Definition
Mental models are internal representations of how some aspect of the external world works, used by the cognitive system to support reasoning, prediction, planning, explanation, and counterfactual simulation. The concept traces to Kenneth Craik's 1943 The Nature of Explanation, which proposed that thought operates by constructing "small-scale models" of external reality and running them under hypothetical inputs to anticipate outcomes. Philip Johnson-Laird's mental-model theory (1983 onward) formalized the framework for reasoning research, distinguishing it from rule-based logical accounts.
A mental model is structural rather than propositional: it represents a system's components and their relations directly, allowing inference by manipulation of the model itself rather than by application of formal rules over verbal descriptions. This makes mental-model reasoning fast and concrete but also limits it — models are typically partial (omitting details deemed irrelevant), iconic (preserving structural relations rather than literal appearance), and biased (the constructed model tends to reflect the most readily available exemplar of the situation type).
Mental models are constructed dynamically from long-term schemas (templates) plus current perceptual input plus task-relevant goals. The construction process is resource-intensive and vulnerable to information environment: what schemas the constructor has available, what features of the situation are emphasized in input, and what counter-considerations are made salient all shape which model gets built. Once constructed, a model tends to be retained and refined rather than reconstructed from scratch — making first impressions disproportionately influential.
Context
Within Prismatic's threat model, mental models are the durable substrate at which sustained epistemic annihilation campaigns ultimately operate. An attacker who succeeds in shaping which mental model the target constructs of some domain — the trustworthiness of a perpetrator, the structure of a financial situation, the topology of a social conflict — has installed a structure that will subsequently be retrieved automatically and applied as default interpretation in countless downstream situations.
The kompendium distinguishes between belief-level attacks (changing what the target thinks is true) and model-level attacks (changing how the target construes the situation). Belief-level attacks are reversible by counter-evidence; model-level attacks are not, because the model determines what counts as relevant evidence in the first place. EA techniques in the structural-corruption family (EA-1.4) target mental models specifically.
The Decision Engine's design treats subject testimony as embedded in implicit models that may themselves require investigation. When systematic patterns appear across many of a subject's statements suggesting a corrupted underlying model — for example, all interpretations of a perpetrator's behavior cohering with an idealized rather than evidence-warranted picture — the platform surfaces the model-level pattern explicitly rather than evaluating individual statements in isolation.
Interactive: Schema Network
A mental model is a network of concepts (nodes) connected by relations (edges). Adjust schema coherence to see how internal consistency strengthens or weakens the binding edges. Adjust model accuracy vs reality to see which nodes diverge from the world they purport to represent — and the resulting prediction error.
Internal consistency of the network's relations
How well the model corresponds to the external world
See Also
- Schema Theory — Templates from which models are built
- Default Mode Network — Self-model substrate
- System 1 / System 2 — System 1 retrieves models
- Cognitive Flexibility — Capacity to revise models
- Neuroplasticity — Mechanism of model installation
- Kompendium — EA technique taxonomy
- Epistemická anihilace — Threat model
Connect & Contribute
Created by Tomáš Korcak (korczis) | Open Source under GHL