A schema is a mental structure that organizes knowledge — a pattern of related information, typically thoughts or behaviors, that defines the bits it holds and the inter-relationships between them. Schemas are the structured knowledge we carry and apply to whatever we encounter, and the information inside a schema can itself be made of schemas, nesting all the way down. Cognitive scientists formalize them in several ways, including frames, scripts, and semantic networks. Because a schema arrives before perception does, it shapes what we notice, how we interpret it, and what we expect next: incoming information gets slotted into an existing structure, ambiguity gets resolved in the structure's favor, and gaps get filled with the schema's defaults rather than the world's facts. This is why two people can witness the same event and walk away with different accounts — they ran it through different schemas. The same mechanism guides behavior, since a schema for a situation carries an implicit script for how to act in it.
For my work, schemas are the load-bearing unit of belief. If you want to model how people process information and form convictions, you cannot stop at attitudes or stated opinions — you have to get at the underlying structures that organize what someone already knows and govern how new information is absorbed or rejected. Beliefs do not shift because we supply better facts; they shift, or refuse to, because of the schema doing the interpreting. That is the level I try to make computational.