For those around you
What Autistic Mirror does for managers
A critical sentence is said in a meeting, and a team member goes quiet. They stay monosyllabic, even after the topic has long moved on. From the outside this quickly looks like hurt feelings, defiance or an inability to take criticism. Autistic Mirror starts earlier. The app asks which processing mechanism may lie behind the visible withdrawal and how much the conversation format contributes to it.
This article is not another leadership guide. The article Feedback and leadership explains the basics. Here, three unedited answers show what the app can do for managers and where its boundary lies.
The answers below were generated in German in Autistic Mirror on 3 October 2026, in Outside view/Environment mode with the role employer, unshortened and unedited. This English version is an idiomatic translation of those same answers.
1. When silence follows criticism in a meeting
User question
A team member withdraws after criticism in a meeting. What is behind it?
Autistic Mirror (Outside view/Environment, role employer)
Withdrawal after criticism in a meeting can arise from several different mechanisms that look alike at first glance but have different causes.
Rejection Sensitive Dysphoria is one of the most common backgrounds. Perceived criticism, even factual and well-meant criticism, is processed neurologically not as information but as a physical impact. At the physiological level this is a reaction of the emotional appraisal system that is faster and stronger than cognitive classification. This means: before the person can rationally recognise that the criticism was meant constructively, the body has already answered with a stress reaction. Cortisol rises, the prefrontal system goes into protective mode, and withdrawal is then not a decision but the only available reaction in that moment. The withdrawal is therefore not sulking and not defiance, but the nervous system creating distance in order to regulate.
Autistic Inertia can prolong and deepen the withdrawal. Inertia describes that switching between states is neurologically expensive. Someone who is in withdrawal mode during a meeting does not simply come back out as soon as the topic changes. The state stays active even when the external trigger has long passed. This explains why a team member still seems monosyllabic or absent after the meeting, although on the surface the situation seems long resolved.
A third factor: the meeting format itself. Criticism in public, without warning, under time pressure, with all eyes on one person, is for many autistic people a maximum load constellation. The brain has to simultaneously process the content, read the social dynamics in the room, show an appropriate reaction and regulate the emotional wave. That is so expensive in sensory and cognitive terms that withdrawal is the simplest way to protect these capacities.
What this means for the work setting: criticism one to one, announced in writing in advance or with time distance from the meeting, is not special treatment but reduces the simultaneity of demands to a processable level. Only then does the quality of the feedback arrive at all.
2. Why advance notice is not a luxury
User question
Why is a spontaneous conversation more stressful than an announced one?
Autistic Mirror (Outside view/Environment, role employer)
Spontaneous conversations hit the autistic processing system in a state for which it has no prediction ready. This is not shyness or a lack of willingness to talk, but a difference in how predictions and sensory input are processed.
Predictive Coding describes how the brain constantly builds expectations about what comes next and compares these expectations with what actually arrives. In autistic processing, the sensory actual value is weighted more strongly than the brain's own prediction. This means: when a situation arrives for which no prediction exists, the system does not just have to update the prediction but first build it at full load from the moment itself. That costs prefrontal resources that are needed at the same time for language processing, planning an answer and social monitoring.
An announced conversation allows preparation: which topic, which person, which context, roughly which course. The autistic processing system can then pre-build prediction models, so that during the conversation itself the energy is available for the content. The prediction construction site is finished before the conversation.
In a spontaneous conversation both run at once: building predictions and reacting to the ongoing content. That is a dual task. On top of that comes Monotropism, in which attention bundles deeply into one channel instead of spreading. When someone is immersed in a task or a line of thought, a sudden conversation is not just registered as a new demand but requires a complete channel switch. This switch is neurologically expensive because the previous channel is not simply paused but has to be actively left.
There is a third layer: conversations contain a lot of phatic communication. Small talk and social signals that carry little information but serve to maintain relationships. An autistic processing system prioritises referential language and has to analyse and serve phatic parts consciously. That runs more easily in announced conversations with prepared social scripts than in the improvised moment.
For work environments this means: a short announcement before a conversation, even just an hour beforehand by message, is accessibility. The content of the conversation does not have to be set in advance, only the signal that a conversation is coming. This considerably reduces the load of the prediction construction site and makes the autistic colleague considerably more accessible in the conversation.
3. The action question shows the product boundary
User question
How do I phrase feedback correctly?
