Annual review: why good work can do badly in the conversation

Autistic Mirror (outside view, role employer)

In their standard form, annual reviews mostly measure skills that structurally conflict with autistic neurology, not the work performance itself.

Monotropism is the central mechanism here. Autistic attention bundles intensely on a few channels instead of spreading out. That means deep specialist expertise, often above-average quality of results, but at the same time little resource for the parallel attention an annual review demands. An annual review is not a factual conversation. It is a complex social format with implicit expectations, self-presentation, emotional subtext and simultaneous processing of body language, tone of voice and conversational dynamics. In a monotropic system, this parallel processing costs a disproportionate amount of capacity.

On top of that comes predictive coding. The brain constantly compares expectations with what actually happens and raises the alarm when predictions do not hold. In autistic people this system is often coupled with a lower tolerance of uncertainty. An annual review is structurally unpredictable: topics change spontaneously, the assessment criteria are rarely fully transparent, and the manager's reaction to answers is hard to anticipate. This prediction deficit activates stress responses that impair verbal processing and the retrieval of information. The person knows more than they can put into words in that moment.

Masking often runs in the background as an invisible permanent load. Masking means actively suppressing autistic communication and behaviour in order to meet neurotypical expectations, which continuously ties up prefrontal capacity. In the annual review this pressure jumps sharply. It is at once an evaluation context and a social context in which eye contact, self-presentation and affective mirroring are read as competence. These resources are then missing for the actual task: presenting one's own contribution clearly and convincingly.

The gap between performance and review outcome is therefore not a sign of a lack of self-reflection or ambition. It is a sign that the review format contains barriers that do not make the actual performance visible.

In concrete terms, the format can be made less of a barrier: communicate topics and assessment criteria in writing beforehand so that no surprises arise. Offer the option of submitting evidence of performance in writing before, or instead of, reporting on it orally. Keep the conversation shorter and more focused rather than bundling many topics into one format. Ask direct, concrete questions instead of open invitations to self-presentation. These adjustments improve how meaningful the review is for everyone involved, not only for autistic team members.

Autistic Mirror (outside view, role employer)

Autistic processing tends towards a literal interpretation of language: a sentence is checked for its information content, not for the social subtext it is meant to carry. "Show more presence" is not an instruction that can be carried out, because the term has no reference. Presence for what, in which situation, measured against which behaviour? For a neurotypical colleague, implicit convention fills the gap automatically: she has stored hundreds of social contexts in which "presence" takes a particular form as a concept. This automatic filling-in does not run along in autistic processing.

This is linked to predictive coding. The autistic brain weights sensory and linguistic input more strongly against its own predictions, instead of routinely filling gaps with contextual knowledge. The same happens with formulations of social expectations: the system waits for a reference that is not present in the sentence, and finds none. Not because understanding is missing, but because the sentence is semantically underdetermined. What counts as feedback is in reality a hint.

On top of this comes monotropism. Attention bundles in a tunnel on one channel, with great depth in return. That means the team member has probably worked on their tasks intensively, focused on what matters in terms of content. Formal presentation, visibility, casual networking in corridor conversations do not run in parallel, because that is a different channel. "Be more proactive" names the missing parallel channel without explaining which concrete behaviour is expected.

The double empathy problem sits on a third level. The manager gives feedback that is completely clear to them, because they were socialised in a communication culture in which these phrases function as instructions. The autistic team member sits in a different processing style and hears the same sentence as an empty signal. Both leave the conversation with the impression of having communicated. In fact it was a one-way street.

Concrete feedback works here: not "more presence", but "In team meetings I would expect you to share assessments of ongoing projects without me explicitly asking, for example once a week." That can be carried out, because it has a reference.

Autistic Mirror (outside view, role employer)

What happens in the annual review is a classic overlap of several mechanisms that reinforce each other.

Masking runs at full speed in an annual review. Autistic reactions that could come across as socially undesirable are suppressed prefrontally: regulating facial expression, dosing eye contact, controlling tone of voice, decoding ambiguous signals from the other person in real time. This runs in parallel with the actual content of the conversation and permanently costs cognitive capacity that is then missing for the actual self-presentation. From your side as the person leading the conversation, this looks like tension or reserve. What is actually happening is a dual task under conditions that are none for neurotypical conversation partners.

Intolerance of uncertainty adds to this. An annual review is structurally unpredictable: which questions will come, in which order, how the reaction turns out, what the consequences are. A system that calibrates particularly demandingly through predictive coding has to keep recalculating when predictability is missing. That is energetically expensive. What is unannounced costs more than what is unpleasant itself, because the resources flow into correcting expectations instead of into the answer.

There is also a specific problem with self-presentation: many autistic people find it difficult to formulate their own achievements as a coherent narrative, not because the achievements are missing, but because this kind of representation needs a different processing route. Monotropism means that attention focuses on concrete tasks, not on meta levels such as "what does my work mean strategically for how the team perceives it". This self-reflective framing does not automatically run along while the work is being experienced. Producing on demand in the conversation what was never stored as a narrative is an additional real-time achievement.

Sensory filter exhaustion completes the picture. If the conversation takes place in the afternoon or a demanding day lies behind it, filter capacity is already lower anyway. Sensory details in the room (lighting, pitch of voice, background noise) then arrive less dampened, which further lowers the capacity available for language processing.

What this means in practice: the conversation does not evaluate the employee, but the format. A written advance document with concrete questions that she can answer calmly before the conversation creates a different starting point. Not because she "should" be prepared, but because the format then becomes more accessible, the results can be retrieved more reliably, and the conversation itself takes place under less of a double load.