How to compare types of clarifying questions?

⬅️ [Inder/defer token](<./Inder_defer token.md>) | ⬆️ [Ideas](<./README.md>) | [Geocache app](<./Geocache app.md>) ➡️

Generate multiple clarifying questions for each agent to be compared.

Compute the expected value of the maximum entailment from each question of agent 1 to agent 2. Basically asking the question how often does one agent extract the same information as another agent?

Expected minimum loss is a simpler system that better captures the distribution, but is more sensitive to style differences instead of content differences.

⬅️ [Inder/defer token](<./Inder_defer token.md>) | ⬆️ [Ideas](<./README.md>) | [Geocache app](<./Geocache app.md>) ➡️