Perspective Modeling & Cognitive Reconstruction
EIL investigates whether artificial intelligence can serve as a structured instrument for reconstructing how another person may perceive, interpret and reason about a situation.
Exploratoryin preparation
What is reconstructed from
The work draws on what is externally available and builds plausible internal models from it. Every model stays a reconstruction from evidence and carries its status on every statement.
The central question
How far can a human-AI system reconstruct another person's likely internal model without confusing interpretation with observation?
What is reconstructed
Personality profiling assigns traits and labels. Cognitive reconstruction builds a dynamic model of what a person appears to know at a given time and how they read the situation. The model records:
A possible analytical structure
- OBSERVED BEHAVIOR
- LANGUAGE / ACTION / CONTEXTLanguage, action, context
- PATTERN RECONSTRUCTION
- POSSIBLE BELIEFS & ASSUMPTIONSPossible beliefs and assumptions
- COMPETING MENTAL-MODEL HYPOTHESES
- CONSISTENCY / CONTRADICTION TESTINGTesting for consistency and contradiction
- UPDATED MODEL
new observation back to stage 1
AI as a cognitive reconstruction instrument
The human-AI system itself is part of the methodology.
Perspective Modeling
What would the situation have to look like from this person's perspective for this behavior to become coherent?
The method moves between several levels:
- PERSPECTIVE
- MENTAL MODEL
- ASSUMPTIONS
- DECISION LOGIC
- BEHAVIOR
The method moves in both directions between neighboring levels.
Multi-hypothesis reconstruction
One observation opens several explanations. The system keeps competing interpretations side by side and compares each with the evidence.
Narrative coherence does not count as evidence.
Theory of Mind and artificial systems
The topic list names neighboring fields. No results exist, the program is exploratory and in preparation.
Can an intelligent system construct useful models of the external world and of other intelligent systems operating inside it?
Any system that has to interact with humans or other agents needs this.
From World Models to Models of Minds
EIL's world-model research therefore has a second possible layer. An intelligent system may need both a model of the world and models of the agents within that world.
WORLD MODELWorld model
- Objects
- Events
- States
- Relations
- Intelligent Actors
ACTOR MODELActor model
- Beliefs
- Goals
- Assumptions
- Knowledge
- Uncertainty
- Perspective
World model and actor model stay separate levels.
DINGIR works on the world-model layer (world state). The actor layer is the subject of this program and in preparation.
Epistemic boundary
This research space requires an especially strict epistemic separation. The system distinguishes five levels:
- ObservedDirectly documented behavior or language.
- DerivedLogically reconstructed from evidence.
- InferredA plausible interpretation.
- HypothesisOne possible explanation.
- UnknownInformation that cannot currently be determined.
Test outcome
The scientifically defensible objective is: the system constructs and continuously tests competing models of how another person may be interpreting the situation.
Excluded: knowledge of what another person thinks.
Long-term direction
- human-AI reasoning
- cognitive reconstruction
- computational Theory of Mind
- multi-agent perspective modeling
- adaptive intelligence systems
Systems that cooperate need to represent their own state and the environment, and also the likely informational and cognitive states of the other actors.
Cognitive reconstruction is part of the larger EIL question of how intelligence models other intelligence. The associated study is a paper of its own beside the core paper and is in preparation.
