The New Social Image: How AI Competency and AI Proactivity Influence Self- and Peer-Perceptions in the Workplace

Abstract: Human-AI collaboration is considered the most promising way to incorporate AI in the workplace. What remains unexplored are the experiential consequences of this teaming. More specifically, in a team with AI, how humans perceive themselves (self-perception) and how they are perceived by their coworkers (peer perception) in terms of work ownership and job meaningfulness. In a 2x2x2 vignette study (n=50), participants rated perceptions of ownership, affect, job meaningfulness and satisfaction, and role dynamics across two levels (low/high) of AI proactivity and AI competency as within-subject factors, with point-of-view (self perception/peer perception) as between-subjects. Our results showed that AI with low competency or low proactivity generally improved feelings related to ownership, meaningfulness, satisfaction, and role dynamics, and also increased positive affect while reducing negative affect. However, these effects were often influenced by point-of-view. For instance, low AI proactivity resulted in higher job satisfaction from self-perception rather than peer perception. Based on our findings, we argue that designing AI for the future of work solely around performance metrics may not be adequate. Highly competent and proactive AI-driven systems can have undesirable impacts on perceptions of ownership, job identity, social image and team dynamics, and consequently, job meaningfulness.

• The paper demonstrates through a 2×2×2 vignette experiment with 50 participants that highly competent and proactive AI reduces psychological ownership, job meaningfulness, satisfaction, and motivating potential, with MPS falling from 50.21 to 21.54 in key conditions.
• The findings show that AI competency can increase self-efficacy when the system remains nonintrusive, while high proactivity consistently weakens territoriality, accountability, autonomy, and workers’ sense of control.
• The paper reveals a self–peer perception gap in which highly competent, proactive AI may appear supportive or effective to observers but make workers feel dominated, reduce social image, and shift the AI’s perceived role from teammate to superior.

Overview and motivation

This paper examines an underexplored dimension of human-AI teaming: not how humans perceive the AI, but how working with AI shapes how humans perceive themselves (self-perception) and how they are perceived by coworkers (peer perception). The authors, Ghosh, Hassenzahl, and Sadeghian, argue that job meaningfulness depends partly on social recognition and social image at work, and that highly competent or highly proactive AI systems can erode these values by taking over tasks that define a worker's identity. They frame their inquiry around three research questions concerning ownership of work outcomes, job meaningfulness and affect, and role dynamics.

The central claim is provocative relative to mainstream AI development practice: designing workplace AI solely around performance metrics may be inadequate, because a maximally competent and proactive system can damage perceptions of ownership, job identity, social image, and team dynamics — and consequently job meaningfulness (2606.00182).

Study design

The study is a $2\times2\times2$ mixed-design video vignette experiment ($n=50$, 25 per point-of-view) conducted via Prolific. Within-subject factors were AI competency (low/high) and AI proactivity (low/high); point-of-view (self vs. peer) was between-subjects. Participants watched four videos of a scripted office team meeting in which a team lead interacts with three team members and an AI-driven system called "TAI," a prototype combining a Bluetooth speaker in a decorated earthen pot with pre-recorded TTS audio. TAI's behavior followed Meurisch et al.'s reactive/proactive categorization and Kraus et al.'s implicit anticipation: low-competency TAI gave vague or incorrect suggestions; high-competency TAI gave precise, contextually relevant answers; high-proactivity TAI intervened unprompted and frequently.

A methodological strength is the role-matched recruitment: participants who reported to a supervisor answered from the team member perspective (peer perception), while those in supervisory roles answered as the team lead (self-perception). Manipulation checks confirmed participants distinguished all four conditions (e.g., $t(49)=-11.44$, $p<0.001$ for competency between LCHP and HCHP). Measures included a customized Psychological Ownership Questionnaire (Territoriality, Self-Efficacy, Accountability), the Job Diagnostic Survey including the Motivating Potential Score, single-item ratings of meaningfulness, satisfaction, and positive/negative affect, and role ratings of TAI as superior/subordinate/teammate.

