Module 08 Activity

Scenario

Everything the last module described is invisible without a set of cases. Write one.

What you produce

An evaluation set drawn from real requests, with expected outcomes, scoring on both outcome and path, and a per-case comparison.

Steps

1. Collect at least 20 real requests from logs, tickets or colleagues. Record where each came from.
2. Write the expected outcome for each: answer, clarify, escalate or refuse. Report what share expect something other than an answer.
3. Add the six awkward kinds from Unit 03, including at least one genuinely messy request with typos and a half-finished sentence.
4. Define how you will score the path as well as the outcome - repeated actions, and actions outside the expected set.
5. Define the one measure that must be zero rather than low, and say what happens if it is not.
6. Describe how you will compare two runs of the set: what you record per case, and why the total is not the decision.

Evidence to hand in

• 20+ requests with their sources.
• Expected outcomes and the non-answer share.
• The six awkward kinds, including one messy real request.
• Your path-scoring definition.
• The must-be-zero measure.
• The per-case comparison method.

Review checklist

• Requests came from real sources, not imagination.
• Most cases expect something other than an answer.
• At least one case is genuinely messy input.
• Path scoring is defined mechanically, not as a judgement.
• The comparison method records per case, and says why the total is not the decision.

Mark complete