# AI Evidence Check — source storyboard

RomeoApps-owned synthetic learning proof. No customer content or outcome claim.

## Brief

- Audience: professionals who use generative AI in routine knowledge work.
- Problem: confident model language can be mistaken for verified evidence.
- Target behavior: before acting on an AI-generated claim, check source, scope,
  and time.
- Duration: five to eight minutes.
- Modalities: responsive web module and SCORM 1.2 package.
- Languages demonstrated: English and Italian.

## Learning objective

Given an AI-generated business claim and its supporting source, the learner can
choose a safe next action by checking source ownership, claim scope, and
currency.

## Storyboard

1. Cold open — Present a plausible, confident answer and ask for the learner's
   first action before explaining the model.
2. Consequence — Give specific feedback that names the decision risk without
   shaming the learner.
3. Compact model — Teach three checks: source, scope, and time.
4. Worked example — Demonstrate how one claim can pass one check and fail
   another.
5. Transfer case — Present a different business situation and measure the new
   decision.
6. Job aid — Leave the learner with the three-line check.

## Accessibility decisions

- Semantic headings and native buttons.
- Keyboard-operable answer choices.
- Visible focus states and text labels for answer status.
- Feedback announced through a polite live region.
- Layouts tolerate longer translated strings.
- Motion is removed when the user prefers reduced motion.

## Evaluation plan

- Learning: baseline versus transfer-case decision accuracy.
- Behavior: sampled use of the three checks after 14 days.
- Business: avoidable AI-claim corrections per 100 reviewed outputs.

The evaluation plan is a design proposal. The prototype does not collect or
claim real learner or business results.