Adoption of Artificial Intelligence-Based Employee Experience Chatbots–Model

1. Abstract

The Adoption of Artificial Intelligence-Based Employee Experience Chatbots–Model is an advanced psychometric and structural assessment tool developed by Pillai, Ghanghorkar, Sivathanu, Algharabat, and Rana (2024) to evaluate the cognitive, affective, and contextual determinants influencing employees’ behavioral intentions to adopt artificial intelligence (AI)-driven conversational agents within workplace environments. Grounded in Behavioral Reasoning Theory (BRT), the instrument captures the nuanced psychological interplay between values, context-specific reasons (both “reasons for” and “reasons against”), attitudes, and ultimate adoption intentions regarding employee experience (EEX) chatbots. Comprising 43 items structured across 11 discrete latent constructs, the measurement scale is administered using a standardized five-point Likert scale ranging from 1 (“strongly disagree”) to 5 (“strongly agree”).

The eleven constructs operationalized in the instrument include: Personalization (PES), Interactivity (INT), Perceived Intelligence (PEI), Perceived Anthropomorphism (ANP), Reasons Against (RSA), Perceived Risk (PRK), Language Barrier (LBR), Technological Anxiety (TCX), Values of Openness to Change (VOP), Attitude (ATD), and Adoption Intention (AIN). Empirical validation of the model was conducted among organizational employees across technology-intensive domains, including information technology (IT), information technology enabled services (ITeS), and electronic commerce sectors. Psychometric evaluation utilizing Partial Least Squares Structural Equation Modeling (PLS-SEM) confirmed exceptional internal consistency reliability, with Cronbach’s alpha coefficients and composite reliability (CR) metrics consistently exceeding the conservative threshold of 0.80 across all subscales.

Convergent validity was rigorously demonstrated, with average variance extracted (AVE) estimates surpassing the benchmark of 0.50 for all latent dimensions. Discriminant validity was established via both the Fornell-Larcker criterion—demonstrating that squared shared variances were systematically lower than corresponding AVE values—and contemporary structural modeling diagnostics. Procedural and statistical evaluations of common method bias (CMB), including Harman’s single-factor test, revealed that the primary factor accounted for only 26.82% of total variance, confirming the absence of substantial common method distortion. This instrument represents a pivotal contribution to organizational behavior, human resource management (HRM), and human-computer interaction (HCI) research.

2. Keywords

Adoption Intention, Artificial Intelligence Adoption, AI-Based Employee Experience Chatbots, Behavioral Reasoning Theory, Human-Computer Interaction, Technological Anxiety, Perceived Anthropomorphism, Perceived Risk, Openness to Change, Organizational Behavior, Structural Equation Modeling, Psychometrics

3. Authors

The conceptualization, structural design, and empirical psychometric validation of the model were executed by an international consortium of scholars specializing in management information systems, organizational technology, and consumer behavior:

• Rajasshrie Pillai: Pune Institute of Business Management, Pune, India. Specializes in disruptive technological integration and organizational human resource informatics.
• Yamini Ghanghorkar: COEP Technological University, Pune, India. Focuses on organizational behavior, digital transformation, and workforce technology readiness.
• Brijesh Sivathanu: Pune Institute of Business Management, Pune, India. Expert in human resource analytics, technology adoption models, and digital employee experience paradigms.
• Raed Algharabat: College of Business and Economics, Qatar University, Doha, Qatar. Research focus encompasses digital marketing, human-agent interaction, and consumer psychology. ORCID: 0000-0002-1870-1708.
• Nripendra P. Rana (Corresponding Author): College of Business and Economics, Qatar University, Doha, Qatar. Globally recognized scholar in emerging technology adoption, e-governance, and information systems behavioral modeling. Direct correspondence email: [email protected].

