Evals for AI Engineers

Chapter 4. Collaborative Evaluation Practices

A Note for Early Release Readers

With Early Release ebooks, you get books in their earliest form—the author’s raw and unedited content as they write—so you can take advantage of these technologies long before the official release of these titles.

This will be the 4th chapter of the final book. If you’d like to be actively involved in reviewing and commenting on this draft, please reach out to the editor at [e-mail].

In the previous chapter (Chapter 3), we walked through error analysis: reading traces, identifying failure modes, and building a taxonomy of how your system goes wrong. That process depends on human judgment at every step. You decide what counts as a failure. You decide how to categorize errors. You decide whether a trace is acceptable or not.

But what happens when your judgment is not enough?

Sometimes the evaluation criteria are inherently subjective. What one person considers a “helpful” response, another might find verbose or off-topic. A tone that feels professional to an engineer might strike a customer service lead as cold. When you are evaluating qualities like clarity, empathy, or appropriateness, there is no objective ground truth to fall back on.

Or, sometimes an expert is required to make quality judgements. In complex domains—​legal document review, medical advice, financial analysis—​no one person has the expertise to catch every type of error. A software engineer might miss technical inaccuracies ...