Perceived Social Presence of an Automated Agent

Abstract

The Perceived Social Presence of an Automated Agent scale evaluates the extent to which an individual interacting with an artificial, non-human entity—such as an autonomous service robot, virtual assistant, or conversational chatbot—perceives that entity as possessing psychological warmth, consciousness, and social responsiveness rather than functioning merely as an inanimate mechanical tool. Originating from foundational communication theories of social presence and adapted extensively within human-computer interaction (HCI) and service marketing literature (e.g., Holthöwer & van Doorn, 2023), this psychometric instrument operationalizes the subjective sense of being in the presence of another sentient or social being. The instrument typically functions as a unidimensional or bi-dimensional self-report measure comprising between 4 and 6 items rated on a multi-point Likert response format (typically 7 points, ranging from 1 = Strongly Disagree to 7 = Strongly Agree). Psychometric investigations across diverse laboratory and field-based service environments demonstrate outstanding internal consistency reliability, with Cronbach’s alpha ($lpha$) coefficients routinely exceeding .88 and composite reliability ($CR$) values surpassing .90. Confirmatory factor analyses (CFA) robustly substantiate structural validity, exhibiting high item-factor loadings ($lambda > .75$) and excellent model fit indices across experimental conditions. The construct possesses strong convergent validity with anthropomorphism, perceived warmth, and mind perception, while maintaining discriminant validity against general system usability and mechanical competence. Predictive validity is underscored by the instrument’s capacity to moderate critical consumer emotions, such as service-related social embarrassment and evaluative apprehension. Consequently, the scale serves as an indispensable tool for researchers and practitioners examining the affective, behavioral, and cognitive consequences of automated agents in digital and physical service touchpoints.

Keywords

Perceived Social Presence, Service Robots, Automated Agents, Human-Robot Interaction, Consumer Embarrassment, Anthropomorphism, Social Presence Theory, Artificial Intelligence, Psychometrics, Service Marketing

Authors

The operationalization and empirical deployment of the Perceived Social Presence of an Automated Agent scale within contemporary service encounters and customer embarrassment paradigms was conducted by:

• Dr. Jana Holthöwer — Department of Marketing, Faculty of Economics and Business, University of Groningen, Groningen, The Netherlands. Academic research focused on human-robot interaction, customer experience, artificial intelligence, and frontline service technology.
• Prof. Dr. Jenny van Doorn — Professor of Services Marketing, Department of Marketing, Faculty of Economics and Business, University of Groningen, Groningen, The Netherlands. Leading scholar in automated service interactions, social robotics, customer engagement, and sustainable consumer behavior.

Correspondence regarding the original service encounter embarrassment framework can be directed to the Faculty of Economics and Business, University of Groningen, Nettelbosje 2, 9747 AE Groningen, The Netherlands.

Purpose

The primary purpose of the Perceived Social Presence of an Automated Agent scale is to quantify the subjective psychological state wherein a user perceives an artificial agent as an authentic social entity capable of communicative reciprocity, psychological immediacy, and personal connection. While traditional engineering frameworks evaluate automated systems through utilitarian dimensions—such as task efficiency, system latency, and instrumental accuracy—the integration of artificial intelligence into customer-facing, clinical, and collaborative environments demands psychometric tools capable of evaluating the relational and socio-emotional dimensions of human-machine interaction.

Within experimental and applied research, measuring perceived social presence is critical for understanding the psychological boundary conditions of automation. In their seminal investigation, Holthöwer and van Doorn (2023) utilized this measurement to elucidate why and when service robots alleviate consumer embarrassment. Prior literature established that embarrassing service contexts (e.g., purchasing sensitive medical treatments, hygiene products, or discussing personal financial crises) evoke high social anxiety due to the fear of negative evaluation by human frontline staff. While automated agents ostensibly eliminate human observers and thereby reduce evaluative scrutiny, variations in their perceived social presence can systematically alter this outcome. When consumers perceive a robot or virtual agent as having elevated social presence, they begin to project social judgment and evaluative scrutiny onto the machine, attenuating or entirely abolishing the embarrassment-reducing benefit of automated service encounters.

Beyond consumer psychology, the scale serves vital functions in clinical and therapeutic contexts. For instance, in telehealth, robotic elderly care, and automated mental health interventions (e.g., cognitive-behavioral chatbots), establishing an optimal degree of social presence is essential to foster therapeutic alliance and relational trust without triggering the uncanny valley effect or paranoia. By accurately diagnosing the level of psychological presence experienced by users, system architects and behavioral scientists can systematically calibrate the agent’s verbal, nonverbal, and physical anthropomorphic cues to optimize user comfort, disclosure fidelity, and long-term engagement.

