PaperlayerView exact revision on arXiv

Source-linked AI summary

Understanding AI Anxiety in the Workplace: A Multimethod Investigation Using Fear Acquisition Theory and the Technology Acceptance Model

Jaroslaw Grobelny, Mateusz Klakus, Kacper Szymański, Teresa Chirkowska-Smolak

arXiv:2608.14609v1cs.CYcs.AI

TL;DR

AI job replacement anxiety remains insufficiently understood, particularly regarding whether it arises from direct displacement or vicarious perceptions of AI control. Across an experiment and an observational study, the paper tests this mechanism using Integrated Fear Acquisition Theory and TAM, finding that perceived AI-over-human control is associated with greater anxiety and that perceived usefulness moderates the relationship.

  • Problem

    Evidence on how AI job replacement anxiety forms is limited, despite concerns about AI-driven displacement and prior suggestions that vicarious exposure may matter.

  • Method

    Two studies combined a randomized experiment (N = 316) with a large observational study (N = 995) to examine AI-control narratives and moderation by perceived usefulness and ease of use.

  • Results

    Perceived AI-over-human control increased or was associated with higher job replacement anxiety; perceived usefulness moderated this relationship, whereas perceived ease of use did not.

  • Takeaways & Limitations

    The findings support a theory-informed account in which perceptual and vicarious mechanisms can trigger AI job replacement anxiety, with perceived usefulness shaping the response.

  • Takeaways & Limitations

    Because vicarious exposure was operationalized mainly through curated video narratives, interpersonal communication and organizational messaging may involve different psychological processes.

Abstract

from arXiv · show

As artificial intelligence (AI) rapidly diffuses and concerns about job displacement intensify, the psychological mechanisms underlying AI job replacement anxiety remain insufficiently understood. Drawing on Integrated Fear Acquisition Theory and the Technology Acceptance Model, the present research investigates whether AI job replacement anxiety can be elicited through vicarious exposure to narratives emphasizing AI-over-human control, and whether perceived usefulness and perceived ease of use of AI moderate this response. Across two studies, we examine AI job replacement anxiety as a response that emerges through vicarious exposure to narratives emphasizing AI agency and human control loss, rather than through direct personal experience of job displacement. Study 1 employed a randomized experiment (N = 316), demonstrating that such exposure increased AI job replacement anxiety. This effect was moderated by perceived usefulness of AI, but not by perceived ease of use, and remained robust after controlling for core self-evaluations. Study 2 (N = 995) replicated the association between perceived AI-over-human control and job replacement anxiety in an observational design and provided convergent evidence for the moderating role of perceived usefulness, supporting the external validity of the findings. Together, the results provide the first causal evidence that perceptual and vicarious processes can trigger AI job replacement anxiety. By shifting attention from structural labor-market conditions to how AI agency is perceived and communicated, this work offers a mechanism-based account of when and why AI-related job fears arise.

Introduction

The paper examines how AI job replacement anxiety forms through vicarious exposure to narratives portraying AI as gaining control over human roles and decisions. It integrates Integrated Fear Acquisition Theory with TAM to test whether perceived usefulness and ease of use shape this response.

  • AI job replacement anxiety reflects forward-looking fears that AI will make human labor obsolete, including concerns about job loss, reduced human skills, and autonomy.
  • Existing evidence suggests job replacement anxiety may arise vicariously from information about others losing jobs to AI rather than direct personal displacement.
  • The research addresses limited evidence on anxiety formation through a randomized experiment and a large observational study, controlling for core self-evaluations in the experiment.
  • Narratives depicting AI as operating independently and transferring agency from humans to autonomous systems provide the proposed fear-eliciting threat cue.
  • The study hypothesizes that AI-control narratives increase job replacement anxiety compared with human-control narratives, while perceived usefulness and ease of use moderate this relationship.
  • TAM constructs are treated as appraisal filters that shape the credibility and personal relevance of threat cues rather than as direct predictors of anxiety.

Method

The study measured AI perceptions, job replacement anxiety, core self-evaluations, and other psychological and demographic characteristics using established or adapted scales. The experimental sample and sensitivity analysis were specified before analysis.

  • The analytic sample comprised 316 participants after exclusions from an intended sample of 325 adults with current or recent labor-market experience.
  • Sensitivity analysis indicated a minimum detectable effect of R² = .072 under conservative assumptions for the 316-participant regression model.
  • Perceived AI control was assessed with two items measuring whether AI operates independently and makes decisions without human intervention.
  • Perceived ease of AI use and perceived usefulness of AI were measured with four-item TAM-based scales capturing effortlessness and performance enhancement, respectively.
  • AI job replacement anxiety was measured with a six-item subscale on a 5-point Likert scale showing good internal consistency, Cronbach’s α = .83.
  • Core self-evaluations were measured with the 12-item CSES, while controls included general negative AI attitudes, trait anxiety, demographics, employment context, and prior labor-market experiences.

