1. Abstract
The Humanlikeness of the Object scale is an empirical psychometric instrument developed by Garvey, Kim, and Duhachek (2023) within the domain of consumer psychology and human-computer interaction (HCI). As artificial intelligence (AI), autonomous algorithmic agents, and robotic interfaces increasingly assume roles traditionally reserved for human communicators, understanding how recipients psychologically construe these entities is of paramount theoretical and practical importance. The instrument quantitatively assesses the degree to which an observer perceives a non-human entity, technological interface, or automated agent as resembling a human being in both outward aesthetic/physical characteristics and inner behavioral/cognitive attributes. Evaluated typically via multidimensional semantic differential and multi-item Likert formats, the construct captures dual underlying dimensions: morphological/visual humanlikeness (perceived anatomical, visual, or cosmetic resemblance) and behavioral/cognitive humanlikeness (perceived intentionality, empathy, communicative naturalness, and mental capacity). Across experimental investigations examining communications of varying valence—such as corporate announcements, service failure resolutions, and medical or financial disclosures—the instrument demonstrates exceptional internal consistency (Cronbach’s alpha coefficients typically exceeding α = .88), robust construct validity, and pronounced convergent and discriminant validity against related constructs such as social presence, warmth, and competence. By systematically indexing humanlikeness, the instrument allows researchers to predict critical downstream psychological and behavioral outcomes, including attribution of blame, perceived sincerity, emotional coping mechanisms, and overall service evaluation.
2. Keywords
Humanlikeness of the Object, Anthropomorphism, Artificial Intelligence, Human-Robot Interaction, Mind Perception, Algorithmic Communication, Consumer Psychology, Perceived Intentionality, Social Presence, Psychometrics
3. Authors
The scale was developed and operationalized by an interdisciplinary team of researchers specializing in consumer behavior, marketing communication, and human-technology interactions:
- Aaron M. Garvey — Associate Professor of Marketing, Gatton College of Business and Economics, University of Kentucky, Lexington, KY, USA. Expertise: Consumer decision-making, consumer-technology interfaces, and branding psychology.
- TaeWoo Kim — Assistant Professor of Marketing, Kelley School of Business, Indiana University, Bloomington, IN, USA. Expertise: Technological disruptions in consumer welfare, automation perception, and psychological mind perception.
- Adam Duhachek — Professor of Marketing and Margaret A. St. Clair Professor, Kelley School of Business, Indiana University, Bloomington, IN, USA. Expertise: Affective forecasting, emotional processing under threat, stress, and algorithmic morality.
4. Purpose
The primary purpose of the Humanlikeness of the Object scale is to provide an exact, empirically rigorous metric for measuring subjective appraisals of an artificial or non-human entity’s proximity to the human archetype. Over the past two decades, organizational environments have rapidly transitioned from purely human-to-human service delivery to hybrid ecosystems wherein automated algorithmic agents deliver evaluative feedback, marketing disclosures, healthcare notifications, and financial determinations. While earlier research broadly categorized entities into binary taxonomies (i.e., human versus machine), Garvey, Kim, and Duhachek (2023) recognized that consumer evaluation operates along a graded continuum governed by nuanced psychological construals.
The theoretical rationale for the instrument stems from cognitive appraisal theories and mind perception frameworks. When individuals interact with automated systems, they inevitably make cognitive attributions regarding the agent’s underlying capacities. In high-stakes communicative interactions—such as the delivery of unfavorable outcomes (e.g., loan rejections, diagnostic bad news) versus favorable outcomes (e.g., promotions, service upgrades)—the degree of perceived humanlikeness dictates the moral, emotional, and retributive reactions of the recipient. Specifically, highly humanlike objects evoke assumptions of intentional agency, subjective feeling, and moral accountability, whereas low-humanlike (mechanistic) objects are perceived as detached, computational, and dispassionate executors of algorithmic rules.
In research contexts, the scale functions as a vital manipulation check, mediating variable, or boundary-condition moderator. In applied organizational and clinical settings, the scale assists service designers, software engineers, and health communicators in identifying optimal thresholds of anthropomorphic design. It prevents inadvertent psychological backlash, such as the evocative triggering of uncanny valley phenomena or unmerited moral blameworthiness when autonomous agents must inevitably convey adverse communications to vulnerable human stakeholders.