4. Purpose

The rapid digital transformation of internal enterprise architectures has catalyzed widespread integration of artificial intelligence systems designed to streamline organizational processes. Within contemporary human resource ecosystems, AI-based Employee Experience (EEX) chatbots serve as direct, automated interfaces handling complex internal inquiries, benefits administration, continuous onboarding, compliance verification, and performance self-evaluations. Despite substantial institutional capital expenditures on enterprise automation, human resource practitioners and organizational psychologists frequently observe persistent resistance, psychological avoidance, and sub-optimal utilization among personnel.

The primary purpose of the Adoption of Artificial Intelligence-Based Employee Experience Chatbots–Model is to provide an empirically validated, theoretically comprehensive diagnostic instrument capable of assessing both the facilitating cognitive mechanisms and the inhibiting psychological frictions governing an employee’s intention to adopt AI chatbots in the workplace. Unlike conventional technology acceptance models that presume a singular continuum of adoption drivers, this instrument was intentionally constructed to capture the distinct, concurrent forces that motivate or dissuade end users.

In academic research applications, the instrument provides an empirical bridge between classical human-computer interaction (HCI) metrics and advanced organizational psychology theories. It allows researchers to systematically quantify how technological attributes (such as natural language interactivity, perceived cognitive autonomy, and humanlike nonverbal cues) interface with deep-seated personal values and emotional impediments. Scholars can employ this multi-dimensional inventory to evaluate complex moderation-mediation dynamics across cross-cultural, cross-generational, and multi-industry enterprise settings.

In clinical, industrial-organizational (I-O), and human resources consulting contexts, the instrument serves an indispensable diagnostic role. When organizations deploy AI-driven internal support platforms, generic user satisfaction questionnaires often fail to uncover why adoption metrics stagnate. By deploying this 43-item assessment, workplace psychologists and HR change managers can pinpoint specific bottlenecks. For example, the tool distinguishes whether resistance stems from objective interface deficiencies (such as language barriers or poor personalization) or subjective psychological threats (such as intense technological anxiety or systemic perceived risks regarding workplace surveillance and career obsolescence). Consequently, organizational interventions can be precision-tailored to offer targeted upskilling, psychological safety initiatives, or algorithmic transparency protocols.

5. Psychological Construct

The theoretical architecture of the instrument encompasses 11 interrelated latent constructs, capturing cognitive evaluations, psychological traits, functional perceptions, and behavioral dispositions:

1. Personalization (PES)

Personalization refers to the degree to which an employee perceives that the AI chatbot tailors its communicative responses, service delivery, and information architecture to individual user preferences, organizational historical context, and immediate administrative needs. Drawing on human factors engineering, this construct captures the transition of automated software from a generic, static knowledge base to an adaptive system capable of recognizing user-specific roles, departmental requirements, and individual communication styles.

2. Interactivity (INT)

Interactivity reflects the perceived responsiveness, bi-directional dynamic exchange, and real-time synchronicity of the dialogue between the human employee and the conversational AI agent. It measures whether the interaction feels mechanical and rigid or fluid and conversational, assessing the chatbot’s capacity to maintain context over multi-turn exchanges without forcing unnatural conversational restarts.

3. Perceived Intelligence (PEI)

Perceived Intelligence operationalizes the user’s subjective assessment of the chatbot’s cognitive competencies, reasoning capabilities, knowledge depth, and situational problem-solving aptitude. Within organizational environments, this construct determines whether employees deem the agent competent enough to resolve intricate policy dilemmas, compensation inquiries, and operational problems without human intervention.

4. Perceived Anthropomorphism (ANP)

Perceived Anthropomorphism captures the extent to which an employee attributes distinct humanlike characteristics, social presence, natural conversational flow, and psychological agency to the automated agent. Rooted in social response theory and the Computers Are Social Actors (CASA) paradigm, this dimension measures the degree to which users subconsciously apply social scripts to their algorithmic interactions.

5. Reasons Against (RSA)

In accordance with Behavioral Reasoning Theory, Reasons Against constitutes a meta-construct or composite conceptual space reflecting the specific, context-dependent cognitions that justify an employee’s non-adoption or active rejection of the technology. Rather than merely being the mathematical inverse of reasons to adopt, these cognitions represent substantive counter-arguments that rationalize avoidance behaviors.

