When AI does more of the work, how much credit does the person using it get, from others and from themselves?
How does perceived agency in AI affect attributions about user competence?
This explores how the amount of work people believe the AI did changes judgments of the human's skill, both how others rate AI users and how users rate themselves.
This explores how the amount of work people believe the AI did changes judgments of the human's skill: how other people rate someone who used AI, and how that person rates themselves. The corpus has no single study that varies how agentic the AI seems and then measures competence ratings. Read together, though, several notes show the same AI contribution being credited in opposite directions depending on who is doing the judging.
Start with how others see AI users. In experiments with more than 4,000 participants, people who used AI expected colleagues and managers to see them as less competent and less diligent, and they were less willing to say they had used it Do people fear judgment when they use AI at work?. The more the AI looks like the one doing the work, the less credit the human expects to get. That expectation changes behavior: people hide their AI use, which also hides the evidence anyone would need to judge them fairly.
On the user's own side, the error runs the other way. The 'LLM Fallacy' describes users folding AI output into their sense of their own ability How does AI-assisted work reshape how people see their own abilities?. This is a self-perception mistake, separate from trusting a wrong answer or over-relying on the tool. One likely mechanism is fluency. Polished output feels easy to process, and people read that ease as a sign that they understood or produced it Does processing ease mislead users about their own competence?. Self-assessment is a poor check on this. Across three studies, self-rated AI competence and measured AI competence correlated at only .055 Can self-ratings replace objective performance scores for AI competence?. So observers tend to give the user too little credit while the user gives themselves too much, for the same work.
What might change where the credit goes? One lever is how much the person actually steered the output. Feelings of ownership over AI-written text rose with the user's influence over it, but personalizing the AI made no difference Does user control over AI text shape feelings of ownership?. That suggests the useful variable is not how agentic the AI looks but how clearly the human's contribution can be seen. The other lever is time plus visible results. Revealing that a partner is an AI first triggers bias against it, but the bias reverses once people repeatedly see outcomes. Disclosure without that feedback calibrates nothing Does revealing AI identity help or hurt user trust?.
Here is the part you may not have expected to want. When people form a picture of an AI partner, perceived competence carries most of the weight, explaining 49% of the variation in their impressions How do users mentally model dialogue agent partners?. And people judge that competence mostly from how confident the AI sounds, not from whether it is right Do users worldwide trust confident AI outputs even when wrong?. Competence judgments about AI and about AI users may rest on the same weak surface cues. If so, the fix that recurs across these notes is making the boundaries of contribution visible: who did what, and whether it worked.
Sources 8 notes
Across four experiments with 4,439 participants, people using AI expected others to judge them as less competent and diligent, and reported lower willingness to disclose AI use to managers and colleagues. The gap suggests a social cost that users foresee and act on.
Research shows the LLM Fallacy operates through misattribution of AI outputs to personal capability, independent of output accuracy or reliance behavior. It requires interventions that clarify human-machine contribution boundaries, not just better system accuracy or forced verification.
High-quality AI output triggers a metacognitive heuristic: users experience fluency as a signal of their own capability, even though they didn't generate it. This self-directed fluency illusion systematically inflates perceived competence because LLMs optimize for fluency regardless of user understanding.
A pooled analysis of three studies found a correlation of only .055 between self-reported and objective measures of AI competence, with confidence intervals including zero. This provides no basis for substituting self-assessment for demonstrated performance.
Study 1 found that greater user control over generated text raised sense of ownership, while personalizing the AI model had no impact on the AI Ghostwriter Effect.
Show all 8 sources
Users initially avoid AI partners when identity is revealed, but this preference reverses after repeated interactions with visible results. The learning mechanism—observing consistent outcomes—is essential; disclosure without feedback produces no calibration.
The Partner Modelling Questionnaire reveals that perceived competence dominates user impressions (49% of variance), followed by human-likeness (32%) and communicative flexibility (19%). This three-factor structure reflects how people evaluate dialogue partners against both functional and social standards.
Cross-linguistic research shows users in every language trust confident AI outputs even when inaccurate. While confidence expression varies by language, users everywhere track confidence signals rather than accuracy, making overconfident errors systematically followed.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- The LLM Fallacy: Misattribution in AI-Assisted Cognitive Workflows
- Evidence of a social evaluation penalty for using AI
- Humans learn to prefer trustworthy AI over human partners
- Understanding Reader Perception Shifts upon Disclosure of AI Authorship
- Penalizing Transparency? How AI Disclosure and Author Demographics Shape Human and AI Judgments About Writing
- LLM Evaluators Recognize and Favor Their Own Generations
- Assistant or Actor? Student Trust, Control, and Delegation Regret When Using a General-Purpose AI Agent
- Being honest about using AI at work makes people trust you less, research finds