Ownership outcomes (RQ1)

All three ownership sub-scales showed significant main effects of both competency and proactivity, consistently favoring the less capable or less assertive AI:

Sub-scale	Effect direction	Key statistic
Territoriality	Higher when AI less competent / less proactive	Proactivity main effect $F(1,48)=122.96$, $\eta^2=0.324$
Self-Efficacy	Higher self-efficacy with high competency, but higher with low proactivity	Competency $F(1,48)=5.98$; proactivity $F(1,48)=27.88$
Accountability	Higher when AI less competent / less proactive	Competency $F(1,48)=29.62$

Two nuances are notable. First, self-efficacy was the one ownership measure where high AI competency helped — but only when paired with low proactivity, suggesting that expertise without intrusion preserves confidence. Second, accountability showed a competency × point-of-view interaction: peers attributed more accountability to the team lead when TAI was incompetent than the lead attributed to himself, indicating that observers credit humans for compensating for AI shortcomings more than the humans credit themselves.

Job meaningfulness, satisfaction, and affect (RQ2)

The JDS results show that every motivating characteristic — skill variety, task identity, task significance, autonomy, feedback, and overall MPS — declined significantly when TAI was more competent or more proactive. The MPS dropped sharply in the high-competency/high-proactivity condition (self-view mean 21.54 vs. 50.21 for high-competency/low-proactivity). Meaningfulness and satisfaction showed the same pattern, with strong interaction effects: the drop in satisfaction with high proactivity was only significant when competency was also high ($t(48)=9.53$, $n=50$0). Qualitative responses captured this vividly: with a highly competent, proactive TAI, participants said the team lead became "TAI's assistant" and that "TAI took away the whole glory."

Point-of-view moderated several effects, revealing a divergence between self- and peer-perception:

• Positive affect: high AI competency increased positive affect from the self-perspective (TAI seen as supportive) but decreased it from the peer perspective (the lead seen as deferring to TAI). Conversely, high proactivity hurt self-perception more than peer perception.
• Negative affect: high proactivity raised negative affect more strongly in self-perception ("bossy," "intrusive"), whereas low competency frustrated only the lead himself while peers admired his composure and leadership.
• Job satisfaction: the benefit of low proactivity was larger from the self-perspective than the peer perspective.

These divergences constitute the paper's most interesting finding: the same AI configuration can simultaneously flatter the human collaborator in others' eyes and undermine her own sense of agency, creating a trade-off between social image and experienced autonomy.

Role dynamics (RQ3)

Perceived role shifted systematically with TAI's configuration. A highly competent and proactive TAI was rated as a superior (mean 5.20 self-view on a 7-point scale vs. 1.56–2.28 in other conditions), described as "dominant and assertive" and "acting like a know-it-all." Low-competency/high-proactivity TAI remained a subordinate despite its intrusiveness because it was constantly corrected. High-competency/low-proactivity TAI occupied the most favorable position — perceived as both subordinate and teammate, "helping the team but not overstepping." Notably, high proactivity increased perceived superiority more in self-perception than peer perception, meaning the lead felt dominated even when observers did not fully register it.

Design implications

The authors draw two design conclusions. First, AI should augment rather than replace human initiative: preserving opportunities to exercise skills, make decisions, and drive tasks to completion sustains ownership and meaningfulness, going beyond Shneiderman-style supervision-based human control. Second, AI competency and proactivity should be tailored to the intended social role and team context; they suggest adaptive designs that modulate proactivity based on engagement or emotional cues, stepping back when users show frustration. They also note contextual boundaries: in socially visible coordination roles, a proactive AI threatens social image, whereas for solo workers this effect may largely disappear.

Limitations and open questions

The paper concedes several constraints plainly. The vignette method measures hypothetical intentions rather than actual behavior, raising external-validity concerns, though the authors argue internal validity is supported by successful manipulation checks and realistic scenario design. The study used a single artifact design (TAI) in a single meeting context; anthropomorphism, physical movement, tone of speech, and longer-term adaptation were not manipulated, and the authors acknowledge these could alter the results. Task performance and productivity were not measured, so the paper cannot say whether the meaningfulness gains of a less capable AI come at a performance cost — the authors explicitly flag this trade-off as unexamined. Finally, interactions unfolded over minutes, not the weeks or months over which trust, resistance, and role renegotiation actually develop.

Conclusion

Through a factorial vignette study, this paper demonstrates that AI competency and proactivity have systematic, often adverse effects on human self- and peer-perceptions of ownership, meaningfulness, affect, and role dynamics. Less competent or less proactive AI improved nearly every experiential measure, and point-of-view moderated key effects, exposing tensions between how workers see themselves and how colleagues see them. The paper's argument — that performance-centric AI design neglects experiential consequences that matter for meaningful work — is well supported within its scope, though longitudinal field studies and performance trade-off analyses remain necessary to establish its generalizability.