Hvordan man strukturerer efterhandlingsevalueringer (AAR'er) for at fremme psykologisk tryghed, ærlig selvkritik og læring fra rodårsagerne

An after-action review (AAR) is a structured conversation held soon after a training exercise or a real event, to establish what was meant to happen, what did happen, why, and what to do next. It sounds simple. In practice, many AARs go wrong in the same two ways: people defend their decisions, and the record of the event is thin. Both make it hard to learn anything.

This guide is written for public safety training directors and training officers who facilitate AARs. It focuses on de-escalation and incident coordination, always within the authority your people have been delegated. It shows how to separate a skill gap from a climate problem, how to set up the conditions for honest discussion, how to run a worked scenario, and how to tell whether the review changed anything.

Start with the behaviour, not the lecture

Picture this illustrative scenario (not a real case). A team finishes a tabletop exercise on a tense, multi-agency incident. In the review, one officer says the radio handover "went fine". Another says the timeline was "roughly" as planned. Nobody mentions that two people disagreed about who held the next decision. The written event record has gaps, and the conversation drifts to what the exercise designers got wrong.

Before you plan a response, ask what is actually behind that silence. There are at least four different explanations, and they need different fixes:

1. A missing skill. People do not know how to describe a decision, cite what they knew at the time and say what was unclear. This is a learning need.
2. An authority gradient. Junior staff will not challenge a senior colleague's account in front of the room. Practising the skill will not help until the gradient changes.
3. Incitamenter. If review findings feed into appraisal or discipline, honest self-critique is a risk, however well it is facilitated.
4. Historie. If the last person who spoke up was ignored or sidelined, the room has learned something, and a new format will not undo it.

Psychological safety is the term researchers use for this climate. Amy Edmondson's study of 51 work teams in a manufacturing company defined it as a shared belief that the team is safe for interpersonal risk taking, and found it was associated with team learning behaviour (Edmondson, 1999, Administrative Science Quarterly). That is an association in one setting, not proof that any particular AAR format will create safety in yours. Treat it as a reason to examine the climate, not as a promise.

The practical consequence: a facilitation technique can only do so much. Psychological safety also depends on how leaders behave outside the session, so plan to look at both.

Set the conditions before anyone speaks

The US Army's training circular on AARs (TC 25-20, A Leader's Guide to After-Action Reviews, 1993) describes an AAR as a professional discussion of an event, focused on performance standards, that lets participants discover for themselves what happened, why it happened and how to sustain strengths and improve on weaknesses. Note the phrase "discover for themselves". A review that turns into a verdict delivered by the most senior person in the room is a debrief of a different kind.

Agree these conditions in advance, in writing, with the participants' supervisors:

• Purpose and boundary. State whether this review is for learning only, or whether anything said can feed a separate formal process. If the honest answer is "it depends", say that plainly. People will not be candid on the strength of a vague reassurance.
• Approved responsibilities. Be clear about what each role was authorised to decide during the exercise or event. Staff can only reflect honestly on a decision if they know whether it was theirs to make. Where information or authority was insufficient, the right behaviour is to say so and escalate, and the review should treat that as good practice.
• How leaders will respond. Rehearsing speaking up with staff is only half the job. Supervisors need a corresponding response: thanking the person, asking a follow-up question, and not commenting on the individual. Brief them first, and ask them to speak last.
• Beviser først. Gather the event record, timeline, radio or log extracts and any exercise injects before the session. Defensiveness often fills the gap that missing evidence leaves.

Design a behaviour-based learning journey

Staff do not become good at honest review by attending one. Good design starts from the behaviour you want, then matches preparation, practice and feedback to it. The Carnegie Mellon Eberly Center's guidance on aligning assessments, objectives and instructional strategies makes the point well: if you want people to analyse, practise and assess analysis, not just recall. It is design guidance for aligning learning activities, not evidence about any sector outcome or any platform.

For this task, the behaviour is: analyse what happened and agree a learning action. In Bloom's terms that sits at apply and evaluate. Those labels describe the thinking demanded, not a delivery format, and they do not explain motivation. If someone will not speak, Bloom will not tell you why.

A workable journey has three stages:

Self-paced preparation frees the live session for the part that needs people in the room: ambiguity. Facts can be learnt alone. Whether a decision made with partial information was reasonable is a conversation.

