Advice-Giver Warmth

1. Abstract

The Advice-Giver Warmth scale is a specialized psychometric instrument designed to evaluate the degree to which an individual perceives a source of guidance, recommendations, or consultation as warm, approachable, benevolent, and interpersonally engaged. Adapted and validated in contemporary consumer psychology and human-computer interaction research by von Walter, Wentzel, and Raff (2023), the scale captures consumers’ and decision-makers’ evaluations of interpersonal warmth across both human and artificial intelligence (AI) or algorithmic advice systems. Grounded conceptually in the Stereotype Content Model (SCM; Fiske, Cuddy, Glick, & Xu, 2002), the scale operationalizes the primary social judgment dimension of perceived positive intent and friendliness. The measure comprises 4 items assessed on a 7-point Likert scale ranging from 1 (“Strongly disagree”) to 7 (“Strongly agree”). Items assess key attributes of perceived warmth: being warm, friendly, helpful, and well-intentioned. Psychometric evaluations across multiple empirical investigations demonstrate high internal consistency (Cronbach’s $\alpha ge .90$; Composite Reliability $> .90$), strong factor determinacy, robust convergent validity, and clear discriminant validity against the orthogonal dimension of advice-giver competence. In predictive models, perceived warmth serves as a critical psychological mediator that explains variance in consumers’ trust, advice adherence, and service relationship resilience when transitioning from interpersonal service interactions to automated or algorithmic advisory platforms.

2. Keywords

Advice-Giver Warmth, Stereotype Content Model, Algorithmic Advice, Social Perception, Human-Computer Interaction, Trust in AI, Advice Utilization, Perceived Benevolence, Consumer Behavior, Psychometrics

3. Authors

The scale was adapted and validated in the context of service relationships and algorithmic versus human advice systems by the following researchers:

• Benjamin von Walter — Professor of Marketing, School of Management, OST – Eastern Switzerland University of Applied Sciences, St. Gallen, Switzerland.
• Daniel Wentzel — Professor and Chair of Marketing, School of Business and Economics, RWTH Aachen University, Aachen, Germany.
• Stefan Raff — Assistant Professor of Technology and Marketing, University of St. Gallen, St. Gallen, Switzerland.

4. Purpose

The overarching purpose of the Advice-Giver Warmth scale is to quantify an individual’s subjective appraisal of an advisor’s benevolent intentions, emotional accessibility, and prosocial orientation. Advice-seeking is fundamentally a vulnerable behavioral act: the advice recipient acknowledges an informational deficit, decision ambiguity, or analytical limitation, exposing themselves to the potential opportunism, indifference, or misalignment of the advice source. Consequently, individuals rapidly screen prospective advisors not only for their cognitive capability, but crucially for their perceived sociability and morality.

While the psychometric measurement of interpersonal warmth has a rich history within social psychology, the modern proliferation of automated decision support systems, robo-advisors, generative AI agents, and algorithmic intermediaries created an urgent need for an instrument sensitive to both human and non-human sources of advice. Von Walter, Wentzel, and Raff (2023) operationalized this instrument to investigate how consumers navigate the introduction of algorithmic advice within existing service relationships. In retail, financial consulting, healthcare triage, and digital commerce, organizations increasingly substitute or complement human advisors with computational models. The scale provides researchers and practitioners with an empirical tool to test whether advice recipients perceive an advice source as possessing genuine benevolence or whether the advisor is viewed as cold, transactional, and detached.

In clinical, organizational, and behavioral research settings, the scale addresses several specific applied challenges:

• Explaining “Algorithm Aversion” vs. “Algorithm Appreciation”: While algorithms are frequently judged as analytically competent or superior in calculating statistical probabilities, decision-makers often resist them due to perceived deficits in empathy, relational caring, and personalized intent. Measuring advice-giver warmth isolates the precise emotional and moral deficit driving advice discounting.
• Service Relationship Continuity: The scale enables service firms to diagnose whether migrating clients from frontline human advisors to algorithmic dashboards damages the relational capital, perceived loyalty, and emotional rapport built over years of interpersonal exchange.
• Advisory Compliance and Adherence: Across domains such as medical regimen compliance, wealth management, and strategic managerial consulting, advice acceptance is rarely a purely rational calculation. Perceived warmth fosters psychological safety and subjective comfort, dramatically increasing the probability that an individual will act upon recommended actions.

