Skip to main content

Testing and publishing a Role Play

Always try a role-play as a learner would before you publish it. The test view shows the conversation, the live evaluation, and the role-play details.

Testing and publishing a Role Play

Always try a role-play as a learner would before you publish it. The test view shows the conversation, the live evaluation, and the role-play details.

This is step three of the flow, after designing the role-play with the AI Designer.

Test before publishing

  1. In the designer, once a preview exists, click Test role play. You can also test later from the role-play details page using Preview.

  2. Read the Briefing that sets up the scenario, then click Got it to start.

  3. Type a reply in "Reply as the learner..." and continue the conversation. The header shows the current Stage, the criteria count and the Turn number.

  4. Use Restart to start over, or Retry after a finished session.

When the session finishes, the result appears under two tabs — Evaluation and Conversation.

Reading the evaluation

  • An overall result — Passed or Did not pass — with the score as a percentage and as points (for example, 8 / 10).

  • A breakdown per criterion, each with its score and a short feedback note.

  • An overall Feedback summary, plus the session length and number of turns.

Use the test to check the persona's tone, the difficulty, and whether the rubric rewards the behaviour you actually want.

Publish

When you are happy with the draft, click Publish. The role-play status changes to Live and it becomes available to learners inside the Learn Content item — see The learner experience: practicing a Role Play.

Statuses and deleting

  • Draft — created but not yet available to learners.

  • Live — published and available.

  • Archived — removed from the application. Deleting a role-play archives it: this removes the role-play and all its sessions, and cannot be undone.

Once learners start running it, track results in Reviewing Role Play sessions, evaluations and feedback.

Did this answer your question?