Architecture

Why every mechanism carries a source

Statements about autistic neurology are only verifiable when the source is directly linked. Without a link, a statement remains a claim. With a link, it becomes an anchor that any reader can follow back to the original study.

From this observation a structural change to Autistic Mirror emerged. Blog articles that cover a mechanism now end with a collapsible "Sources" block. When a validated original publication exists for the mechanism, it is linked there, with authors, year and DOI.

How the source block is built

Behind the block sits a curated lexicon in the platform's code. Every source passes three checks before it is admitted. First, the publication that describes the mechanism must be named, if possible the first description. Second, the DOI must resolve and point to exactly that publication, issued by a scientific publisher with peer review. Third, the source abstract must cover the statement that the article cites. Only when all three checks pass does the entry receive that source.

Current state: 97 entries. 89 of them carry a peer-reviewed source. They cover the mechanisms that appear most often in the app and on the blog, such as monotropism, predictive coding, masking, autistic burnout, double empathy, interoception, time blindness and polyvagal theory. They also include co-occurring conditions with their own cause, such as EDS, POTS, epilepsy and Tourette, and the intersectional axes for which robust studies exist.

Eight entries deliberately carry no source. They are terms from the autistic community, such as sensory processing fatigue, spoon theory or restraint collapse, for which no peer-reviewed first description exists. They are in the lexicon because many people use them, and they are marked there as community terms. They do not get an invented source.

What deliberately does not happen

There are no AI-generated sources. There are no secondary citations of the "cited in" kind. There are no content summaries that are retrofitted with a plausibly sounding DOI. If no validated lexicon entry exists for a statement, the passage stays without a source. The statement can then only appear in a generic form ("research on the locus coeruleus shows ...") or not at all.

This rule is strict, because it is the only one that prevents drift. The moment a single statement carries an invented DOI, the anchor is damaged for every other one.

Where the block is visible

The source block sits at the end of the article, in both the German and the English version. The same data structure feeds the source reference in the personal glossary inside the app. Chat responses themselves contain no citations. When the app explains a term and the term is in the lexicon, the same entry carries it into the glossary with the same validated source.

Maintenance happens in clearly bounded waves. Each wave checks a closed group of mechanisms, documents the three check steps per entry and writes the result as a code change. Entries are not added silently. New sources only enter through a new wave, never through a one-off edit on the blog. In addition, a check tool compares every DOI in the lexicon against the Crossref registry: whether it exists, whether the first author and year match the citation, and whether the work has been retracted.

What remains

Traceability is not a gesture. It is a verifiable property of the platform. Whoever reads about a mechanism on Autistic Mirror can travel from the statement to the original study in two clicks. Whoever finds no source knows that none was promised at that point on purpose. Both are part of the same discipline.

Autistic Mirror is an AI chat that explains autistic neurology individually and in relation to the specific situation. Mechanism instead of advice. Explanation instead of correction. The app is not a medical device and does not replace medical or therapeutic treatment.