A stuck student will describe the problem in their own words rather than give up at an empty text box.
Falsified if Most inputs are under ten words, or the ask is opened and abandoned.
Signal ask_submitted 5 observed mean 20 words
Every screen in this journey assumes something about how students behave. A prototype that only completes the task learns nothing about whether those assumptions hold. This page is the register: what is assumed, what would prove it wrong, and what the prototype has actually observed so far.
Counts are from this prototype's own use, including this demo. They are not a result yet. They show that the measurement itself works.
Falsified if Most inputs are under ten words, or the ask is opened and abandoned.
Signal ask_submitted 5 observed mean 20 words
Falsified if Students see the tags and restart instead of continuing.
Signal tags_shown 5 observed
Falsified if Students routinely pick rank two or three, or none at all.
Signal expert_chosen 5 observed
Falsified if Decline rate climbs, or requests sit untouched.
Signal status_set 7 observed
Falsified if Unknown. Nothing here observes it yet.
Signal not measured yet, and that is the honest gap
None of these counts are strong enough yet to support a decision, and saying so is the point. The pilot that would move them is small: one cohort, four weeks, thresholds set in advance. If students write under ten words, the free-text front door is wrong and this becomes guided triage. If they routinely pick the second or third expert, the ranking is decoration and the matcher needs replacing before anything else gets built. Both outcomes are cheap to act on because both are reversible.