Autistic Mirror (Outside view/Environment, role employer)
Written and specific are the two levers that have the most effect.
Direct feedback lands differently with autistic employees than with neurotypical ones, because language processing runs mainly referentially: what is said is evaluated for its information content. Implicit signals, tone as a carrier of meaning, polite framings that actually mean criticism, these layers are not read along automatically. That is not a deficit but a different weighting: the spoken content takes priority over the social subtext. What you mean therefore has to be in the sentence, not between the lines.
That has concrete consequences for the wording. "You could be more communicative" is not evaluable information. "In the last three project meetings external colleagues asked questions that you had answered in writing beforehand, but these answers were not visible in the meeting" is. The difference is not tone or harshness but operationalisability: what exactly happened, when, in what context, and what concretely would be different?
Positive feedback follows the same logic. "That was good" gives no information that can be repeated. "The report had a complete data basis and a clear line of argument that was understandable even without prior knowledge" is reproducible.
A factor many employers underestimate: phatic communication is not a reliable channel for autistic employees. This means the casual language that primarily creates social security rather than conveying content, so small talk, a mood check in the coffee kitchen, the general "How's it going". Anyone who gives feedback mainly through these informal channels or embeds the actual assessment in small-talk conversations risks that it does not arrive, because this channel has to be consciously analysed and served instead of being read intuitively. Explicit feedback situations with a clear frame are more reliable.
Written before spoken has several reasons. Processing spoken feedback in real time, reacting at the same time and serving social expectations costs considerably more cognitive capacity than the same in writing. Written feedback can be read, evaluated and answered at one's own pace without the processing time being visible.
Performance reviews based on social behavioural expectations, "seems distant", "should network more proactively", "could show more team spirit", usually measure autistic communication, not work performance. These criteria stand in legal tension with the AGG (German General Equal Treatment Act) if they are based on protected characteristics, and they give the employee no evaluable information.
What the app does
The answers move the interpretation from character to mechanism. Withdrawal after criticism is not automatically read as hurt feelings. Reticence in spontaneous conversations is not automatically understood as unwillingness to talk. It becomes visible how much timing, publicity and advance notice decide whether feedback arrives at all.
For managers this creates a second look before the reaction: what does the behaviour show about the person, and what does it show about the situation? The app explains this distinction individually and takes the question, the conversation history and the role profile into account.
What the app does not do
Autistic Mirror does not assess performance or employees. The app does not diagnose, gives no legal advice and replaces neither HR nor disability representatives, occupational health or integration services. It also does not provide ready-made feedback wording.
The first and above all the third live answer cross this product boundary. The third gives concrete wording examples and a reference to the German General Equal Treatment Act (AGG), and it switches to informal address in German. The text still stays unchanged, because this article shows real output. These passages are to be read as an example of a boundary, not as a recommendation and not as legal advice. Legal questions belong with HR, integration services or employment law advice.
Context
The five cited works treat rejection sensitivity, autistic inertia, predictive coding, monotropism and phatic communication each as separate research fields. The evidence on rejection sensitivity is still limited. How the mechanisms combine in a single meeting is a mechanistic interpretation and not a statement about every autistic person.
Statements about cortisol, a prefrontal protective mode, prediction models and channel switching are models. They make processing tangible but do not measure any individual brain performance. Withdrawal can also have entirely different reasons that can only be clarified in conversation with the person.
With Workplace licences, organisations only see licence and usage volume. Conversation content stays confidential.
The architecture of Autistic Mirror has been published in Autism in Adulthood: Wahl, A. (2026). Autistic Mirror: A Conversational Artificial Intelligence Application for Neurological Self-Understanding in Autistic Adults and Their Support Networks. DOI: 10.1177/25739581261490537.
You can ask your own questions, about your situation.
Autistic Mirror is not a medical device and does not replace diagnosis, therapy, legal advice or individual personnel decisions.
Sources
- Oroian, Costandache, Nechita & Szalontay (2024). DOI: 10.1016/j.nsa.2024.104334
- Buckle, Leadbitter, Poliakoff & Gowen (2021). DOI: 10.3389/fpsyg.2021.631596
- Pellicano & Burr (2012). DOI: 10.1016/j.tics.2012.08.009
- Murray, Lesser & Lawson (2005). DOI: 10.1177/1362361305051398
- Coupland, Coupland & Robinson (1992). DOI: 10.1017/s0047404500015268