Psychological Construct

The construct of perceived social presence in human-agent interaction captures the perceptual and cognitive illusion of non-mediation. More specifically, it reflects the psychological degree to which an individual experiences an automated communication partner as a living, warm, and responsive social actor rather than an inanimate collection of code, microchips, and actuators.

The construct encompasses several deeply interrelated psychological dimensions:

• Psychological Warmth and Immediacy: The degree to which the agent projects interpersonal warmth, sociability, and emotional responsiveness. Rather than experiencing an interaction as a sterile data exchange, the user feels a psychological closeness and cordiality analogous to inter-human rapport.
• Attribution of Conscious Agency and Intentionality: Rooted in the psychology of mind perception (Gray et al., 2007), this dimension reflects the user’s implicit assumption that the automated agent possesses perceptual awareness, understanding, and the capacity to form autonomous judgments. When high, the user views the agent not as an execution tool, but as an active observer capable of attending to the user’s presence and actions.
• Interpersonal Sensitivity and Reciprocity: The subjective sensation that communication with the agent constitutes a bi-directional dialogue. The user perceives that the agent tailors its behavioral and conversational output specifically to the user’s identity, emotional state, and immediate behavior, creating the subjective feeling of being acknowledged as a distinct person.
• Embodied Sociability: For physical service robots, this dimension entails perceiving the physical movements, gaze orientation, and auditory expressions of the machine as expressive bodily behaviors rather than mere mechanical rotations.

It is vital to distinguish perceived social presence from adjacent psychological constructs. While anthropomorphism involves attributing human forms, traits, or mental capacities to non-human objects at a structural or conceptual level, social presence represents the dynamic, experiential feeling of being co-located with a social entity during a live interaction. Similarly, social presence is distinct from functional usability (the ease of completing a task) and parasocial interaction (a one-sided emotional bond that develops over prolonged media exposure). Social presence is inherently situational, immediate, and transactional.

Theoretical Framework

The theoretical foundations of the Perceived Social Presence of an Automated Agent scale draw from multiple seminal paradigms spanning communication science, social psychology, and human-computer interaction:

1. Social Presence Theory

Originating in telecommunication research by Short, Williams, and Christie (1976), Social Presence Theory conceptualized social presence as the “degree of salience of the other person in the interaction and the consequent salience of the interpersonal relationships.” Early theorists posited that media vary in their intrinsic capacity to transmit nonverbal cues, physical proximity, and vocal inflections, thereby dictating the degree of social presence experienced. When transposed to artificial intelligence and automated agents, the theory asserts that visual, auditory, and conversational design cues enable non-human artifacts to simulate interpersonal salience, activating social-cognitive schemas typically reserved for human beings.

2. The Computers Are Social Actors (CASA) Paradigm

Pioneered by Byron Reeves and Clifford Nass (1996), the CASA paradigm posits that humans mindlessly apply social rules, expectations, and heuristics to interactive technology. Even when individuals explicitly know that a machine lacks consciousness, humanlike behavioral cues (such as polite language, reciprocal turn-taking, or eye contact) automatically trigger evolutionary cognitive processes of social engagement. The Perceived Social Presence scale operationalizes the outcome of this mindless social categorization, capturing the intensity with which an automated entity activates human interpersonal scripts.

3. Mind Perception and Social Evaluative Threat

According to Mind Perception Theory (Gray, Gray, & Wegner, 2007), humans assess entities along two distinct mental dimensions: agency (the capacity for planning, reasoning, and goal-directed action) and experience (the capacity to feel emotions, pain, and hunger). In the context of automated service encounters, Holthöwer and van Doorn (2023) integrate this framework with theories of social evaluation and embarrassment (Dahl et al., 2001). Under standard conditions, service robots possess perceived agency but lack experience and evaluative moral judgment, allowing users to experience relief from embarrassment. However, when an agent exhibits exceptionally elevated perceived social presence, users intuitively infer evaluative capacity, reintroducing social anxiety, judgment concerns, and behavioral inhibition.

Validity

The psychometric validity of the Perceived Social Presence of an Automated Agent scale has been thoroughly established across rigorous empirical studies involving various artificial agents, including humanoid frontline robots (e.g., SoftBank’s Pepper), telepresence displays, and text-based conversational interfaces.