A worked scenario and debrief

Here is the case to build the practice stage around. It is illustrative, and a qualified public safety practitioner should validate the case and the answer criteria before you use it.

Sagen. In a de-escalation exercise, a response team was sent to a tense, crowded public space. The team leader asked a junior member to keep verbal contact with an agitated person while the rest of the team coordinated with a second agency by radio. Afterwards, the written record shows a 6-minute gap and no note of who confirmed the handover. In the review, the team defends the sequence and the record is not discussed.

What the facilitator asks the participant to do. Explain the decision, cite relevant evidence and identify what further information or support was needed.

A strong response sounds like this.

• Hvad jeg gjorde: "I kept talking to the person because I was the only one in verbal contact. I did not stop to log the handover."
• The evidence I had: "I could hear the radio but not the confirmation. The timeline shows my last note before the gap."
• What I did not know and what I needed: "I was unsure whether I was authorised to ask the other agency to repeat the handover. A clearer instruction on that would have helped."

How the recipient responds. The facilitator does not correct or judge. They ask: "What made that hard to do at the time?" and "What would make it easier next time?" The team leader then adds their own account, including what they could have made clearer. Hearing a senior person own a gap is often what makes the second and third contributions possible.

The transfer question. Finish with one forward-looking prompt: "What would make repeating this behaviour easier at work?" The answers are often process points (a handover confirmation line on the log, a clearer statement of who may request repeat-backs) rather than individual shortcomings. Those are your root causes.

Facilitation moves that protect honesty

A few habits make the difference between a review and an interrogation:

• Ask open "what" and "how" questions before any "why". "Why did you do that?" invites a defence. "What were you seeing at that point?" invites an account.
• Ask for the weakest points first from the facilitator and the most senior person. Modelling self-critique is faster than requesting it.
• Chase causes past the first answer. If the explanation stops at "someone made a mistake", keep asking what in the process allowed that mistake to matter. Individual error is rarely the root cause.
• Close with named actions, owners and dates. A review that ends without a learning action teaches people that reviews are theatre.

Measure behaviour and follow through

How would you know learners can now analyse what happened and agree a learning action? Look for evidence, not attendance or enjoyment:

1. Quality of the reflection. Using a rubric, score unseen cases or real reviews on whether participants state the decision, cite evidence, name uncertainty and propose a specific action.
2. Follow-up action quality. Are the agreed actions specific, owned and completed? Track them.
3. Transfer at work. Compare a baseline review with a later unseen case, after enough time in the job to practise.

Here is a sample rubric. Score each criterion 0 to 2, and adapt the criteria with a qualified specialist. This is a proposed rubric, not a validated readiness standard.

Anchors: 0 = absent or unsupported; 1 = partial, with relevant omissions; 2 = complete and justified against agreed criteria. Define critical errors separately, so that an overall score never hides a critical failure.

Be careful with one number in particular: reporting volume. If more concerns are raised after you change your reviews, that might mean people feel safer, or that more is going wrong, or that a new form is easier to fill in. A rise in reporting needs interpretation, not celebration. Pair observation of review sessions with process changes and follow-up actions, define the population, denominator and timing before you start, and check for other operational changes at the same time before attributing anything to the training. Measurement guidance such as CDC's training evaluation resources is useful for framing learning and transfer, but it does not forecast operational improvement.

Constraints that may remain

Two questions are worth carrying into your planning.

Which constraints could remain even after task performance improves? Authority gradients, appraisal links, workload and the way leaders react in the corridor afterwards can all persist. Better facilitation will not change an incentive. If people still cannot afford to be candid, your rubric will show good reflection in the training room and silence in the field.

Where this fits, and what to do next

Start with a task rubric or journey canvas: pick one recurring exercise, write down the review behaviour you want, and check that each stage of your design practises and assesses that behaviour. Then test it with a training officer and an operational specialist before running it.

AhaSlides supports interactive learning, adaptive learning and moving between live and self-paced training. If you want to see how those would handle a review like the one above, book a demo of AhaSlides and ask the team to walk through it against your own case. Confirm the specific authoring, feedback and reporting functions you need during the demo, and treat AI, VR, simulators and integrations as external or unconfirmed until you have seen them work.