Nature Human Behaviour · 24 August 2026 · From issue 26
Quantifying the prevalence and impact of overreaching causal claims in social science
Cause-and-effect wording in social science studies tripled from 2000 to 2024, even without proof of causation.
Calvin Isch, Timothy Dörr et al.
Many social science studies only show that two things happen at the same time. That does not prove one thing causes the other. A check of 194,631 studies found that cause-sounding words tripled between 2000 and 2024, going from 20% to 60% of articles. In tests, AI tools that summarized these studies often made the problem worse, dropping careful wording and adding cause-and-effect claims that were not in the original.
Share of studies using cause-and-effect language in titles or summaries, from 2000 to 2024.
194,631 studies checked; AI summaries often overstated cause and effect further.
Integrity screen: passed (3 checks) Checked 25 August 2026. Retraction record: none. DOI resolves at doi.org. Metadata record found (Nature Human Behaviour). Read the source
This screen checks process signals. It does not tell you the findings are true.