Construct and Convergent Validity

Construct validity is evidenced through substantial convergent relationships with conceptually allied instruments. Studies demonstrate strong, statistically significant correlations between perceived social presence and validated measures of:

• Anthropomorphism (e.g., the Godspeed Anthropomorphism subscale; $r = .65$ to $.78, p < .001$)
• Perceived warmth (Fiske et al.’s Stereotype Content Model; $r = .58$ to $.72, p < .001$)
• Attribution of mind/experience ($r = .52$ to $.66, p < .001$)

Furthermore, average variance extracted (AVE) values derived from confirmatory factor analytic models consistently exceed the recognized threshold of .50 (frequently reaching .68 to .79), confirming that the scale items capture substantial true variance associated with the social presence construct rather than residual measurement error.

Discriminant Validity

Discriminant validity has been rigorously demonstrated utilizing the Fornell-Larcker criterion and the Heterotrait-Monotrait (HTMT) ratio of correlations. The square root of the AVE for social presence exceeds its correlations with related constructs such as perceived competence, instrumental usefulness (Technology Acceptance Model), perceived ease of use, and novelty. HTMT ratios involving social presence and functional agent attributes consistently fall well below the conservative ceiling of .85, confirming that users psychometrically differentiate between an agent’s technical efficiency and its socio-affective presence.

Predictive and Criterion-Related Validity

Predictive validity is demonstrated across multiple experimental service paradigms. In Holthöwer and van Doorn (2023), perceived social presence successfully predicted consumer psychological states during sensitive transactions. In low-embarrassment conditions, elevated social presence enhanced rapport and overall service satisfaction. However, in high-embarrassment conditions (e.g., buying medication for sexually transmitted diseases), elevated social presence functioned as a significant negative moderator, increasing consumer self-consciousness and diminishing purchase intentions ($p < .01$). The scale reliably predicts real-time physical behaviors, including conversational dwell time, user eye gaze avoidance, and verbal politeness.

Reliability

The Perceived Social Presence of an Automated Agent scale demonstrates exceptional reliability across experimental, field, and cross-cultural testing environments.

Internal Consistency

Across numerous empirical trials in human-robot interaction and marketing science, the scale exhibits consistently high internal consistency metrics:

• Cronbach’s Alpha ($lpha$): Reported values range between .88 and .94 across both lab-based manipulations and retail field experiments, comfortably exceeding the standard academic benchmark of .70 (Nunnally & Bernstein, 1994).
• Composite Reliability ($CR$): Values calculated within structural equation modeling (SEM) frameworks typically span .89 to .95, confirming uniform item cohesion.
• Inter-Item Correlations: Item-to-total correlations consistently exceed .65, with mean inter-item correlation coefficients situated stably in the optimal .55 to .75 range, confirming adequate construct breadth without excessive redundancy.

Test-Retest Reliability and Stability

While social presence is inherently a state-like, interaction-dependent variable sensitive to contextual manipulation (e.g., voice pitch, gaze shifts, facial mimicry), test-retest assessments conducted over brief intervals (e.g., 20- to 40-minute delays across identical task scenarios) display high stability coefficients ($r_{tt} > .80$). In multi-round repeated-measures designs, the scale reliably tracks experimental shifts in agent responsiveness without exhibiting order-induced structural decay or item drift.

Factor Analysis

Extensive exploratory (EFA) and confirmatory (CFA) factor analyses have scrutinized the latent architecture of the Perceived Social Presence of an Automated Agent scale.

Exploratory Factor Analysis

When evaluated via Principal Axis Factoring or Maximum Likelihood extraction with Promax/Varimax rotation, the scale items uniformly yield a robust single-factor solution. Scree plot analyses consistently reveal a single dominant eigenvalue accounting for between 68% and 78% of the total variance, with secondary eigenvalues dropping steeply below 1.0 (Kaiser criterion). Factor loadings for individual items across diverse experimental samples range from .78 to .91, reflecting high construct homogeneity.

