human first ai collaboration
This identifier was used as an evidence-review target but was never published as a canonical Brali protocol. It is retained as provenance and deliberately does not link to a public route.
Evidence Ledger · Protocol candidate · reviewed 2026-08-26
Relying on AI at work reduces self-efficacy, ownership, and meaning while active collaboration mitigates the effects
The reviewed source can support a conservative practical protocol.
Brali decision: Protocol candidate. This page summarizes a reviewed evidence boundary; it does not reproduce the source and it does not turn one paper into a universal prescription.
What the source supports
When preserving a worker's sense of authorship, competence and connection to the task matters, a meaningful human first pass before AI refinement is a reasonable workflow to test. In the randomized writing task, direct copy-and-paste use produced lower psychological ownership and meaningfulness than both no-AI and human-first collaboration, while the human-first condition was comparable to no-AI on those outcomes.
What it does not establish
- Human-first AI use is universally more productive or more accurate.
- AI-first workflows are harmful in every task.
- Using AI causes a durable loss of skill or intelligence.
- The experiment establishes the best workflow for coding, analysis, research, design or decision making.
- The correlational follow-up proves causal effects outside writing tasks.
- Any amount of editing after AI generation is equivalent to the studied human-first condition.
- Avoiding AI is psychologically superior to active collaboration.
Important limitations
- The causal experiment used short occupation-specific writing tasks rather than the full range of knowledge work.
- The passive condition was intentionally extreme: participants used AI-generated content directly without modification.
- The active condition tested one workflow only: human draft first, then AI refinement.
- The study did not instrument the external AI tool deeply enough to analyze prompt content and detailed interaction sequences.
- Baseline AI skill and confidence were not comprehensively modeled.
- The broader follow-up survey was correlational and cannot establish direction of causality.
Source context
Relying on AI at work reduces self-efficacy, ownership, and meaning while active collaboration mitigates the effects
- Source type: preregistered-randomized-experiment-plus-correlational-follow-up
- Design: Preregistered randomized experiment with 269 analyzed participants completing occupation-specific writing tasks under no-AI, copy-and-paste AI, or human-first-then-AI-refinement conditions, followed by a correlational survey of 270 workers about broader real-world AI-use patterns.
- Population: Working adults across occupations including consulting, data analysis, HR, management and marketing; the follow-up survey sampled workers in the United States and United Kingdom.
- Exposure / intervention: No AI; passive use defined as directly using AI-generated content without modification; or active collaboration defined as producing an initial human draft and then using AI to review or refine it.
- Outcomes: AI-independent self-efficacy; Psychological ownership; Work meaningfulness; Task enjoyment; Outcome satisfaction
Citation: Lee EH, Yin Y, Jia N, Wakslak CJ. Scientific Reports. 2026;16:13583.
DOI: 10.1038/s41598-026-42312-6
What this changes in Brali
Editorial note: Protocol direction: for work where ownership and independent capability matter, write the task goal, constraints, judgment criteria and a meaningful first-pass artifact before asking AI to improve it. Keep the final accept/reject judgment human. Do not market this as a performance-optimization result.
Use the boundary, not just the headline
When an AI agent or a person retrieves this decision, preserve both the supported claim and the unsupported or overstated claims. Dropping the boundary would turn a reviewed source into a stronger claim than Brali actually maintains.
All Evidence Decisions → · Research & trends → · Machine-readable review →