Nature Study: Causal Overclaiming is Widespread and Increasing in Social Sciences
KordingLab · x · 2026-09-01
A study published in Nature Human Behaviour investigates the prevalence and impact of overreaching causal claims in the social sciences using large language models and experimental methods.
Key Findings:
- These overclaims are common and are increasing over time.
- They affect readers' interpretation, impacting both humans and large language models.
The research quantifies a long-standing critique regarding the exaggeration of causal relationships in research lacking sufficient evidence.
More from Research
- Inria Researcher Discusses Applying Childlike Curiosity to Self-Improving Robots — pyoudeyer · 2026-09-01
- Sensori model learns health representations from 24-hour wrist movement — iScienceLuvr · 2026-09-01
- Frontier LLMs Fail at Counterfactual Reasoning: Best Score Just 64.6% — AryHHAry · 2026-09-01
- Local model benchmarks: Opus 4.8 vs. DeepSeek, Qwen, and GLM across agent & coding tests — perelmanych · 2026-09-01
- MR-JEPA: A General Purpose Video Foundation Model for Cardiac MRI — iScienceLuvr · 2026-09-01
- Nutrient open-sources 0.4B model to fix number hallucinations in documents — Div_pradeep · 2026-09-01