Passive, copy-based use of AI erodes workers' confidence, ownership and sense of meaningful work — and some harms persist after returning to manual tasks; by contrast, active, draft-first collaboration with AI preserves worker agency and avoids these psychological costs.

Relying on AI at work reduces self-efficacy, ownership, and meaning while active collaboration mitigates the effects

Elena Hayoung Lee, Yidan Yin, Nan Jia, Cheryl Wakslak · March 15, 2026 · Scientific Reports

openalex rct medium evidence 7/10 relevance DOI Source PDF

How workers use AI matters: passive copying of AI outputs undermines self-efficacy, ownership, and perceived meaningfulness (with some persistent declines), whereas active, draft-first collaboration preserves these psychological connections at levels similar to working without AI.

Artificial intelligence (AI) promises major productivity gains, but it also raises fundamental questions about how technology can reshape people's relationship to their work. Historical debates over industrialization warned that technological change could undermine people's connection to work and sense of meaning. Similar concerns now surround AI, where the key issue may not be whether AI is used, but how it is used. Across a pre-registered experiment (N = 269) and a follow-up survey (N = 270), we examine how different modes of AI use affect the confidence individuals have in completing work without AI assistance (self-efficacy), their sense of ownership over task output, and the meaning they perceive in their work. Participants completed occupation-specific writing tasks under one of three conditions: no AI use, passive AI use (copying AI-generated content), or active collaboration (drafting first and then using AI to refin). We find that passive use undermined self-efficacy, psychological ownership, and work meaningfulness, with declines in efficacy and meaningfulness persisting even when participants returned to manual work. In contrast, collaborative AI use preserved psychological connection to the task, producing outcomes comparable to independent work. Although passive use initially boosted enjoyment and satisfaction, these benefits reversed once participants resumed manual work. A complementary real-world survey mirrored these patterns across tasks beyond writing. Together, these findings show that the psychological consequences of AI use hinge on how it is integrated into human workflows, underscoring that strategies promoting active, collaborative use may help capture AI's productivity benefits while preserving human workers' agency, competence, and connection to their work.

Summary

Main Finding

How AI is integrated into work matters for workers' psychology. Passive use of AI (copying AI-generated output) reduces workers' self-efficacy, psychological ownership of output, and perceived meaningfulness of work — and some reductions persist after returning to manual work. In contrast, active, collaborative use (drafting first and then using AI to refine) preserves these psychological connections at levels comparable to working without AI. A complementary real-world survey replicated these patterns across tasks beyond writing.

Key Points

Experimental design (pre-registered): N = 269; three conditions:
- No AI (manual work)
- Passive AI use (copying AI-generated content)
- Active collaboration (human drafts first, then uses AI to refine)
Main psychological outcomes: self-efficacy (confidence to complete tasks without AI), psychological ownership of outputs, perceived meaningfulness of work; also measured enjoyment/satisfaction.
Passive AI use:
- Undermined self-efficacy, psychological ownership, and meaningfulness.
- Initial increases in enjoyment/satisfaction were observed but reversed once participants returned to manual work.
- Declines in efficacy and meaningfulness persisted after returning to manual tasks.
Active collaborative use:
- Preserved self-efficacy, ownership, and meaningfulness at levels similar to independent (no-AI) work.
- Did not produce the same lasting psychological costs seen with passive use.
Real-world survey (N = 270) across diverse tasks reproduced the experimental pattern, suggesting external validity beyond the lab writing tasks.

Data & Methods

Pre-registered laboratory-style experiment:
- Sample: N = 269 participants.
- Task: occupation-specific writing tasks (experimental manipulation centered on how AI was used).
- Between-subjects conditions: no AI, passive copy of AI outputs, active collaboration (draft then refine with AI).
- Measured self-reported self-efficacy, psychological ownership, meaningfulness, and enjoyment/satisfaction; also measured persistence of effects after returning to manual work.
Complementary real-world survey:
- Sample: N = 270 respondents reporting on AI use across multiple task types.
- Assessed whether patterns from the experiment generalized to broader, real-world workflows.
Pre-registration increases confidence in hypothesis testing; outcomes are primarily self-reported psychological measures rather than solely objective productivity metrics.

Implications for AI Economics

Adoption effects depend on mode of use, not just presence of AI:
- Productivity gains from AI may be offset by psychological costs if firms adopt passive, copy-based workflows that erode worker efficacy and ownership.
Human capital and skill dynamics:
- Persistent declines in self-efficacy after passive AI exposure suggest potential for skill atrophy, slower reversion costs when tasks must be done without AI, and lower willingness to perform or learn tasks.
Labor-market outcomes and firm performance:
- Reduced meaningfulness and ownership can increase disengagement, turnover, and lower intrinsic motivation—affecting long-run productivity and the returns to AI investments.
Design and policy prescriptions:
- Encourage workflows that require active human contribution (e.g., draft-first, decision-making checkpoints, co-creation interfaces) to preserve worker agency.
- Invest in training and interface design that scaffold collaborative use rather than substitution.
- Measure both short-run productivity and longer-run psychological/skill outcomes when evaluating AI deployments.
Research and evaluation:
- Economic evaluations of AI should include psychological and human-capital externalities (e.g., effects on self-efficacy, skill depreciation, job satisfaction) to fully account for welfare and productivity dynamics.
Potential trade-offs:
- Firms may face short-run boosts in satisfaction or speed from passive AI use, but these can produce downstream costs when workers must perform manually or when long-term engagement matters.

