Feedback Instructions: criteria, grading & the Feedback Chat

The Feedback Instructions section defines how Cesura assesses the learner: the criteria it grades against and the report it writes, plus the optional AI feedback chat that talks the learner through the results. Two pages.

Feedback Report

Where you define what the learner is graded on and how the report reads. The Feedback Report (n) badge counts your criteria.

Feedback criteria

Click Add criterion for each thing you assess. Each criterion has three parts:

• Feedback Topic — a 2–4 word label describing a behaviour, not a trait. Prefer "Open clarifying questions" over "Communication", or "Reflecting emotions" over "Empathy".
• Feedback Instructions — the rubric the AI scores by. The most reliable scores come from a clear structure: what the criterion measures and why; what excellent performance looks like; a 0–100 scale broken into the same bands the report uses (see Grade bands below); and any conditions that force a failing score. The more specific the rubric, the more consistent the score is across runs.
• Judged on — what the criterion is graded against. This setting appears only when the simulation has a briefing or asks for a debrief. See Judge a criterion on the briefing or the debrief below.

Aim for a focused set — around 2–5 criteria. Too many leave learners unsure what to improve.

Hide a criterion from the learner

Each criterion has an eye toggle. Turn it off and the criterion is marked Hidden from learner: it is still evaluated, and you and your instructors still see its score in the report — but the learner never sees it, in their report, their PDF, the feedback chat, or the grade passed back to your LMS.

• A hidden criterion is excluded from the overall grade, so the number the learner can add up from their own rows matches the one you see. (If the simulation uses the AI-generated overall grade rather than a weighted average, the grade still reflects the hidden criterion — switch to a weighted average if it must not.)
• The AI is instructed to keep a hidden criterion out of the written summary, but that instruction is not a guarantee.
• At least one criterion must stay visible — otherwise the learner's report would show no feedback at all.

Use it for things you assess but do not want to hand back mid-course — an integrity check, an instructor-only observation, or a criterion you are still calibrating.

Overall Grade Method

• AI Generated — the AI decides the overall grade holistically. This is the default for a new simulation.
• Weighted Average — each criterion gets a Weight and the overall grade is their weighted average. Use Normalize to 100% to balance the weights; the page shows a running Total and confirms "Weights balanced".

You can choose either method whatever the simulation's purpose. A saved simulation keeps the method it has.

Feedback Report Instructions

Report-level guidance for the tone and philosophy of the written feedback — for example: empowering rather than judgmental, one key strength plus one thing to improve, quote specifics from the conversation, and one concrete recommendation for next time. This shapes how every criterion's feedback reads.

Judge a criterion on the briefing or the debrief

When the simulation has a briefing or asks for a debrief, each criterion has a Judged on setting: a checkbox for each thing the simulation collects.

• The learner's briefing — the answers and document the learner gave before the conversation. Offered when the simulation has a briefing.
• The conversation — the transcript.
• The learner's debrief — what the learner wrote after the conversation. Offered when the simulation asks for a debrief.

Tick one or more. With more than one ticked, the evaluator also names where they disagree, for example a gap between what the learner did and what they wrote. A part the learner was asked for and did not give is not held against them. The last ticked box cannot be cleared. By default a criterion is judged on the conversation and the debrief; the briefing counts only where you tick it.

A criterion whose ticked boxes differ from the default shows them as a chip in its header, for example Conversation only, Debrief or Briefing + Conversation. The chip names only the boxes ticked on the card. In a simulation with neither a briefing nor a debrief, Judged on is hidden and every criterion is graded on the conversation.

A criterion judged only on material the run does not have scores 0, with a note saying so: for example, a criterion judged on The learner's debrief alone when there is no write-up to grade. If the simulation doesn't collect anything a criterion is judged on, the editor warns you on the criterion that every learner scores 0 on it, and keeps Judged on visible so you can fix it: tick something the simulation collects, or turn on what the criterion needs.

Grade bands

Cesura grades every criterion on the same five bands the report shows:

Score	Band in the report
90–100	Excellent
80–89	Very Good
60–79	Good
30–59	Further Attempt Required
0–29	Not Satisfactory

The pass mark is 60. If your rubric names different ranges, the grader maps them onto these bands and reads your ranges as descriptions of each level, so you don't have to rewrite older rubrics.

Note: Grades on existing simulations can differ from earlier runs of the same simulation, because the grader and the report used to apply different bands.

How the written feedback reads

The feedback in the report speaks to the learner, by the name they gave: it describes what they did and what to try next, rather than justifying the grade. In a simulation with the Screening purpose, the report is instead written about the candidate, for your organization.

Every completed run gets a feedback report, including runs of simulations with the Feedback Chat turned off.

Feedback Chat

The optional AI conversation (the "evaluator") that discusses the results with the learner after the report. This page defines who the feedback chat assistant is and how it should run:

• Enable Feedback Chat Phase — turn the feedback chat on or off. (You can also toggle this on Feedback Chat Intro.)
• Evaluator Name, Profession, Gender — the assistant's identity. Give it a credible, specific persona, for example "Dr. Michal Cohen, organizational psychologist".
• Feedback Chat Instructions — how the feedback chat should run: its role, the 2–3 points to dig into, its style (non-judgmental, curious, open questions), the conversation structure, and a safety protocol for emotional distress.

The Exam and Screening purposes turn the Feedback Chat off and lock it. See Choose a simulation's purpose.

The feedback chat's opening line and its Minimum rounds are set on the Feedback Chat Intro page, in the Simulation Display section.

In this series

1. The simulation editor, explained
2. Create a simulation with AI
3. General settings
4. Simulation Display
5. Simulation Instructions
6. Feedback Instructions (you are here)
7. The debrief
8. Formatting content with Markdown
9. Best practices & troubleshooting
10. Set up a simulation as an exam
11. Enable & configure the Realtime Guide