6. Perceived Risk (PRK)

Perceived Risk measures the multi-faceted subjective expectation of loss or negative repercussions associated with deploying an internal organizational AI chatbot. This subscale encompasses concerns regarding data privacy violations, algorithmic bias in HR decisions, confidentiality breaches regarding personal medical/benefit queries, and potential adverse career consequences resulting from automated logging of interactions.

7. Language Barrier (LBR)

Language Barrier measures the communication friction experienced when the AI fails to comprehend idiomatic phrasing, workplace jargon, technical dialects, or multi-lingual syntax commonly found in diverse, multinational workforces. It captures the cognitive fatigue and frustration employees experience when modifying their natural communicative patterns to accommodate rigid natural language processing (NLP) models.

8. Technological Anxiety (TCX)

Technological Anxiety captures a state of psychological apprehension, fear, or affective discomfort evoked when an employee interacts with or considers relying upon complex AI-driven systems. Drawing from established literature on technostress, this construct assesses deep-seated fears regarding personal technical inadequacy, user error, or broader anxieties surrounding human labor displacement by intelligent software.

9. Values of Openness to Change (VOP)

Derived from Schwartz’s Theory of Basic Human Values, Openness to Change represents a higher-order motivational orientation reflecting independent thought, readiness for novel experiences, cognitive flexibility, and enthusiasm for operational innovation. In this scale, it acts as a foundational psychological disposition predisposing individuals toward embracing disruption.

10. Attitude (ATD)

Attitude constitutes an overall enduring affective and evaluative disposition toward utilizing AI-based employee experience chatbots. It encapsulates whether an employee views the prospect of delegating their internal administrative and experience management to an AI as positive, desirable, beneficial, and psychologically acceptable.

11. Adoption Intention (AIN)

Adoption Intention represents the focal outcome variable: an employee’s conscious, subjective behavioral likelihood, plan, and affirmative commitment to actively utilize, interact with, and rely upon AI-based EEX chatbots for workplace and human resource inquiries in their day-to-day organizational life.

6. Theoretical Framework

The architectural foundation of this measurement model is anchored in Behavioral Reasoning Theory (BRT), originally formulated by John R. Westaby (2005). Classical models of human decision-making, such as the Technology Acceptance Model (TAM; Davis, 1989) and the Unified Theory of Acceptance and Use of Technology (UTAUT; Venkatesh et al., 2003), operate on the core premise that user acceptance is governed primarily by utilitarian beliefs—such as perceived usefulness and perceived ease of use—which linearly determine attitudes and intentions. However, organizational researchers have recognized that these traditional frameworks exhibit critical empirical limitations: they understate the unique psychological mechanisms underlying rejection, treat non-adoption merely as an absence of motivational drivers, and fail to explain why individuals hold positive attitudes toward an innovation yet refuse to adopt it in practice.

Behavioral Reasoning Theory resolves these conceptual limitations by establishing an integrated cognitive framework composed of four structural tiers:

1. Individual Values: Deep-seated global beliefs that guide human behavior and prioritize subjective goals across domains. In this model, Openness to Change (VOP) serves as the guiding value orientation influencing specific cognitive beliefs.
2. Context-Specific Reasons: BRT explicitly differentiates between “reasons for” adoption (facilitators) and “reasons against” adoption (inhibitors). The scale operationalizes reasons for adoption through functional system evaluations: Personalization (PES), Interactivity (INT), Perceived Intelligence (PEI), and Perceived Anthropomorphism (ANP). Conversely, reasons against adoption are operationalized through specific barrier constructs: Perceived Risk (PRK), Language Barrier (LBR), and Technological Anxiety (TCX). Importantly, BRT posits that reasons for and against adoption are not polar opposites on a single continuum; an employee may simultaneously perceive high technological intelligence (a reason for) while experiencing acute technological anxiety (a reason against).
3. Attitude: The overall positive or negative psychological evaluation of the target behavior. Under BRT, context-specific reasons directly shape attitude, providing post-hoc cognitive justifications that reinforce or alter the individual’s global disposition. Crucially, reasons also maintain a direct structural link to behavioral intention, bypassing attitude altogether.
4. Adoption Intention: The proximate behavioral decision to execute the target action, determined concurrently by foundational values, context-specific cognitive reasons, and generalized attitude.

