Research question · Not an established conclusion
Compare task types, complementary errors, and the design of reliance—not just average gains.
Public database records · This page may be cached; not live monitoring or human validation.
Adding AI does not establish that a combined system beats either contributor. The comparison must use the same task and evaluation criteria, against the better standalone performer—not only a human baseline.
Scope: complementary performance, correlated errors, appropriate reliance, and task type. The records include experimental and classification settings; they do not represent every form of long-term organizational collaboration.
The relationships below come from an existing public claim, not title similarity. Summaries and relationship text are HIE record interpretations, not verbatim author statements.
Qualifies
Across 106 experiments, combinations were worse than the better standalone agent on average, with task type explaining important variation.
Supports
Randomized access to ChatGPT improved speed and quality on bounded professional writing tasks, especially for lower-skill participants.
Supports
AI assistance improved productivity in a specific customer-support setting, with heterogeneous effects by worker experience.
Qualifies
Explanations increased acceptance of AI advice but did not increase complementary team performance.
Connected proposition · Separate from question progress
Human-AI collaboration does not reliably outperform the better standalone human or AI; results depend strongly on the task and interaction design. →Claim status: Contested · No human review
HIE editorial interpretation · Not a new empirical finding
The records offer different lenses: task type may shape gains, complementary prediction errors may help, and explanations alone do not guarantee appropriate reliance. These are not interchangeable universal laws.
Still unresolved: do effects transfer across populations, models, and time? What division of work reduces shared errors? Which apparent gains reflect changed evaluation criteria?
Verification references: author-institution pages and original papers. Checking these references does not constitute human scientific review.
Proposed hypothesis
Hypothesis: explicit delegation and calibrated reliance could improve combined performance on some tasks. A proposed comparison uses human-only, AI-only, unrestricted collaboration, and bounded delegation with fixed tasks, scoring, model, and sample. Track errors and costs. No recruitment or execution has occurred.
Support would require beating the better standalone baseline and independent replication. No improvement, gains explained only by extra resources, or offsetting risks and costs would challenge or qualify the hypothesis. Preregistration and authorization must precede execution.
Notes & experiments: no executed results yet. Counterevidence, replication designs, and authorized resources can be proposed; no participation is implied.
Participation & contribution boundaries →