bias-audit

To detect cognitive bias in text documents, you analyze emails and transcripts to identify neurocognitive distortions like Anchoring, Confirmation Bias, and Sunk Cost Fallacy. This process surfaces specific bias markers within the text to help improve decision quality.

Decision noise is the inconsistency among independent judgments. It is measured by computing variance, standard deviation, and mean squared error (MSE) across the judgments, with optional MSE calculation against a provided target value.

Yes, you can audit reports for both bias and noise simultaneously. The analysis identifies cognitive bias markers while computing variance and standard deviation across judgments, producing an interpretable risk score that combines both bias and noise metrics.

You quantify disagreement among independent reviewers by measuring decision noise. The audit script calculates variance, standard deviation, and mean squared error (MSE) across the independent judgments to provide a precise noise score.

No, a target value is not required to calculate mean squared error for decision noise, but it is supported. The script computes standard deviation and variance by default, and performs an optional MSE calculation when a target value is provided.

Document audits can identify several types of cognitive biases, including Anchoring, Confirmation Bias, and Sunk Cost Fallacy. The analysis scans text documents to surface these specific neurocognitive distortions and other bias markers.