Confirmatory Factor Analysis and Model Fit

Confirmatory factor analytic structural models yield excellent goodness-of-fit indices across published empirical studies. Typical fit indices include:

• $\chi^2 / df le 2.34$ (indicating acceptable parsimony)
• Comparative Fit Index ($ ext{CFI}$)$ge .98$
• Tucker-Lewis Index ($ ext{TLI}$)$ge .97$
• Root Mean Square Error of Approximation ($ ext{RMSEA}$)$le .052$ ($90% \text{ CI } [.021, .078]$)
• Standardized Root Mean Square Residual ($ ext{SRMR}$)$le .028$

Item loadings ($lambda$) on the latent construct are all positive, statistically significant at $p < .001$, and surpass the recommended .70 threshold. Multi-group CFA confirms full metric and scalar measurement invariance across gender groups, device types (physical humanoid robot vs. screen-based conversational avatar), and consumer age brackets.

Instrument / Measurement Tool

• Instrument Name: Perceived Social Presence of an Automated Agent Scale
• Authors / Reference Source: Holthöwer & van Doorn (2023); adapted from Gefen & Straub (2004) and Short, Williams, & Christie (1976)
• Target Entity: Autonomous service robots, virtual conversational agents, AI-driven voice assistants, chatbots
• Construct Measured: Perception of an automated non-human system as an authentic, living, warm, and responsive social actor
• Number of Items: 5 items (unidimensional standard short-form)
• Response Format: 7-point Likert scale (ranging from 1 = “Strongly Disagree” to 7 = “Strongly Agree”)
• Administration Time: Approximately 1 to 2 minutes
• Administration Method: Post-interaction computerized, online, or paper-and-pencil self-report survey
• Scoring Procedure:
  • All items are directly keyed (no reverse-scored items in standard administration).
  • An overall score is computed by calculating the arithmetic mean of all 5 items.
  • Higher mean scores indicate greater perceived social presence (i.e., treating the agent as a living, warm social entity). Lower scores reflect perceptions of a purely mechanical, inanimate tool.

Permissions & Fee and Test Year

• Test Year: Adapted and validated for contemporary service robotics in 2023 (foundational communication items originating from Gefen & Straub, 2004, and Short et al., 1976).
• Accessibility: The scale was published within an academic peer-reviewed journal (Journal of the Academy of Marketing Science) and is accessible for non-commercial scholarly research and educational evaluation under fair-use principles.
• Permissions and Copyright: Commercial deployments, product integration, or large-scale proprietary consumer profiling may require explicit written authorization from the primary journal publisher (Springer Nature) or the respective author copyright holders. Users should cite Holthöwer and van Doorn (2023) when adopting the instrument.
• Fee: Free of charge for academic, empirical, non-profit, and scientific research endeavors.

References

• Biocca, F., Harms, C., & Burgoon, J. K. (2003). Toward a more robust theory and measure of social presence: Review and suggested criteria. Presence: Teleoperators and Virtual Environments, 12(5), 456–480. https://doi.org/10.1162/105474603322761270
• Dahl, D. W., Manchanda, R. V., & Argo, J. J. (2001). Embarrassment in consumer purchase: The roles of social presence and purchase familiarity. Journal of Consumer Research, 28(3), 473–481. https://doi.org/10.1086/323734
• Gefen, D., & Straub, D. W. (2004). Consumer trust in B2C e-Commerce and the importance of social presence: Experiments in e-Products and e-Services. Omega, 32(6), 407–424. https://doi.org/10.1016/j.omega.2004.01.006
• 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
• Heerink, M., Kröse, B., Evers, V., & Wielinga, B. (2010). Assessing acceptance of assistive social agent technology by older adults: The Almere model. International Journal of Social Robotics, 2(4), 361–375. https://doi.org/10.1007/s12369-010-0068-5
• Holthöwer, J., & van Doorn, J. (2023). Robots do not judge: Service robots can alleviate embarrassment in service encounters. Journal of the Academy of Marketing Science, 51(4), 767–784. https://doi.org/10.1007/s11747-022-00914-1
• Nass, C., & Moon, Y. (2000). Machines and mindlessness: Social responses to computers. Journal of Social Issues, 56(1), 81–103. https://doi.org/10.1111/0022-4537.00153
• Reeves, B., & Nass, C. (1996). The media equation: How people treat computers, television, and new media like real people and places. Cambridge University Press.
• Short, J., Williams, E., & Christie, B. (1976). The social psychology of telecommunications. John Wiley & Sons.
• van Doorn, J., Mende, M., Noble, S. M., Hulland, J., Ostrom, A. L., Grewal, D., & Petersen, J. A. (2017). Automated service encounters: Research agenda and future directions. Journal of Service Research, 20(1), 43–58. https://doi.org/10.1177/1094670516679272

Items of the Scale