5. Psychological Construct
The psychological construct evaluated by this instrument is Humanlikeness, conceptually demarcated as an observer’s cognitive and perceptual assessment that an artificial, synthetic, or non-human target embodies the defining operational properties of human beings. Far from a superficial aesthetic judgment, humanlikeness represents a multi-tiered psychological construct encompassing both exterior phenotypic representations and interior cognitive-affective capacities.
Dimension 1: Morphological and Aesthetic Resemblance
This subdimension addresses the external, sensory-driven appraisal of the target entity. It assesses the extent to which the visual appearance, vocal cadence, physical articulation, and aesthetic form replicate anatomical human structures. In digital agents, this involves the presence of facial symmetry, organic ocular movement, lifelike skin texturing, and natural prosody in synthetic speech. In physical robotics, it encompasses humanoid bipedalism, gesturing arms, and facial mimicking mechanics. High ratings on morphological humanlikeness signify that the observer’s visual processing systems process the target using facial recognition and biological motion schemata typically activated exclusively by human conspecifics.
Dimension 2: Cognitive and Behavioral Intentionality (Agency)
Rooted in cognitive science, this facet captures perceived mental capability, problem-solving dexterity, goal-directed behavior, and perceived reasoning. The human mind is distinguished by its capacity to form intentions, hold beliefs, exercise forethought, and execute deliberate decisions. When respondents perceive high cognitive humanlikeness, they credit the object with conscious deliberation, strategic discernment, and self-directed autonomy rather than rote adherence to deterministic computer script. For instance, in an automated banking transaction, a highly cognitive-humanlike entity is perceived as actively “deciding” the outcome based on understanding, rather than merely parsing statistical data.
Dimension 3: Affective and Relational Experience (Patience and Empathy)
The final component centers on the perceived capacity of the non-human object to register emotional states, express empathy, exhibit vulnerability, and engage in social reciprocity. While agency refers to the capacity for “doing,” experience refers to the capacity for “feeling.” In the context of the Garvey et al. (2023) research program, assessing this dimension is pivotal: automated communicators devoid of affective humanlikeness are seen as incapable of subjective malice or joy, making them psychologically distinct targets of interpersonal attribution. Consequently, perceived affective humanlikeness dictates whether an individual anticipates genuine warmth or cold indifference from the artificial interface.
6. Theoretical Framework
The Humanlikeness of the Object scale is grounded in three converging psychological and philosophical frameworks: Anthropomorphism Theory, the Theory of Mind Perception, and Attribution Theory.
Anthropomorphism and the Uncanny Valley
Classic anthropomorphism theory, pioneered by Epley, Waytz, and Cacioppo (2007), posits that humans possess an innate, chronically accessible cognitive tendency to project human-typical characteristics, intentions, and emotional states onto non-human agents. This cognitive process is driven by three key motivations: elicited agent knowledge (using self-knowledge as an inductive base), effectance motivation (the desire to master and predict one’s environment), and sociality motivation (the fundamental need for social connection). Garvey et al. (2023) build upon this foundation by demonstrating that humanlikeness is neither uniformly positive nor universally desired. Rather, according to Mori’s (1970) Uncanny Valley hypothesis and modern expectancy-violation models, high humanlikeness sets steep cognitive expectations. When an entity appears exceedingly humanlike, subtle machine imperfections or sterile social cues generate cognitive dissonance, transforming perceived benevolence into perceived deceit or uncanny revulsion.
The Dual-Axis Theory of Mind Perception
Synthesizing the foundational psychometric work of Gray, Gray, and Wegner (2007), Garvey et al. ground their conceptualization of humanlikeness in the dual dimensions of Agency (the capacity for self-control, planning, and intentional thought) and Experience (the capacity for feeling pain, pleasure, fear, and love). The Humanlikeness of the Object scale captures how consumers mentally locate an artificial intelligence interface within this psychological space. Unlike organic animals, which are often attributed high experience but low agency, traditional machines are perceived as possessing moderate agency but negligible experience. Garvey, Kim, and Duhachek demonstrate that engineering humanlikeness into communication algorithms elevates perceived agency and experience, fundamentally restructuring how message targets process the emotional and relational weight of delivered news.
Attribution Theory and Moral Agency
Heider’s (1958) attribution theory and Weiner’s (1985) locus of causality framework explain how people interpret causal responsibility for events. When bad news (e.g., loan refusal, product unavailability, adverse diagnostic result) is communicated, individuals actively search for a blameworthy locus of causality. If the communicative source exhibits high humanlikeness, the recipient imputes intentionality, perceived malice, or callous disregard, escalating resentment toward the firm. Conversely, when the communicator exhibits low humanlikeness (i.e., is perceived as a purely mechanistic algorithm), the recipient views the news as the product of impartial, unfeeling, systematic computation, attenuating personal hostility and mitigating adverse brand attitudes.