5. Psychological Construct

The construct of Advice-Giver Warmth represents a focal pillar of human social cognition adapted to consultative interactions. According to universal models of social perception, when humans encounter a novel social actor or an unfamiliar advisory agent, they immediately and instinctively seek answers to two fundamental survival questions: “What are this entity’s intentions toward me?” (Warmth) and “Does this entity have the capacity to enact those intentions?” (Competence). Advice-giver warmth directly assesses the first question.

Within this measurement tool, warmth is operationalized as a composite of four interrelated experiential and dispositional facets:

• Affective Warmth: The basic emotional temperature of the interaction (reflected directly in the item “The advice giver is warm”). It reflects the degree of perceived tenderness, human connection, and psychological comfort radiated by the source. In automated systems, affective warmth is communicated through conversational tone, responsive dialogue pacing, and anthropomorphic visual cues.
• Interpersonal Friendliness: Evaluated via the item “The advice giver is friendly”, this facet gauges perceived approachability, sociability, and pleasantness. A friendly advisor minimizes social distance, reduces status asymmetries, and mitigates the anxiety inherent in complex decision-making contexts.
• Prosocial Helpfulness: Captured by “The advice giver is helpful”, this attribute bridges pure affect with functional cooperation. In an advice relationship, helpfulness reflects an active, outward orientation toward assisting the recipient in resolving their dilemma, signaling that the advisor is willing to invest cognitive or computational resources for the user’s benefit.
• Benevolent Intentionality: Measured by “The advice giver is well-intentioned”, this moral and ethical dimension forms the core of interpersonal trust. Benevolence implies that the advisor prioritizes the recipient’s best interests over self-serving, commercial, or opportunistic goals. In algorithmic settings, perceived intentionality reflects the user’s belief that the underlying algorithm was coded to optimize client outcomes rather than to maximize corporate extraction.

Crucially, within psychometric theory, advice-giver warmth is conceptualized as an emergent perception resulting from the interaction between the advisor’s observable behavioral cues (e.g., tone of voice, linguistic sentiment, interface responsiveness, explanatory depth) and the recipient’s internal cognitive schema. It operates orthogonally to perceived competence: an advisor can be perceived as exceptionally capable, analytically rigorous, and statistically infallible while simultaneously being rated as entirely devoid of warmth.

6. Theoretical Framework

The theoretical architecture underpinning the Advice-Giver Warmth scale synthesizes three prominent frameworks from social psychology, cognitive science, and human-technology interaction:

The Stereotype Content Model (SCM)

Developed extensively by Susan Fiske, Amy Cuddy, and Peter Glick (2002, 2008), the Stereotype Content Model posits that interpersonal, intergroup, and brand impressions are universally structured along two distinct dimensions: warmth and competence. Evolutionary psychology suggests that the warmth dimension carries primacy; distinguishing friend from foe (determining intent) was far more consequential for ancestral survival than determining an actor’s skill level. In advice environments, recipients rapidly scan an advisor’s profile, demeanor, or conversational outputs for cues indicating moral integrity, communal values, and empathy. The 4-item scale directly extracts the warmth axis of the SCM, tailoring its semantic anchors to an advisory context.

Mind Perception Theory

In evaluating non-human or algorithmic advisors, warmth evaluations interface with Mind Perception Theory (Gray, Gray, & Wegner, 2007). Gray and colleagues demonstrated that humans perceive minds along two dimensions: Agency (the capacity for planning, thinking, reasoning, and self-control) and Experience (the capacity to feel pain, pleasure, warmth, joy, and emotional connection). While people readily attribute high agency to algorithms, machines, and analytical software, they systematically deny them the capacity for experience. Consequently, algorithmic advice givers face an inherent baseline deficit in perceived warmth. When service providers deploy AI advisors, they frequently trigger consumers’ cognitive expectation of coldness, demanding robust psychometric instruments to track whether design interventions (such as natural language processing or visual avatars) can bridge the experiential gap.

The Computers Are Social Actors (CASA) Paradigm

The Computers Are Social Actors (CASA) paradigm, pioneered by Byron Reeves and Clifford Nass (1996), proves that human beings mindlessly apply fundamental interpersonal social rules, norms, and evaluations to computer technologies, even when consciously aware that machines lack biological life and subjective consciousness. When an algorithmic system provides advice, users automatically process subtle communicative signals (e.g., politeness, reciprocal disclosure, personalized phrasing) as indicators of social warmth. The Advice-Giver Warmth scale leverages this social-psychological reality, allowing direct comparative measurement between human advisors and digital artifacts on identical semantic dimensions.

