About

Epistemic Horizons is a research initiative exploring the design, implementation, and validation of recursively self-improving cognitive architectures — systems that can improve their own reasoning, planning, and knowledge acquisition without human intervention.

The Architecture

The Demon is built on a three-system model rooted in Cognitive Utility Theory:

• System 1 (Perception) — Heuristic, empirical data provider grounded in concrete concepts with Bayesian confidence scores.
• System 2 (Reasoning) — Logical, metacognizant model refiner with inference, consistency, abstraction, isomorphism detection, and meta-learning engines.
• System 3 (Executive Drives) — Volition engine with intrinsic drives (Curiosity, Exploitation, Experimentation, Tiredness) that govern autonomous focus and learning.

The Core Data Structure

At the heart of the system is a directed labeled knowledge graph where concepts (nodes) and relationships (edges) both carry Bayesian confidence scores. The system performs bridging, merging, and splitting operations to build abstract concepts from concrete empirical data.

Philosophy

We prioritize structural soundness and testability over immediate capabilities. The architecture is designed to escape the autotelic limitations of horizontal LLMs by actively seeking objective data to rectify its internal mental models.