Data Collection Procedure

A preregistered randomized 2 × 2 × 2 experiment exposed participants to curated social-media video compilations manipulating perceived control, ease of use, and usefulness. Participants evaluated the videos and then completed outcome measures.

  • Participants were randomly assigned to eight cells crossing AI versus human control, low versus high ease of use, and low versus high usefulness.
  • The procedure used individual computer workstations in a controlled laboratory setting, with informed consent and subsequent debriefing.
  • Participants viewed three compilations of short user-generated social-media videos, each lasting approximately six minutes and containing 5–10 videos.
  • Control compilations contrasted AI-over-human agency with human-over-AI agency through topics including workplace surveillance, AI monitoring, and technological singularity.
  • Ease-of-use videos emphasized either simple learning and low implementation demands or complexity and difficulty applying AI at work.
  • Usefulness videos emphasized either AI’s workplace assistance and performance benefits or its limitations and shortcomings.

Results

The results supported the predicted association between perceived AI control and job replacement anxiety, but not the predicted ease-of-use moderation. Perceived usefulness significantly moderated the relationship in the opposite direction from the hypothesis and remained positively associated with anxiety overall.

  • AI-over-human control produced higher descriptive anxiety than human control, with means of 3.74–4.08 versus 2.90–3.78.
  • The experimental manipulations significantly changed perceived AI control, ease of use, and usefulness, while baseline AI-control perceptions did not differ between conditions.
  • The regression model was significant, F(10, 306) = 20.90, p < .001, explaining 40.6% of variance in AI anxiety.
  • AI-control exposure predicted higher job replacement anxiety, b = 0.96, SE = 0.15, t = 6.26, p < .001, supporting H1.
  • Ease of use did not moderate the exposure–anxiety relationship, b = −0.07, SE = 0.18, t = −0.37, p = .710, rejecting H2.
  • Usefulness moderated the relationship opposite to prediction, while usefulness independently predicted anxiety, b = 1.00, SE = 0.12, t = 8.08, p < .001.
  • The AI-control difference was larger at low usefulness, b = 0.92, than high usefulness, b = 0.33, although both were statistically significant.

Study 2

Study 2 assessed the proposed model in a large-scale observational design to examine whether Study 1’s experimentally identified associations generalized to a naturalistic setting. Because perceived AI-over-human control was measured rather than manipulated, Study 2 could not establish causality or directly replicate Study 1’s mechanism.

  • Study 2 measured all focal variables as observed constructs in a large-scale observational design to examine the model’s external validity.
  • Because perceived AI-over-human control was not experimentally manipulated, Study 2 could not establish causality.Its purpose was to assess whether experimentally identified associations generalized to a naturalistic setting, not to provide a second causal test.

Method

The method describes participant eligibility and sampling for Study 2, alongside measures of AI anxiety, control variables, perceived AI control, perceived ease of use, and perceived usefulness. A sensitivity analysis evaluated the study’s ability to detect small regression effects.

  • Participants had to be adults who were currently employed and not platform-based workers.
  • 1,019 individuals remained after screening and attention-check exclusions from 1,715 survey initiators.The passage reports 1,423 completers, 1,218 cases after screening exclusions, and a final analytic sample of 1,019.
  • With N = 995, α = .01, and power = .99, sensitivity analysis yielded a critical R² of .017 for the regression models.
  • The study measured AI replacement anxiety, core self-evaluations, demographic and employment variables, prior AI work experience, and related controls.
  • Perceived AI-over-human control was assessed with 10 items covering AI control over humans and human control over AI.The measure showed acceptable internal consistency, Cronbach’s α = .84, with Feldt’s 95% CI [.82, .85].
  • Perceived ease of use and perceived usefulness were measured with six-item Technology Acceptance Model scales adapted to AI technologies.Usefulness captured beliefs that AI enhances performance across work-related contexts, while ease of use captured effortlessness of interaction.

Data Collection Procedure

Study 2 used an ethics-approved online questionnaire with consent, eligibility screening, attention checks, and compensation procedures. Regression analyses examined AI replacement anxiety as a function of perceived AI-over-human control, its interactions with AI-use perceptions, and core self-evaluations.

  • Participants gave electronic informed consent after instructions, and screening questions excluded ineligible respondents before the questionnaire continued.
  • Data collection ran from June 3 to July 10, 2025, and the average survey took 13 minutes and 57 seconds.
  • Three embedded attention checks excluded participants who responded incorrectly, while non-monetary panel compensation reduced incentives for automated responses.
  • Study 2 variables were positively intercorrelated: AI anxiety correlated with perceived AI-over-human control, ease of use, and usefulness.Core self-evaluations showed small but significant associations with focal constructs.
  • The regression modeled AI replacement anxiety from perceived AI-over-human control, interactions with ease of use and usefulness, and core self-evaluations.Influential observations were identified using a preregistered, conservative, multi-criterion procedure.
  • The model explained 28.2% of variance in AI job replacement anxiety, R² = .282, adjusted R² = .277, with no detected heteroskedasticity or multicollinearity concerns.The overall model was statistically significant, F(4, 990) = 64.53, p < .001; VIFs were ≤ 1.30.
  • Perceived AI-over-human control was positively associated with anxiety, b = 0.55, while usefulness moderated this association and ease of use did not.The usefulness interaction was b = −0.07, p = .03; the ease-of-use interaction was b = 0.05, p = .143.