7. Validity

Empirical validation studies for the Advice-Giver Warmth scale confirm exceptional psychometric performance across multiple validity criteria:

Content and Face Validity

The scale possesses high content and face validity, derived from the established operational definitions of warmth, sociability, and benevolence in the social cognition literature (Cuddy et al., 2008; Fiske et al., 2002). The four items comprehensively represent affective orientation (warm), interpersonal stance (friendly), functional prosocial behavior (helpful), and underlying ethical motivation (well-intentioned). Expert panels in consumer research and psychometrics have repeatedly confirmed that these items concisely encapsulate the warmth construct without introducing semantic redundancy or construct contamination.

Convergent Validity

Convergent validity has been established through strong item factor loadings and average variance extracted (AVE) statistics. In the structural equations tested by von Walter et al. (2023), standardized factor loadings for all four items consistently exceed $.80$, frequently reaching between $.85$ and $.94$. The Average Variance Extracted (AVE) values consistently surpass the recommended $.50$ threshold, typically ranging between $.72$ and $.84$, confirming that the latent warmth construct captures far more variance from its indicators than is attributable to measurement error.

Discriminant Validity

Discriminant validity is critical, particularly given the historical pairing of warmth and competence. Using the Fornell-Larcker criterion, the square root of the AVE for the Advice-Giver Warmth scale is substantially greater than its bivariate correlation with related constructs, including perceived competence (typically measured by items such as competent, capable, intelligent, and skillful), service quality, and technological novelty. Furthermore, modern assessments using the Heterotrait-Monotrait (HTMT) ratio of correlations yield values well below the conservative $.85$ threshold (often between $.35$ and $.60$), confirming that warmth and competence remain psychometrically distinct latent variables in advisory contexts.

Predictive and Nomological Validity

Nomological and predictive validity have been documented through experimental and field settings. In the investigations conducted by von Walter et al. (2023), perceived advice-giver warmth served as a key psychological mediator. Specifically, when existing customers encountered algorithmic rather than human advisors, lower perceived warmth directly accounted for a decline in overall relationship satisfaction and lower acceptance of the recommended advice. Furthermore, when algorithmic systems were engineered to provide explanations or anthropomorphic cues, increases in warmth scores statistically mediated enhanced consumer trust and willingness to adopt the advice.

8. Reliability

The Advice-Giver Warmth scale exhibits high internal consistency reliability across varied empirical samples and experimental conditions. In psychometric evaluations testing consumer responses to financial, legal, and retail advice, the scale has repeatedly surpassed established standards for instrument reliability:

• Cronbach’s Alpha ($\alpha$): Across experimental studies examining human versus algorithmic advisors, Cronbach’s alpha coefficients for the 4-item scale consistently range between $\alpha = .90$ and $\alpha = .95$, demonstrating exceptional internal consistency without evidence of destructive item redundancy.
• Composite Reliability (CR): Structural equation modeling confirms that the composite reliability of the latent construct regularly exceeds $.91$ (often reaching $.94$), significantly outperforming the standard $.70$ benchmark proposed for social science research.
• Split-Half and Inter-Item Correlations: Inter-item correlations among the four items consistently range between $.68$ and $.85$, indicating that each individual item contributes meaningful, non-trivial variance to the common latent core. Corrected item-total correlations uniformly exceed $.75$.
• Test-Retest Stability: In longitudinal experimental paradigms with repeated measurement intervals, the scale displays high stability across static stimuli ($r_{tt} > .80$). When advice conditions change (e.g., transitioning from an automated script to an interactive agent), the scale demonstrates immediate sensitivity to experimental manipulation, confirming that it accurately reflects dynamic shifts in situational social perception.

9. Factor Analysis

The factorial validity of the Advice-Giver Warmth scale has been thoroughly confirmed using both Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA).

Exploratory Factor Analysis (EFA)

When subjected to EFA (using Principal Axis Factoring or Maximum Likelihood estimation with Promax or Varimax rotation) alongside items measuring advisor competence, system usability, and consumer trust, the four warmth items load cleanly onto a single, dominant factor. The initial eigenvalue for the warmth dimension typically exceeds $3.0$, accounting for over $75%$ of the total variance among the items. Cross-loadings on adjacent factors (such as competence or technical capability) are minimal, routinely remaining below $.20$.

Confirmatory Factor Analysis (CFA)

Confirmatory factor analytic models demonstrate that a single-factor, unidimensional structure provides an excellent fit to empirical data. In multi-group CFA models comparing human advisors directly against algorithmic advisors, the measurement model exhibits strong metric and scalar invariance, proving that respondents interpret the construct identically across both human and automated advice contexts.