7. Validity
Empirical evaluation of the Humanlikeness of the Object construct across multiple experimental studies in Garvey, Kim, and Duhachek (2023) has yielded extensive psychometric validity evidence.
Construct and Convergent Validity
Construct validity has been evidenced through rigorous structural testing. In multi-sample exploratory and confirmatory factor analyses, scale items intended to index humanlikeness load uniformly onto their theoretical factors with robust factor loadings generally exceeding .75. Convergent validity is evidenced by strong, statistically significant correlations with established standardized batteries of anthropomorphism, such as the Anthropomorphic Tendency Scale (ATS) and the Robotic Social Attributes Scale (RoSAS), with correlation coefficients ranging between r = .62 and r = .78 (p < .001). Furthermore, experimental manipulations of agent characteristics (e.g., presenting a message delivered by a text terminal versus an embodied digital avatar with a biological name) produced pronounced between-condition shifts on the humanlikeness index (F > 85.0, p < .001, partial η² > .30), confirming the instrument’s high experimental sensitivity.
Discriminant Validity
Discriminant validity was established via average variance extracted (AVE) comparisons against conceptually adjacent but functionally distinct consumer constructs, including general technological competence, interface usability (System Usability Scale), and perceived communicative clarity. The AVE for the Humanlikeness latent factor consistently surpassed the squared inter-construct correlations (Fornell-Larcker criterion), demonstrating that measuring humanlikeness does not simply capture a general halo effect or generic positive brand evaluation, but specifically reflects perceived human ontology.
Predictive and Criterion Validity
The scale possesses remarkable predictive validity regarding consumer emotional coping, counterfactual thinking, and behavioral intentions. As demonstrated by Garvey et al. (2023), perceived humanlikeness mediates the interaction between message valence (good news vs. bad news) and organizational evaluation. Specifically, under bad-news conditions, higher scores on humanlikeness uniquely predict increased consumer anger, elevated vindictive word-of-mouth intentions, and heightened attribution of negative intent. Under good-news conditions, humanlikeness reliably predicts positive affective gratitude and organizational loyalty.
8. Reliability
The reliability of the Humanlikeness of the Object measurement instrument has been rigorously demonstrated across diverse experimental contexts involving thousands of participants recruited across university behavioral laboratories and nationwide consumer panels.
Internal Consistency
Internal consistency metrics for the scale consistently exceed rigorous psychometric benchmarks. Across the empirical studies reported by Garvey, Kim, and Duhachek (2023):
- Overall multi-item scale Cronbach’s alpha coefficients regularly fall within the range of α = .88 to α = .95, denoting high inter-item covariance and minimal measurement error.
- Composite reliability (CR) calculated via structural equation modeling demonstrates coefficients consistently above .90, well exceeding the established .70 threshold recommended for psychometric research.
- McDonald’s Omega (ω) coefficients mirror these findings, consistently demonstrating values ranging between .89 and .94, confirming that the scale maintains high internal consistency without suffering from item redundancy.
Test-Retest Stability and Cross-Context Invariance
In longitudinal and cross-scenario replication studies, test-retest reliability across brief latency periods (1 to 2 weeks) yielded intraclass correlation coefficients (ICC) ranging from .78 to .84. Measurement invariance testing across diverse stimulus platforms (e.g., text-only algorithmic chat interfaces, voice-activated smart assistants, realistic photo-rendered humanoid avatars, and physical humanoid robotics) showed configural, metric, and scalar invariance, verifying that the scale items measure the same underlying construct across varying technological modalities.
9. Factor Analysis
The structural dimensionality of the Humanlikeness of the Object scale has been validated using both Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA).
Exploratory Factor Analysis (EFA)
Initial factor structure investigations utilizing principal axis factoring with promax (oblique) rotation revealed an unambiguous, parsimonious factor solution. The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy consistently surpassed .89, and Bartlett’s test of sphericity achieved significance (χ² > 1200, p < .001), indicating the suitability of the correlation matrices for factor extraction. Eigenvalues for the primary factor routinely exceed 3.5, accounting for upwards of 62% to 74% of the total variance across operationalized items. Depending on whether researchers incorporate physical morphology items alongside psychological mind items, EFA either identifies a strong, unified unidimensional general factor (General Perceived Humanlikeness) or a correlated two-factor solution distinguishing Physical/Appearance Humanlikeness from Mental/Behavioral Humanlikeness (inter-factor correlation r ≈ .56 to .68).
