Do What We Say, Not What We Do (PDF) Preamble: AI Bias and the WEIRD Inheritance AI systems are biased. This is not speculation. This is measured, published, and peer-reviewed. In 2010, researchers at Harvard documented that 96% of subjects in top psychology journals came from Western industrialized nations, which house just 12% of the […]
bias
The Case for AI Provider Plurality in Evidence-Based Research
This white paper by Basil C. Puglisi tests whether multi-AI provider plurality, with human arbitration, can correct the WEIRD bias that AI inherits from its training data. Family structure and the division of caregiving and earning serve as the test case. ChatGPT refused to engage the evidence, Perplexity researched it, and Claude confirmed the synthesis.