(Notes: results are based on self-reported psychological measures from a pre-registered experiment and a corroborating survey; further field and longitudinal work is needed to quantify long-run productivity and labor-market impacts.)

Assessment

Paper Typerct Evidence Strengthmedium — Strong internal validity for short-term causal effects on self-reported psychological outcomes due to random assignment and pre-registration, and partial external validation via a separate survey; limited by reliance on self-reported measures rather than objective productivity or long-run labor outcomes, modest sample size, and short follow-up. Methods Rigormedium — Design strengths include pre-registration, randomization across three clearly defined conditions, and replication in a complementary survey; limitations include reliance on self-reported psychological scales, short-term measurement of persistence, unspecified representativeness of samples, and lack of field/behavioral productivity measures. SamplePre-registered lab-style experiment N = 269 participants completing occupation-specific writing tasks randomized to no-AI, passive-copy AI, or active collaboration (draft then AI-refine) conditions; complementary real-world cross-sectional survey with N = 270 respondents reporting on AI use across diverse tasks. Themeshuman_ai_collab productivity skills_training adoption org_design IdentificationRandomized controlled experiment with pre-registered between-subjects assignment to three conditions (no-AI, passive copy of AI output, active human-first collaboration with AI); complementary cross-sectional survey (N=270) used to replicate pattern in real-world task reports. GeneralizabilityLaboratory writing tasks may not represent the range of real-world job tasks (limited task types)., Evidence is short-term; persistence of effects beyond brief follow-up is unknown., Primary outcomes are self-reported psychological measures, not objective productivity, wages, or employment outcomes., Sample representativeness unclear (likely convenience/online sample), limiting population-level inference., AI model(s), interface, and specific prompt/workflow details may affect results and may not match enterprise deployments or domain-specific tools.

Claims (15)

Claim	Direction	Confidence	Outcome	Details
Passive use of AI (copying AI-generated output) reduces workers' self-efficacy. Worker Satisfaction	negative	high	self-efficacy (confidence to complete tasks without AI)	n=269 0.6
Passive use of AI reduces psychological ownership of the produced outputs. Worker Satisfaction	negative	high	psychological ownership of outputs	n=269 0.6
Passive use of AI reduces perceived meaningfulness of work. Worker Satisfaction	negative	high	perceived meaningfulness of work	n=269 0.6
Some declines (in self-efficacy and meaningfulness) from passive AI use persist after participants return to manual work. Worker Satisfaction	negative	high	self-efficacy; perceived meaningfulness (measured post-return to manual work)	n=269 0.6
Passive AI use produced an initial increase in enjoyment/satisfaction that reversed once participants returned to manual work. Worker Satisfaction	mixed	medium	enjoyment/satisfaction	n=269 0.36
Active, collaborative AI use (human drafts first, then uses AI to refine) preserves self-efficacy at levels comparable to independent (no-AI) work. Worker Satisfaction	null_result	high	self-efficacy (confidence to complete tasks without AI)	n=269 0.6
Active, collaborative AI use preserves psychological ownership of outputs at levels comparable to independent work. Worker Satisfaction	null_result	high	psychological ownership of outputs	n=269 0.6
Active, collaborative AI use preserves perceived meaningfulness of work at levels comparable to independent work and does not produce the lasting psychological costs seen with passive use. Worker Satisfaction	null_result	high	perceived meaningfulness of work (including post-return)	n=269 0.6
A complementary real-world survey (N = 270) across diverse tasks reproduced the experimental pattern, suggesting external validity beyond the lab writing tasks. Worker Satisfaction	positive	medium	self-reported relationships between AI-use mode and psychological outcomes (self-efficacy, ownership, meaningfulness) across real-world tasks	n=270 0.36
The experiment was pre-registered, used occupation-specific writing tasks, and employed a between-subjects design with three conditions (No-AI, Passive AI, Active collaboration). Other	null_result	high	n/a (methodological claim)	n=269 0.6
Outcomes reported are primarily self-reported psychological measures rather than objective productivity metrics. Other	null_result	high	measurement type (self-reported psychological outcomes)	n=269 0.6
Firms that adopt passive, copy-based AI workflows risk psychological costs that could offset short-run productivity gains from AI. Firm Productivity	negative	speculative	inferred organizational outcomes (productivity offsets, not directly measured)	0.06
Persistent declines in self-efficacy after passive AI exposure suggest potential for skill atrophy and slower reversion when tasks must be performed without AI. Skill Obsolescence	negative	speculative	inferred human-capital outcomes (skill atrophy, reversion costs; not directly measured)	0.06
Design and policy interventions that encourage active human contributions (e.g., draft-first workflows, co-creation interfaces, training) can help preserve worker agency and mitigate psychological costs. Training Effectiveness	positive	medium	inferred mitigation of psychological harms (not directly measured at firm scale)	0.36
Economic evaluations of AI adoption should include psychological and human-capital externalities (effects on self-efficacy, skill depreciation, job satisfaction) to fully account for welfare and productivity dynamics. Other	positive	speculative	recommended evaluation scope (inclusion of psychological/human-capital measures)	0.06