By implementing BRT, Pillai et al. (2024) synthesized complementary insights from the Computers Are Social Actors framework (Nass et al., 1994) and Innovation Resistance Theory (Ram & Sheth, 1989). Anthropomorphic and interactivity cues trigger social cognition, leading the employee to treat the software as a collaborative partner. Simultaneously, the psychological risk barriers capture the perceived threat to cognitive autonomy, privacy, and identity. Consequently, the model offers an exceptionally robust, dual-pathway psychometric lens capable of mapping both adoption acceleration and algorithmic resistance.

7. Validity

The Adoption of Artificial Intelligence-Based Employee Experience Chatbots–Model underwent extensive, multi-phase psychometric validation utilizing Partial Least Squares Structural Equation Modeling (PLS-SEM), demonstrating exceptional construct, convergent, and discriminant validity across corporate sample cohorts.

Content and Face Validity

To establish rigorous content validity, the authors adapted candidate measurement items from highly validated, peer-reviewed foundations in technology acceptance, innovation resistance, and social computing literature (e.g., Claudy et al., 2015; Gupta & Arora, 2017; Westaby, 2005). The contextual re-specification of these items to internal enterprise conversational agents was evaluated through iterative qualitative reviews by a panel of academic subject-matter experts and senior human resource executives. This panel evaluated semantic clarity, organizational realism, and construct alignment, ensuring that the 43 items precisely mirrored workplace communication dynamics.

Convergent Validity

Convergent validity demonstrates that items representing a specific latent construct share a high proportion of common variance. In the empirical validation by Pillai et al. (2024), convergent validity was evaluated via two primary criteria: individual indicator factor loadings and the Average Variance Extracted (AVE). All 43 standardized indicator loadings exceeded the conventional benchmark of 0.708, confirming that each item shared more than 50% of its variance with its assigned latent construct. Furthermore, the calculated AVE coefficients for every construct surpassed the required 0.50 threshold established by Fornell and Larcker (1981), indicating that the latent constructs explained the vast majority of variance in their respective measurement indicators rather than measurement error.

Discriminant Validity

Discriminant validity—the degree to which a given latent construct empirically diverges from other theoretical constructs in the structural system—was confirmed using rigorous classical and contemporary criteria:

• Fornell-Larcker Criterion: The square root of the AVE for each latent variable was substantially higher than its bivariate correlation with any other construct in the model. Equivalently, the squared variance shared between any two constructs was systematically lower than their individual AVE values.
• Cross-Loading Inspection: Every individual survey item loaded substantially higher on its hypothesized latent construct than on any cross-construct, demonstrating clear structural boundaries.

Predictive and Nomological Validity

Nomological validity was verified through structural path significance in the hypothesized directions. As predicted by Behavioral Reasoning Theory, functional drivers (Personalization, Interactivity, Intelligence, Anthropomorphism) demonstrated significant positive relationships with Attitude and Adoption Intention. Conversely, the inhibitor dimensions (Perceived Risk, Language Barrier, Technological Anxiety) displayed strong, statistically significant negative relationships with Attitude and directly suppressed Adoption Intention. The high coefficient of determination ($R^2$) values observed across Attitude and Adoption Intention verified that the operationalized constructs account for substantial variance in real-world employee adoption behavior.

8. Reliability

The reliability of the Adoption of Artificial Intelligence-Based Employee Experience Chatbots–Model was evaluated through multi-faceted internal consistency diagnostics. High internal consistency ensures that the observed measurement indicators yield stable, repeatable, and coherent representations of their target theoretical constructs.

