How do we evolve raw, disconnected thoughts into coherent, actionable thinking? We explore the primitives and processes required to build a system for computational thought.
To represent "thought" computationally, we must first decompose it into its atomic units. These primitives allow us to move beyond simple text storage into structured knowledge representation.
The 'What'
Often represented as a predicate logic statement (e.g.,Causes(Rain, WetGround)) or a semantic triple. It transforms unstructured text into a standalone claim or fact.
The 'Why'
A rationale is a relationship. It captures the argumentative structure: Support, Attack, Justification, and Evidence. Without rationale, a graph is just a heap of facts.
The 'How Sure'
Thoughts are rarely binary truths. To model human thinking, we must quantify Confidence, Source Reliability, and Plausibility. This allows the system to weigh conflicting evidence.
The Template
Raw text often maps onto known structures like a "Medical Diagnosis" frame (Symptom → Cause → Treatment) or a "Problem-Solution" frame. We use Slots and Fillers to capture this.
How do isolated rationales become a "coherent thought"? It is a process of synthesis, conflict resolution, and linearization.
Gather all isolated rationales from the available corpus.
Ex: "Sky is blue", "It is raining", "Rain comes from clouds".
Organize propositions into a hierarchy using Rhetorical Structure Theory. Identify Nuclei (core ideas) and Satellites (supporting details). This establishes the relative importance of each piece of information.
When Rationale A conflicts with Rationale B, the system must use the Epistemic State to weigh the evidence.
Ex: "Market will rise" (Low Confidence) vs "Market will crash" (High Reliability Source).
Finally, traverse the resolved graph to produce a linear flow. This linearity—a clear path through the forest of facts—is what we perceive as a "coherent thought".
A breakdown of the core vocabulary used in this research, including why each concept is strategically relevant to the system.
A statement that expresses a judgment or opinion. In this context, it is the atomic unit of meaning (e.g., "The sky is blue").
Transforms unstructured text into discrete, addressable units that the system can manipulate.
The underlying reason or logical basis for a proposition. It connects isolated facts into an argument.
Allows us to model reasoning (the "why"), not just knowledge (the "what"), enabling conflict resolution.
The degree of validation or certainty attached to a piece of information (e.g., "Confirmed", "Hypothetical", "Dubious").
Essential for weighing conflicting evidence. A system without uncertainty cannot handle the nuance of human thought.
A cognitive framework or concept that helps organize and interpret information (e.g., a "Diagnosis" template).
Provides the structural "skeleton" for raw thoughts, allowing the system to anticipate and prompt for missing information.
In Rhetorical Structure Theory (RST), the central part of a text span, which is most essential to the writers purpose.
Identifying the nucleus helps the system determine what must be preserved during summarization or synthesis.
In RST, the supporting part of a text span which is incomprehensible without the Nucleus.
Provides evidence or context. Knowing what is a satellite allows the system to prune detail without losing the core message.
The process of flattening a graph or network of thoughts into a sequential narrative flow (like writing a paragraph).
Humans consume information linearly. To communicate the "coherent thought" back to the user, the graph must be flattened.
This system maps academic concepts to the primitives defined in this research document. This "Translation Map" helps you understand the origin of our terminology and identifying relevant concepts in the source material.
| Primitive | Source (Origin) | Term Mapping | Status |
|---|---|---|---|
| Structuring The definition of "Nucleus" as the central idea and "Satellite" as supporting material is the direct basis for our Structuring primitive. | Rhetorical Structure Theory (RST) Mann & Thompson (1988) Paper | "Nucleus / Satellite Schema" | Adopted |
| Epistemic State Jøsang’s "Opinion Triangle" provides the mathematical basis for quantifying uncertainty, which we adapt into our Epistemic State. | Subjective Logic Jøsang (2016) Book | "Opinion (Belief, Disbelief, Uncertainty)" | Adapted |
| Propositions The RDF standard establishes the "Triple" as the atomic unit of data, which directly informs our definition of a Proposition. | Semantic Web / RDF W3C Recommendation Concept | "Subject-Predicate-Object Triple" | Adopted |
| Conflict Resolution Dung’s framework for "Attack Relations" is the primary candidate for modeling how conflicting rationales interact in our system. | Abstract Argumentation Dung (1995) Paper | "Attack Relations / Admissible Sets" | Auditioning |