Confirmatory Factor Analysis (CFA)
Confirmatory factor analytic routines utilizing maximum likelihood estimation confirm excellent goodness-of-fit indices across empirical models. Standard fit indices meet or exceed contemporary psychometric standards:
- Comparative Fit Index (CFI): .96 to .99 (Threshold: ≥ .95)
- Tucker-Lewis Index (TLI): .95 to .98 (Threshold: ≥ .95)
- Root Mean Square Error of Approximation (RMSEA): .042 to .058, 90% CI [.031, .069] (Threshold: ≤ .06)
- Standardized Root Mean Square Residual (SRMR): .025 to .038 (Threshold: ≤ .08)
- Chi-Square to Degrees of Freedom Ratio (χ²/df): 1.65 to 2.30 (Threshold: ≤ 3.0)
All standardized factor loadings (λ) are statistically significant (p < .001) and range from .74 to .93, indicating that each item reliably reflects the underlying latent construct without severe cross-loadings or structural misspecification.
10. Instrument / Measurement Tool
The operational features and structural parameters of the Humanlikeness of the Object scale are structured as follows:
- Target Construct: Perceived humanlikeness of an artificial intelligence agent, automated algorithm, or non-human entity.
- Test Format: Self-administered psychometric questionnaire; adaptable for paper-and-pencil surveys, laboratory computer terminals, and mobile-optimized online platforms (e.g., Qualtrics, Gorilla Experiment Builder).
- Item Formats: Multi-item battery utilizing semantic differential bipolar anchors and single-stimulus Likert-type response scales.
- Response Anchors: Typically operationalized using a 7-point Likert scale (ranging from 1 = “Not at all humanlike / Strongly disagree” to 7 = “Extremely humanlike / Strongly agree”) or 7-point semantic differential scales (e.g., Machine-like vs. Human-like, Artificial vs. Natural, Inanimate vs. Animate).
- Administration Time: Approximately 1 to 3 minutes, making it highly efficient for laboratory experiments, longitudinal tracking, and field surveys.
- Scoring Procedure:
- Negatively keyed items (e.g., mechanistic, algorithmic anchors) are reverse-coded prior to aggregation.
- Subscale scores are generated by averaging the observed scores within the respective morphological and behavioral domains.
- An overall Composite Humanlikeness Index is computed by averaging all items; higher total mean scores signify greater perceived anthropomorphism and psychological resemblance to a human communicator.
11. Permissions & Fee and Test Year
The scale was developed and published in 2023 in the Journal of Marketing:
- Test Year: 2023.
- Publishing Entity: American Marketing Association (AMA) / SAGE Publications.
- Licensing and Academic Use: The scale was established as an academic measurement tool for scholarly and empirical inquiry. Qualified academic researchers may implement, adapt, and administer the instrument for non-commercial educational and empirical investigations under fair use principles, provided that proper scholarly attribution is formally accorded to the authors (Garvey, Kim, & Duhachek, 2023).
- Commercial Applications: Commercial deployments, proprietary software integration, or inclusion in commercial testing batteries may require formal copyright clearance and written licensing agreements from the American Marketing Association or the respective copyright holders.
- Fee: There are no per-administration fees for non-profit academic research, though access to the original archival journal article may require an institutional subscription or individual article purchase through the publisher’s platform.
12. References
- 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
- Garvey, A. M., Kim, T., & Duhachek, A. (2023). Bad news? Send an AI. Good news? Send a human. Journal of Marketing, 87(1), 10–25. https://doi.org/10.1177/00222429211066991
- Gray, H. M., Gray, K., & Wegner, D. M. (2007). Dimensions of mind perception. Science, 315(5812), 619–619. https://doi.org/10.1126/science.1134475
- Heider, F. (1958). The Psychology of Interpersonal Relations. John Wiley & Sons. https://doi.org/10.1037/10628-000
- Mori, M. (1970). The uncanny valley. Energy, 7(4), 33–35.
- Weiner, B. (1985). An attributional theory of achievement motivation and emotion. Psychological Review, 92(4), 548–573. https://doi.org/10.1037/0033-295X.92.4.548