In psychometric structural modeling, reliance solely on Cronbach’s alpha is often critiqued due to its unweighted assumption of tau-equivalence, which can underestimate reliability. To ensure methodological rigor, Pillai et al. (2024) assessed internal consistency utilizing both Cronbach’s alpha ($lpha$) and Composite Reliability (CR; Dillon-Goldstein’s $
ho$). The composite reliability metric weights individual items based on their actual factor loadings within the structural measurement model.

The statistical findings established exceptional internal reliability across all 11 constructs:

• Cronbach’s Alpha ($lpha$): Every subscale demonstrated an alpha coefficient exceeding 0.80, substantially surpassing the standard threshold of 0.70 for research instruments and satisfying the stringent 0.80 benchmark recommended for high-stakes psychological diagnostics.
• Composite Reliability (CR): The composite reliability metrics for all 11 latent variables consistently exceeded 0.80, with several core constructs reaching values above 0.90, confirming an exceptional degree of shared variance and internal cohesion among the item blocks.
• Average Variance Extracted (AVE): Reliability is further reinforced by AVE values strictly greater than 0.50 throughout the instrument, establishing that random measurement error accounts for a minimal fraction of the observed score variance.

The instrument exhibits negligible item-specific error variance, confirming that the 43 items reliably assess their designated cognitive, affective, and behavioral constructs across diverse technological environments.

9. Factor Analysis

The dimensionality, factor structure, and common method properties of the scale were comprehensively validated through variance-based structural modeling and diagnostics designed to confirm the operationalization of the Behavioral Reasoning Theory framework.

Measurement Model Specification

The measurement architecture was tested within a reflective structural equation modeling paradigm via PLS-SEM. Unlike covariance-based SEM (CB-SEM), which prioritizes global goodness-of-fit minimization, PLS-SEM is particularly well-suited for complex frameworks incorporating both facilitative and inhibiting direct and indirect pathways derived from behavioral reasoning paradigms. The eleven-factor reflective structure exhibited robust factor loadings across all 43 indicators, with all empirical loadings surpassing the standardized 0.70 threshold ($p < 0.001$). No cross-loadings approached problematic thresholds, affirming distinct empirical conceptualization.

Assessment of Common Method Bias (CMB)

Because organizational surveys capturing perceptual, affective, and intentional data from a single cohort of respondents are susceptible to common method variance, the authors implemented both procedural remedies (e.g., maintaining respondent confidentiality, randomizing item placement, and providing precise semantic instructions) and rigorous statistical diagnostic tests:

• Harman’s Single-Factor Test: An unrotated exploratory factor analysis was executed incorporating all 43 measurement items. The analysis revealed multiple distinct factors with eigenvalues greater than 1.0. Crucially, the initial, dominant factor accounted for only 26.82% of the total variance. Because this value falls substantially below the conservative 50% threshold, it demonstrates that the variance is not driven by a singular general method factor.
• Full Collinearity Assessment: In accordance with PLS-SEM best practices, variance inflation factors (VIFs) were computed across the latent variables. All inner-model VIF scores remained well below the threshold of 3.3, ruling out the presence of pathological multicollinearity or common method contamination.

In summary, the factor analytical procedures verified an eleven-dimensional structure that exhibits structural stability, robust factor loadings, and freedom from common method artifacts.

10. Instrument / Measurement Tool

The Adoption of Artificial Intelligence-Based Employee Experience Chatbots–Model is structured as a standardized, self-report organizational inventory:

