• Chi-Square to Degrees of Freedom Ratio ($\chi^2/df$): Ranged between 1.45 and 2.10, comfortably within the desirable threshold of < 3.0.
• Comparative Fit Index (CFI): Values regularly exceeded .96 (benchmark > .95).
• Tucker-Lewis Index (TLI / NNFI): Values consistently surpassed .95 (benchmark > .95).
• Root Mean Square Error of Approximation (RMSEA): Values ranged between .038 and .054, with 90% confidence intervals below .08, indicating close approximate fit.
• Standardized Root Mean Square Residual (SRMR): Observed values were consistently below .045, well under the .08 ceiling.

Item loadings across the five indicators were uniformly high and statistically significant ($p < .001$), with standardized coefficients consistently falling between .74 and .89. These analytical outcomes confirm that the five manifest indicators operate as an essentially unidimensional, psychometrically clean reflective measurement model.

10. Instrument / Measurement Tool

The structural and administrative specifications of the Human-like Interaction subscale are detailed below:

• Instrument Type: Self-administered psychometric rating scale / subjective perception questionnaire.
• Measurement Paradigm: Multi-item reflective measurement scale.
• Target Unit of Analysis: Consumer perceptions of a specific physical, digital, or cyber-physical product.
• Item Inventory: 5 manifest indicator items.
• Administration Format: Paper-and-pencil, computer-assisted personal interviewing (CAPI), or web-based survey interfaces.
• Estimated Completion Time: Approximately 1 to 2 minutes for the 5-item subscale (or 5 to 7 minutes if administered alongside the entire 7-dimension Product Intelligence battery).
• Response Format: Seven-point Likert scale:
  • 1 = Strongly Disagree
  • 2 = Disagree
  • 3 = Somewhat Disagree
  • 4 = Neutral (Neither Agree nor Disagree)
  • 5 = Somewhat Agree
  • 6 = Agree
  • 7 = Strongly Agree
• Scoring Protocol: All items are positively framed (no reverse coding required in standard implementation). Researchers calculate an aggregate score by either:
  • Unweighted Mean Scoring: Summing the scores across all five items and dividing by 5, preserving the original 1 to 7 interpretative continuum.
  • Latent Factor Scoring: Extracting standardized latent factor scores within a structural equation modeling environment (e.g., via AMOS, Mplus, or lavaan in R), which weights items based on their empirical factor loadings.
• Score Interpretation: Higher mean scores (e.g., > 5.0) reflect a strong perception that the product interacts, communicates, and exchanges dialogue in an anthropomorphic, natural manner. Lower scores (< 3.0) signify that the consumer perceives the product’s interactive modality as strictly mechanical, impersonal, or deficient in conversational naturalness.

11. Permissions & Fee and Test Year

The Human-like Interaction scale was published in 2007 in the Journal of the Academy of Marketing Science (Volume 35, Issue 3, pages 340–356). The journal is published by Springer Nature on behalf of the Academy of Marketing Science. The copyright for the academic publication resides with the Academy of Marketing Science / Springer Nature.

For non-commercial, academic research, and educational purposes, researchers are generally permitted to use the scale items in surveys, theses, and scholarly inquiries without monetary licensing fees, provided that appropriate attribution and bibliographic citation are accorded to Rijsdijk, Hultink, and Diamantopoulos (2007). For commercial deployment, proprietary market consulting, software integration, or commercial benchmarking tools, formal permission or licensing may need to be secured from the copyright holder (Springer Nature via RightsLink) or directly through the original authors.

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
• Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: A three-factor theory of anthropomorphism. Psychological Review, 114(4), 864–886. https://doi.org/10.1037/0033-295X.114.4.864
• Nass, C., Steuer, J. S., & Tauber, E. R. (1994). Computers are social actors. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 72–78). https://doi.org/10.1145/191666.191703
• Oliver, R. L. (1980). A cognitive model of the antecedents and consequences of satisfaction decisions. Journal of Marketing Research, 17(4), 460–469. https://doi.org/10.1177/002224378001700405
• 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. (2003). ‘Honey, have you seen the car keys?’ A study on consumers’ and designers’ perceptions of product intelligence. Journal of Product Innovation Management, 20(3), 204–216. https://doi.org/10.1111/1540-5885.2003004
• Rijsdijk, S. A., & Hultink, E. J. (2009). How smart products create customer value. Journal of Product Innovation Management, 26(1), 24–40. https://doi.org/10.1111/j.1540-5885.2009.00334.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/s11747-007-0040-6

13. Items of the Scale