Feedback that lands: criticism for autistic employees

Autistic Mirror (Outside view/Surroundings, role employer)

Feedback like "already pretty good, maybe it could still be polished a little" contains no instruction. It contains an assessment ("pretty good"), a modal particle ("maybe"), a verb without an object ("polished") and not a single concrete pointer to what should change. For someone who processes language primarily referentially, that is, by what is said rather than by what is socially implied, this sentence carries the following information: The report is good, and there is a vague possibility of improvement, without it being clear what, where or why.

Two mechanisms overlap here.

The Double Empathy Problem describes how autistic and non-autistic people enter a conversation with different communication expectations. Non-autistic communication often contains a social layer: softeners like "maybe" or "could" usually signal an expectation in this context, not an option. Autistic processing takes this linguistic softening literally. That is not a misunderstanding on one side but a structural difference in what language is supposed to do. Your employee may well have understood the sentence completely, namely that the report is good.

On top of this comes phatic communication. Softeners like "maybe" and "already pretty good" often have a relational function in non-autistic communication: they are meant to make criticism more bearable, soften hierarchy, save face. A monotropic, information-oriented attention profile processes language mainly by its content. The relational function of the softener is not automatically read along, because that would require a separate level of analysis to become active, which does not switch on automatically.

The result: the report was handed in unchanged because the information "something should be changed here" was simply not transmitted.

What transmits with this employee: "Page 3, second paragraph: the interpretation of the figures is missing. Please add it by Friday." That is not a different tone but different information.

Autistic Mirror (Outside view/Surroundings, role employer)

Rejection Sensitive Dysphoria (RSD) is the primary mechanism here. The brain does not process perceived criticism, rejection or evaluation as neutral informational input but as a bodily alarm. What this means: the signal "you were evaluated, and not entirely positively" triggers the same physiological cascade as a real threat. The HPA axis, the control loop of hypothalamus, pituitary gland and adrenal gland, releases cortisol, the heart rate rises, the body mobilises. How friendly the wording was plays a smaller role than expected, because the trigger is not the tone but the signal "evaluation". This is not an overreaction in the psychological sense but a neurobiological response that largely escapes voluntary control.

That also explains the exhaustion in the days afterwards. In many autistic people, cortisol peaks subside more slowly than in the neurotypical average. The body stays in a heightened state of activation for longer, even when the situation is long over. What looks from the outside like days of sensitivity is physiological aftermath, not an ongoing emotional reaction to the content of the conversation.

A second factor amplifies this effect as a precondition: the annual review is a high-load format. It contains many hard-to-predict elements, a formal evaluation situation, possibly new information, a different social frame than everyday work. Intolerance of uncertainty describes how a brain that works intensively with predictive models has to recalibrate constantly in an environment with many unknown variables. That costs capacity before the first word is spoken. The actual criticism then hits a system that is already running under increased load.

In addition, annual reviews often contain small, implicit evaluation signals that do not appear in everyday conversation: pauses, broken-off sentences, shifts in tone, the difference between "that was good" and a concrete example. Phatic communication, the social layers in language that signal relationship rather than information, is often analysed consciously by autistic people instead of being processed automatically. This means every one of these signals is actively decoded, which creates additional cognitive load and strains the overall system further.

For the format itself there are structural adjustments that lower the load without changing the content: send the points of criticism in writing beforehand, so the conversation no longer contains unknown variables. That gives the employee time to process the information before he is physically in the room. Concrete, behaviour-describing wording instead of evaluative adjectives ("In the last three projects the status updates were missing at the agreed times" instead of "you could communicate more proactively") reduces the interpretive effort and with it the uncertainty load. Ideally, you discuss the kind of adjustment with the employee himself beforehand, because the sensory and cognitive profile is individual.

Autistic Mirror (Outside view/Surroundings, role employer)

Autistic processing is often strongly bottom-up. This means: information is built up from the concrete detail, not downwards from an abstract overall judgement. Bottom-up processing means that details arrive first and the brain assembles an overall picture from them step by step, instead of starting the other way round with a conclusion that then looks for details as evidence. For feedback this has a direct consequence: an overall verdict like "You could be more proactive" provides no processable signal, because it contains no concrete anchor point from which the brain can build.

Effective feedback names an observable situation, the concrete action within it and what effect it had. Not "communication was sometimes unclear", but "in Tuesday's meeting the decision stayed open because it was not clear which option you prefer, and that blocked the planning round." That is not a change of style out of kindness but simply the format that is processable.

Written is structurally better than spoken. Spoken feedback runs in real time, with a parallel social channel, tone of voice, eye contact, implicit evaluations. That takes up processing capacity which is then no longer available for the content. In writing, pace and repetition can be self-determined. For complex feedback: prepare it in writing, then discuss it, not the other way round.

Monotropism is another relevant factor. Autistic attention bundles intensively on one channel instead of spreading broadly. Someone who receives several pieces of feedback at once cannot simultaneously decode the social context, place the criticism and formulate a response. One point per conversation is genuinely better than a complete list, not because the person cannot bear it, but because processing needs time to go into depth.

A common problem in standard feedback formats is mixing the evaluation of work performance with the evaluation of communication style. Directness, little small talk, factual wording are autistic communication, not a performance gap. When feedback criteria like "comes across as approachable" or "takes an active part in team dynamics" are applied to autistic employees, the expectation of masking is disguised as a performance goal. That should not appear in feedback categories.

Finally, it makes sense to discuss the format itself with the person concerned: How much advance notice do they need? Do they prefer written summaries in advance? Are there phrasings they experience as ambiguous? The person knows their own processing style best. Asking for this information directly is more efficient than any general guideline.