• Instrument Type: Standardized diagnostic inventory / self-report survey scale.
• Test Type: Original empirical model operationalizing Behavioral Reasoning Theory in artificial intelligence interactions.
• Target Population: Adult organizational workforce members, particularly employees operating within information technology (IT), information technology enabled services (ITeS), digital commerce, and modern corporate enterprises utilizing or piloting AI HR tools.
• Number of Items: 43 items distributed across 11 discrete subscales.
• Administration Format: Computer-assisted web interviewing (CAWI), enterprise intranet surveys, or paper-and-pencil psychometric batteries. Average completion time: 10 to 15 minutes.
• Response Scale: Standard 5-point Likert scale formatted as:
  • 1 = Strongly Disagree
  • 2 = Disagree
  • 3 = Neutral (Neither Agree nor Disagree)
  • 4 = Agree
  • 5 = Strongly Agree
• Scoring and Computational Procedures:
  • Construct scores are generated by calculating the arithmetic mean of the items comprising each subscale. Alternatively, in structural modeling contexts, latent variable scores can be computed via PLS weightings.
  • High scores on Personalization (PES), Interactivity (INT), Perceived Intelligence (PEI), and Perceived Anthropomorphism (ANP) indicate highly favorable functional appraisals of the AI system (Reasons For Adoption).
  • High scores on Reasons Against (RSA), Perceived Risk (PRK), Language Barrier (LBR), and Technological Anxiety (TCX) signify acute psychological, functional, and emotional barriers to adoption.
  • High scores on Values of Openness to Change (VOP) reflect a strong baseline individual disposition toward novelty and technological experimentation.
  • High scores on Attitude (ATD) and Adoption Intention (AIN) indicate strong affective readiness and high behavioral commitment to utilizing conversational agents in the workplace.

11. Permissions & Fee and Test Year

The conceptual model and psychometric scales were published in 2024 in Information Technology & People (Emerald Publishing). The foundational scholarly article is titled “Adoption of artificial intelligence (AI) based employee experience (EEX) chatbots” (Pillai et al., 2024).

Regarding academic and organizational application permissions:

• Academic and Non-Commercial Research: The conceptual framework, structural model, and psychometric dimensions may be utilized, adapted, and referenced by university researchers, graduate students, and independent academics for non-commercial scientific inquiries, provided appropriate formal academic citation is rendered to Pillai et al. (2024) and Emerald Publishing.
• Commercial and Enterprise Application: Corporate entities, human resources consultancies, and commercial software vendors wishing to embed this inventory into commercial diagnostic software, employee pulse-survey platforms, or proprietary change-management toolkits must consult Emerald Publishing’s rights and permissions clearance department or contact the corresponding author, Prof. Nripendra P. Rana ([email protected]), to secure necessary enterprise licensing or reproduction approvals.
• Fees: Academic usage generally incurs no licensing fees under fair-use scientific research guidelines, provided the publication is cited appropriately. Commercial implementations may be subject to contractual or licensing conditions as determined by the copyright holders.

12. References

• Claudy, M. C., Garcia, R., & O’Driscoll, A. (2015). Consumer resistance to innovation—A behavioral reasoning perspective. Journal of the Academy of Marketing Science, 43(4), 528–544. https://doi.org/10.1007/s11747-014-0399-0
• 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
• Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18(1), 39–50. https://doi.org/10.1177/002224378101800104
• Gupta, A., & Arora, N. (2017). Consumer adoption of m-banking: A behavioral reasoning theory perspective. International Journal of Bank Marketing, 35(4), 733–747. https://doi.org/10.1108/IJBM-11-2016-0162
• Nass, C., Steuer, J., & Tauber, E. R. (1994). Computers are social actors. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 72–78). ACM. https://doi.org/10.1145/191666.191703
• Pillai, R., Ghanghorkar, Y., Sivathanu, B., Algharabat, R., & Rana, N. P. (2024). Adoption of artificial intelligence (AI) based employee experience (EEX) chatbots. Information Technology & People, 37(1), 449–478. https://doi.org/10.1108/ITP-04-2022-0287
• Ram, S., & Sheth, J. N. (1989). Consumer resistance to innovations: The marketing problem and its solutions. Journal of Consumer Marketing, 6(2), 5–14. https://doi.org/10.1108/EUM0000000002542
• Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User acceptance of information technology: Toward a unified view. MIS Quarterly, 27(3), 425–478. https://doi.org/10.2307/30036540
• Westaby, J. D. (2005). Behavioral reasoning theory: Identifying new linkages underlying intentions and behavior. Organizational Behavior and Human Decision Processes, 98(2), 97–120. https://doi.org/10.1016/j.obhdp.2005.07.001

13. Items of the Scale