aceModerate evidence
More
Coworkers help less when they see colleagues working with AI, a survey of 375 finds
Study title: How Employee–AI Collaboration Influences Coworkers’ Helping Behaviour: An Attribution Theory Perspective
- Evidence type
- Survey
- Sample
- n = 375
- Sources
- 1 sourcePeer-reviewed
- Published
- 3 min read
- Truvace record
- TRV-2026-2576cite or check it
Study title: How Employee–AI Collaboration Influences Coworkers’ Helping Behaviour: An Attribution Theory Perspective
A peer-reviewed study in Behavioral Sciences examined how coworkers react when they see a colleague working with AI. In a two-wave survey of 375 matched coworker evaluations, the more a colleague was seen as collaborating with AI, the less help the respondent reported giving that colleague. The raw correlation was r = -0.530, and the regression coefficient was -0.406.
Read the study abstract
As artificial intelligence (AI) is increasingly integrated into the workplace, employee-AI collaboration is evolving from a personal productivity tool to a social cue that coworkers can observe and interpret. Existing research has largely emphasised the performance and well-being effects of employee-AI collaboration; however, few studies have revealed, from the observer's perspective, its potential negative spillover mechanisms on coworkers' helping behaviour. Based on attribution theory, this study constructs a…
The problem
In a two-wave survey of 375 coworker evaluations in Chinese firms, the more coworkers saw a colleague as collaborating with AI, the less help they gave that colleague (r = -0.530), partly because they read it as laziness or avoiding responsibility.
In brief
Coworkers who saw heavy AI collaboration were more likely to suspect laziness (β = 0.551) or an attempt to dodge responsibility when results were poor (β = 0.583). Both suspicions were linked to less helping. The authors report responsibility avoidance as the stronger pathway.
The link was stronger when the human and the AI depended closely on each other's work, because coworkers could not tell where the employee's effort ended and the AI's contribution began. In the survey, the laziness pathway was not significant when interdependence was low, but it was when interdependence was high. A second experiment with 320 full-time employees was designed to test the causal effect.
Main points
- Study 1 used a two-wave survey: 450 questionnaires were sent out and 375 valid matched coworker evaluations were kept, an effective response rate of about 83.3%.
- Respondents came from firms in regions such as Hubei and Jiangsu in finance, IT services and education that had widely adopted AI, and each rated the same target coworker at both time points.
- Perceived AI collaboration correlated with less helping (r = -0.530) and with laziness attribution (r = 0.624) and responsibility-avoidance attribution (r = 0.599).
- In the regression, AI collaboration predicted laziness attribution (β = 0.551) and responsibility-avoidance attribution (β = 0.583); laziness (β = -0.169) and responsibility avoidance (β = -0.189) each predicted less helping.
- At high task interdependence (+1 SD), the indirect effect via laziness was -0.065 with a 95% confidence interval of [-0.130, -0.015]; at low interdependence it was -0.054 with an interval of [-0.109, 0.003], which includes zero.
- The indirect effect via responsibility avoidance excluded zero at both low (-0.076) and high (-0.101) interdependence.
- Study 2 was an online 2 × 2 scenario experiment with 320 full-time employees, who read about a hypothetical coworker named Zhang San using AI frequently or rarely.
The full story
The research team, writing in Behavioral Sciences in 2026, drew on attribution theory, which holds that observers infer motives from visible cues. Coworkers can see an employee use AI and see the output, but not the prompt design, information filtering, output checking and judgement correction behind it. The authors argue this gap can lead observers to read AI use as shortcut-taking.
Study 1 recruited respondents who used AI at work and could name a coworker whose AI use they could observe. At the first wave they rated the coworker's perceived AI collaboration and task interdependence. At the second wave they rated laziness and responsibility-avoidance attributions and how much help they gave. Measures were adapted from existing five-point scales, and reliability coefficients ranged from 0.885 to 0.943.
The authors tested common method bias with several checks and concluded it was not serious. The model fit the data well (CFI = 0.992, RMSEA = 0.030). Laziness attribution only partly explained the helping drop, since AI collaboration still predicted less helping with it in the model (β = -0.203).
Study 2 randomly assigned participants to four conditions that varied AI collaboration and human-AI task interdependence, then asked them to judge the coworker and report their own willingness to help. The authors designed it to add causal evidence to the survey findings.
Sources
1 on file- Behavioral Sciencesdoi:10.3390/bs16060985
Talk it over
Who carries this cost, and what would have to change to prevent it?
Coworkers help less when they see colleagues working with AI, a survey of 375 findshttps://truvace.com/item/how-employee-ai-collaboration-influences-coworkers-helping-behaviour-an-attribut
Who carries this cost, and what would have to change to prevent it?
The problem In a two-wave survey of 375 coworker evaluations in Chinese firms, the more coworkers saw a colleague as collaborating with AI, the less help they gave that colleague (r = -0.530), partly because they read it as laziness or avoiding responsibility.
Answer in the threadHow do we know? What can this kind of evidence show, and what can’t it?
The evidence This rests on a survey with a sample of 375, from Behavioral Sciences.
Answer in the threadWhat does it leave open? What else would you want to know before acting on it?
The record’s own caveat Study 1 is correlational and rests on coworkers' perceptions of a colleague's AI use rather than usage logs or actual collaboration frequency, and its participants were from firms in Chinese regions such as Hubei and Jiangsu.
Answer in the threadHave you seen AI used this way where you study, work or live? Did it go the same way?
Answer in the thread
The questions are the same on every record. The details under them are taken from this record as written.
What readers say