Representative goodness-of-fit indices from structural evaluations include:

• Chi-Square / Degrees of Freedom ($\chi^2/df$): Values regularly fall between $1.0$ and $2.5$, indicating minimal model misspecification.
• Comparative Fit Index (CFI): Consistently reported between $.98$ and $1.00$.
• Tucker-Lewis Index (TLI): Consistently reported between $.97$ and $.99$.
• Root Mean Square Error of Approximation (RMSEA): Estimates range from $.02$ to $.06$, with the $90%$ confidence interval safely below the conventional $.08$ cutoff.
• Standardized Root Mean Square Residual (SRMR): Typically under $.03$, indicating outstanding residual covariance reproduction.

Individual standardized factor loadings ($lambda$) are uniformly elevated and statistically significant at $p < .001$:

• “The advice giver is warm”: $\lambda \approx .86 – .93$
• “The advice giver is friendly”: $\lambda \approx .84 – .92$
• “The advice giver is helpful”: $\lambda \approx .81 – .89$
• “The advice giver is well-intentioned”: $\lambda \approx .83 – .90$

10. Instrument / Measurement Tool

The operational specifications of the scale are structured as follows:

• Test Type: Psychometric self-report rating scale / perceptual evaluation inventory.
• Target Population: Consumers, decision-makers, clients, or experimental participants receiving guidance, advice, or recommendations from a human consultant, artificial intelligence, or automated algorithmic system.
• Administration Format: Computer-assisted web interviewing (CAWI), paper-and-pencil questionnaires, or embedded real-time software feedback modules.
• Item Count: 4 items.
• Authentic Response Scale: 7-point Likert scale (1 = Strongly disagree, 7 = Strongly agree).
• Scoring Rules: All four items are phrased positively in the direction of the construct. There are no reverse-scored items. An overall perceived advice-giver warmth index is calculated by computing the arithmetic mean across all four items:
  
  $$\text{Advice-Giver Warmth} = \frac{\text{Item}_1 + \text{Item}_2 + \text{Item}_3 + \text{Item}_4}{4}$$
  
  Scores range from $1.00$ to $7.00$, with higher scores indicating higher levels of perceived warmth, friendliness, helpfulness, and benevolent intentions.
• Completion Time: Approximately 30 to 60 seconds, minimizing respondent fatigue and making it ideal for integration into larger experimental batteries.

11. Permissions & Fee and Test Year

The Advice-Giver Warmth scale was adapted and published in its current form in 2023 in the Journal of Retailing by Benjamin von Walter, Daniel Wentzel, and Stefan Raff. As an instrument published within a peer-reviewed academic journal, the scale items are available for non-commercial academic research, educational use, and scientific inquiry under standard academic fair-use conventions, provided that appropriate bibliographic citation is accorded to the original authors and journal.

Commercial practitioners, proprietary software developers, or corporate consulting firms intending to embed the scale within commercial algorithmic platforms, customer relationship software, or proprietary benchmarking tools should review the copyright policies of Elsevier and the Journal of Retailing, or contact the primary authors regarding commercial licensing agreements.

12. References

Below are primary references and theoretical foundations relevant to the Advice-Giver Warmth scale:

• Cuddy, A. J., Fiske, S. T., & Glick, P. (2008). Warmth and competence as universal dimensions of social perception: The Stereotype Content Model and the BIAS map. Advances in Experimental Social Psychology, 40, 61-149. https://doi.org/10.1016/S0065-2601(07)00002-0
• Dietvorst, B. J., Simmons, J. P., & Massey, C. (2018). Overcoming algorithm aversion: People will use imperfect algorithms if they can (even slightly) modify them. Management Science, 64(3), 1155-1170. https://doi.org/10.1287/mnsc.2016.2643
• Fiske, S. T., Cuddy, A. J., Glick, P., & Xu, J. (2002). A model of (often mixed) stereotype content: Competence and warmth respectively follow from perceived status and competition. Journal of Personality and Social Psychology, 82(6), 878-902. https://doi.org/10.1037/0022-3514.82.6.878
• 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
• Reeves, B., & Nass, C. (1996). The media equation: How people treat computers, television, and new media like real people and places. Cambridge University Press.
• von Walter, B., Wentzel, D., & Raff, S. (2023). Should service firms introduce algorithmic advice to their existing customers? The moderating effect of service relationships. Journal of Retailing, 99(2), 280-296. https://doi.org/10.1016/j.jretai.2023.04.001

13. Items of the Scale