Discussion

Across an experiment and a large observational study, the research links AI job replacement anxiety to perceived AI-over-human control and vicarious narratives about AI agency. The findings distinguish usefulness from ease of use and position perceptual, context-sensitive threat appraisals as an additional pathway to anxiety.

  • Across Study 1 (N = 316) and Study 2 (N = 995), the studies examined whether narratives emphasizing AI-over-human control increase job replacement anxiety.
  • The two studies provide converging experimental and observational evidence for a theory-informed model of AI job replacement anxiety emergence.
  • Perceived AI-over-human control is identified as a proximal psychological mechanism underlying AI job replacement anxiety.The account integrates Integrated Fear Acquisition Theory with Technology Acceptance Model constructs.
  • Perceptions of autonomous AI operation functioned as a threat cue eliciting anxiety even without direct personal displacement risk.
  • The findings suggest that AI job replacement anxiety can arise from proximal appraisals of AI agency and control, not solely from structural labor-market conditions.
  • Anxiety may develop through socially transmitted narratives about AI agency before observable labor-market consequences occur.
  • Usefulness and ease of use showed asymmetric roles: usefulness moderated anxiety, whereas ease of use did not condition responses.The usefulness interaction attenuated as usefulness increased, but this did not establish that usefulness buffered anxiety; the authors offer an exploratory explanation involving its independent positive association with anxiety.
  • Perceived AI-over-human control predicted anxiety beyond core self-evaluations, supporting the relevance of situational threat appraisals.

Study limitations

The studies were limited by short-term measurement, curated media exposure, unresolved trait–state distinctions, measurement reliability, and non-equivalent operationalizations across studies.

  • Temporal scope: The studies measured immediate anxiety responses, so they cannot establish durability or longer-term psychological consequences.The authors recommend longitudinal designs to examine stability and temporal dynamics.
  • Exposure format: Vicarious exposure was operationalized through curated, primarily video-based media narratives, leaving other communication modes untested.Interpersonal communication and organizational messaging may engage different psychological processes.
  • Trait–state scope: The designs could not fully distinguish transient situational responses from stable dispositional tendencies in AI job replacement anxiety.The trait–state distinction remains open for future research.
  • Measurement: The Study 1 pre-manipulation measure of perceived AI-over-human control had modest internal consistency (Cronbach’s α = .62), potentially adding measurement error.The authors recommend higher reliability for brief pre-experimental measures.
  • Cross-study comparison: Study 2 associations cannot be interpreted as causal because perceived AI-over-human control was operationalized differently across studies.The convergence is consistent with the proposed model but does not independently replicate its causal structure.

Conclusion

The research concludes that AI job replacement anxiety can arise through perceptual and vicarious mechanisms rather than direct technological exposure or objective employment threats.

  • Conclusion: Across two studies, perceived AI-over-human control functioned as a proximal trigger of AI job replacement anxiety.Narratives emphasizing AI agency elicited job replacement concerns even without explicit displacement cues.

Funding

The research received funding from Adam Mickiewicz University’s Initiative of Excellence programme and reported ethics approval, open materials, preregistration, no competing interests, and author contributions.

  • Funding: The research was funded by the Initiative of Excellence – Research University programme at Adam Mickiewicz University, Poznan, Poland.The funding decision number was IDUB 140/04/POB5/0019.
  • Ethics: The study received ethics approval, obtained informed consent, and followed the Declaration of Helsinki.Approval came from Adam Mickiewicz University’s Faculty of Psychology and Cognitive Science ethics committee.
  • Open science: Data and materials were made openly available through the Open Science Framework, and the study was preregistered.The passage reports no substantial discrepancies between the registered protocol and the conducted study.
  • Declarations: The authors declared no competing interests.
  • Contributions: Author contributions covered conceptualization, methodology, analysis, investigation, writing, supervision, and project administration.The listed contributors were J.G., M.K., K.S., and T.C.-S.
  • AI use: ChatGPT was used exclusively to proofread the final manuscript text, with authors retaining responsibility for the content.

Tables and Figures

The paper includes tables and figures documenting participant characteristics, descriptive statistics, regression analyses, conditional effects, and confidence intervals across the studies.

  • Tables: Table 1 presents demographic and control characteristics for participants in both studies.
  • Figures: Figure 2 displays conditional effects of perceived AI control on replacement anxiety as a function of perceived AI usefulness.The relevant comparison is the AI-control versus human-control effect at low and high perceived usefulness.
Loading 2608.14609v1…