Skip to content
J·BJon BishopIndependent growth marketer

AI · Preregistered systematic review/meta-analysis of 106 experiments and 370 effect sizes.

When combinations of humans and AI are useful

Nature Human Behaviour. · 2024

Vaccaro, M., Almaatouq, A., Malone, T. W.

Summary

Combinations underperformed the best solo system on average; gains were stronger for content creation and when humans were better than AI.

For your marketing

Possible founder application

Benchmark combined workflows against both solo baselines.

Reported result

What the study found

Combinations underperformed the best solo system on average; gains were stronger for content creation and when humans were better than AI.

Research question

What the authors investigated

When does human–AI collaboration outperform its components?

Design

Preregistered systematic review/meta-analysis of 106 experiments and 370 effect sizes.

Sample and setting

Human-participant experiments comparing human-only, AI-only, and combined systems.

Scope and transfer

Limitations to keep in view

Publication bias and heterogeneous short tasks.

Research sources

Nature full text reports the review protocol, inclusion criteria, effect sizes, and limitations.

Additional research source ↗

Prefer to write? Send Jon a message.