Evaluate and Respond to Poor Generative AI Output

Describe a project in which you used generative AI and discovered that its output was not good enough for the intended use. Explain how you recognized the problem, what you did about it, and how you assessed the result. ### Constraints and Clarifying Questions - Use an actual project and identify your personal contribution. - Explain what acceptable output meant for that task; do not assume that fluent output was correct. - Discuss the project's evaluation metrics. If those metrics came from a different project, clearly identify that example and explain the connection. - Separate measured improvements from qualitative observations and untested expectations. ```hint Establish the acceptance standard Identify the evidence that made you question the output, then explain how you checked whether your response addressed that same problem. ``` ### What a Strong Answer Covers - The intended use of the generated output and a concrete quality failure. - How the issue was detected and investigated, with ownership of the response. - A justified corrective action and evidence about the resulting quality. - Meaningful project metrics, their limitations, and what they changed about the decision. ### Follow-up Questions - Could the chosen metric improve while the output remained unsuitable for users? - How did you decide whether to revise the AI workflow or stop using it for that task? - What did you change about future evaluation after this incident?