If a hand-coded audit finds no bias in a coder, can you trust its counts?

What it found

On 3,188 real GoEmotions comments, an audit-gated estimator (use the coder's labels if a 200-item hand-coded paired z test finds no bias, else apply a prediction-powered correction) covered the true prevalence in only 8% of draws when the coder over-counted by 2.6 points, and 27% when it under-counted by 5.1; it reached 88% only with a 12.8-point bias. At zero bias it behaved (corrects 4% of draws, covers 96%). Always applying the prediction-powered correction covered 94–97% at every bias. The coders are a naive-Bayes classifier with a moved cut-off and a word list, not LLMs; one corpus, one seed.

A figure from its own output

Paper, code and results

• README and method
• Experiments and results

More in Machine learning on qualitative data

• Does a low audit rate for coder-negatives pay when the coder has high recall?Does a low audit rate for coder-negatives pay when the coder has high recall? On 3,188 GoEmotions comments the best share falls with recall (50%, 20%, 2%), but the narrowest interval (0.115–0.119 at 200 items) only matches a random audit with no coder (0.1195). Not an LLM.
• Do hypotheses about what separates two sets of documents survive a held-out check?Do hypotheses about what separates two sets of documents survive a held-out check? On 5,313 GoEmotions comments a word-counting proposer's effects are 7.6× too large at 100 discovery comments (1.5× at 2,000); a BH check on 3,000 held-out comments confirms almost nothing with shuffled labels, but 36–59% of confirmed hypotheses are the label's own vocabulary. Not an LLM proposer.
• How much of a human emotion codebook does a codebook induced from the text alone recover?How much of a human emotion codebook does a codebook induced from the text alone recover? On 5,313 GoEmotions comments, NMF and k-means codebooks agree with the human emotion label above chance but weakly (NMI 0.26 vs 0.12 for 27 codes); the best single code per emotion reaches AUC about 0.6. Topic models, not an LLM.
• Should you hand-code the items the coder called positive, or sample the audit by the coder's label?Should you hand-code only what the coder called positive, or sample the audit by the coder's label? On 3,188 GoEmotions comments, a positives-only audit had 0 of 1,000 intervals covering the truth for every coder (it misses the coder's false negatives); stratifying by label barely helped. Not an LLM; one corpus.
• Can you audit the coder's negatives at a lower rate and weight them back?Can you audit the coder's negatives at a lower rate and weight them back? On 3,188 GoEmotions comments, weighting removes the bias of a positives-only audit but the interval widens as the negative share falls (0.126 at 50%, 0.352 at 5%, 200 items) and under-covers below ~10%. Not an LLM; coders miss half the positives.
• With six groups, how many hand-coded items does it take for every group's prevalence to be right?With six groups, how many hand-coded items does it take for every group's prevalence to be right? On 3,188 GoEmotions comments, six 95% intervals covered all six truths in only 71 to 77% of 120 to 240-item equal audits (Bonferroni: 89 to 94%); a random audit could not estimate a 1.5% group at all. Not an LLM; one corpus.

All 23 Machine learning on qualitative data projects →

Other research topics

Proper scoring & elicitation28Effort & peer prediction9Verification games & slashing19Detection & tolerances14Markets as routers13Prediction-market design9Paying contributors8Decentralised training systems23Digital twins & sim-to-real164LLM chip design loops40Robot manufacturing loops53Ambient AI assistant46Self-replicating manufacturing34Simulation-first design64Engineered clothing45Task-specific AI accelerators40Algorithmic chip-design loops42Open-source purpose-built robots44The software-engineering loop43Adversarial human imitation41Open-source residential construction41Autonomous multi-angle capture34Sensing, imaging and 3D31Energy supply and dense energy storage39Robot societies among people31Agentic co-design of low-cost robots27Laundry-hanging robots25Machine learning on qualitative data23Robots on the building site27Civilizations & group psychology1History & human sciences1

Produced with AI assistance (Claude). Results are from the code in this repository and are simulated or stylised unless the abstract says otherwise; nothing here implies affiliation with or endorsement by any lab, researcher or company cited. MIT licensed.