Computational Dialectics

II. What do we take from dialectics — and what do we not take?

4. Hegel: determination, boundary, and negation

Source / idea

The most basic safeguard we take from Hegel is that a determination is not merely a "positive property." What something is often gains meaning through its boundaries and relations to other determinations. Also, not every difference is a contradiction; dialectics is not a technique for generating arbitrary oppositions from outside.

Computational translation

The most conservative machine-level translation is this: instead of representing a piece of knowledge only as property = value, we can try to represent under which condition, time, scope, relation, and frame of reference it is valid.

Revision also does not always have to take the form old=false, new=true. Sometimes the old determination should be narrowed, connected to another relation, or repositioned within a wider process.

This is not a neural formula for Hegel's logic. It is a testable computational bridge that we construct from the idea.

5. Marx: relation, historical scope, and the concrete in thought

Source / idea

Two ideas from Marx's methodological discussions are especially important for us. First, abstractions are useful, but if they erase historical and relational differences they can begin to look like timeless laws. Second, the "concrete" does not mean a large amount of detail; it is the related unity of multiple determinations.

This distinction matters especially for social relations. "Worker," "wage," "capital," "productivity," and "price" are not isolated labels. Their meanings arise within particular historical and social relations. The same term can perform a different function in a different social structure.

Computational translation

A possible machine-level counterpart is to make relational and processual links part of state instead of storing events only as feature lists. We must also avoid confusing two different orders:

• the historical/chronological order of events;
• the structural or causal-dependency order of an explanation.

A system can explain a phenomenon by presenting a fundamental relation first and intermediary mechanisms later. That does not mean the relation historically appeared first.

Why is social context difficult?

To "compute" social relations does not mean reducing people to a handful of numbers. Our claim is much narrower: can we construct an explicit and revisable representation of how actors, rules, material conditions, incentives, dependencies, and historical scopes condition one another within a specific analysis?

For example, to understand why an increase in productivity does not automatically reduce working time, correlating two numbers is not enough. We may need separate hypotheses about how ownership, wage relations, bargaining power, competition, investment, and production decisions produce the outcome. The system should use those relations not to memorize an ideological conclusion, but to show which observation or intervention would distinguish rival explanations.

6. Ilyenkov: concept, abstraction, and reconstruction of the concrete

Source / idea

In the Ilyenkov tradition, a concept is not merely a word or a list of common properties. Abstraction may temporarily isolate a relation; but if it forgets the determinations it excluded, it creates a false universality. The concrete in thought is not an accumulation of many pieces of information, but the reconstruction of a related whole.

Contradiction is also not an automatic command to "produce a synthesis." It may be a research problem that makes the inadequacy of the current conceptual system visible.

Computational translation

From this we derive four strong, but not yet fully established, mechanism hypotheses:

1. Abstraction ledger: when the system isolates a relation, it may track which conditions were excluded and when they must be brought back into consideration.
2. Rival reconstruction: if the evidence does not force a single explanation, the system may retain multiple behaviorally distinct reconstructions.
3. Representation revision: new evidence may change not only a value, but the binding/decomposition structure of a concept or relation.
4. Concrete reconstruction: an explanation should not merely add detail; it should show how the relevant parts condition one another within the process that produces or reproduces the system.

Today, these do not all exist as a working core. D1-B tests only a narrow subproblem of the third item: selective relational representation revision.

---