Evaluate chatbot replies using Grice's quantity, quality, relevance, and manner maxims.
Grice gives you a fast reply audit. For each chatbot answer, ask four questions before shipping. Quantity Did we give exactly enough for the user's current decision? Extra policy details can be useful later but harmful in the first answer. Quality Can we support the answer with the retrieved source, account data, or known rule? If not, the bot should disclose uncertainty or ask for missing context. Relevance Does the answer address the user's question as asked, or did it drift into the closest FAQ article? Relevance failures are common when bots over-index on keyword match. Manner Is the answer…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in