How to Detect Bias in Large Language Models

Research from Wharton's Sonny Tambe finds that LLMs can make biased hiring decisions that traditional auditing methods might not be able to catch.

KEY TAKEAWAYS

• LLMs trained on vast swaths of online data can absorb and replicate human biases.

• The direction of these biases is not always predictable.

• Policymakers and organizations need context-specific audits to understand how these models actually perform in the real world.