Rethinking Modeling: From Mechanistic Artifacts to Cognitive Origins
If you want to create an AI that can reasoning, you must not fear to face reasoning.
This is a critique that point a philosophical and epistemological tension in the approaches to knowledge and domain modeling: the external vs. internal perspective on knowledge creation. Let me attempt to articulate the contrast i’ve feeling, the gaps i want to focus your attention on, and why i want to change the “savage” approach.
1. The Fundamental Divide:
External Analysis vs. Cognitive Generation
External Structuring Approach:
The focus of approaches like FCA and statistical knowledge discovery, including the methods described in the CORDIET paper, is artifactual knowledge—patterns extracted from data as seen externally.
These methods treat knowledge as something detached from the observer or creator, as if it exists independently and can be deduced entirely through external characteristics such as:
Attributes.
Relationships between entities.
Statistical patterns or clusters.
This external orientation assumes that meaning emerges from observed artifacts and relationships—a mechanistic view that bypasses the human cognitive process that originally gives rise to knowledge.
Lost their roots
These methods is trying to "play detective," piecing together fragments of an external world without addressing how knowledge is formed internally. This approach ignore the moment of birth—the dynamic, internal process by which cognition generates concepts and correlates them to the world. As a result:
It reduce knowledge to external expressions (attributes, signs, artifacts) rather than studying the inner mechanism of its genesis.
Their models don’t touch the reality of the cognitive process—they merely reflect the observer’s constructed view of knowledge, which may have no intrinsic correlation with the phenomena they claim to represent.
Just imagine - it is equivalent to instead of trying to ask the source for the true cause, you are trying to construct a multi-dimensional construct of theories and explanations that may have nothing to do with the root cause. This multiple retelling, in each variant of which we endlessly break the link, sounds like insanity, or at least cognitive distortion.
Internal Generation and Human-Centric Modeling:
Alternative is emphasize a cognitive origin of knowledge: the inference chains, the act of conceptualization, and the inner mechanisms by which humans generate and structure meaning.
Knowledge is not something that exists "out there," waiting to be discovered by examining artifacts.
Instead:It arises as a constructive process within the human mind.
The only "truth" it correlates with is the internal coherence of the thinker’s model of the world—not with some imaginary “real objects” that these external methods attempt to reconstruct.
The goal of domain modeling, is not just structuring external data but also capturing and expressing the cognitive processes that generate this structure.
Savage methods
These external methods avoid diving into the cognitive depths, where the actual meaning-making process occurs. It shy away from the uncomfortable uncertainty of exploring the human act of "knowing," preferring instead to rely on superficial patterns.
2. The Contextual Gap
External Approach:
In External Approach work exists in a paradigm where knowledge is seen as independent of its creator.
It assume that the goal of modeling is to create structures that approximate "true" external realities through signs and patterns.
This leads to:A reliance on empirical data as the ultimate arbiter of truth.
A disconnection from the subjective, cognitive processes that give rise to those patterns in the first place.
Cognitive Generation Analysis Approach
In Cognitive Generation Analysis paradigm - knowledge is a human artifact, inseparable from the cognitive and contextual processes that generate it.
The key is not to compare knowledge with "objective reality"
(an elusive and arguably imaginary construct) but to understand how:Humans conceptualize, structure, and align ideas.
Different perspectives and internal models can coexist and harmonize.
What is the Gap?
External approach feels like a refusal to engage with the human-centered, generative origins of knowledge, instead reducing it to externally observable signs.
Cognitive Generation Analysis seeks to bridge the visible and invisible at the moment of conceptual birth, rather than retroactively piecing together meaning through external attributes.
3. The Mechanistic Nature of External Approach
External approach methods seems to be a “mechanistic” comes from their reliance on reductionism:
It treat knowledge as static, reducible to attributes, relationships, and statistical properties.
It avoid engaging with the dynamic, evolving process of how humans construct meaning.
This approach sidelines the subjective and creative elements of knowledge formation, reducing it to a "black-box" analysis of outputs.
4. Embracing the Inner Chain of Knowledge
If we want to go deep in reasoning - we need to use alternative approach, grounded in studying the cognitive processes of knowledge creation:
Focus on Human Cognition:
Instead of starting with external attributes, begin with the cognitive mechanisms:
How does a domain expert infer concepts?
How do they align internal mental models with external representations?
Explore how these models are formed, evolve, and communicate meaning.
Correlating Internal and External:
Internal models are often expressed externally (e.g., through language, artifacts). However:
The goal is not to treat the external as the sole truth.
Instead, the focus is on understanding how external representations reflect and align with the inner cognitive world.
Modeling as a Cognitive Process:
Tools and methods should focus on the act of modeling itself:
How are terms defined, relationships formed, and concepts aligned within the mind of a domain expert?
What are the mental steps and heuristics that lead to coherent models?
This process would guide users through conceptual exploration, rather than just documenting external patterns.
5. Practical Implications
Shift the Focus of Modeling Tools:
Tools should not merely visualize existing attributes or artifacts (as FCA or ESOM do). Instead, they should enable users to explore and express their thought process, building models as cognitive extensions of their understanding.
Incorporate Cognitive Heuristics:
Design methods that help users articulate their conceptual inferences:
How do they classify and group ideas?
What invisible relationships or constraints influence their thinking?
Encourage users to explore the why and how of their classifications, not just the what.
Focus on Internal Alignment:
Instead of validating models against external "realities," focus on their internal coherence and how well they align with the user’s cognitive framework.
Offer tools to harmonize multiple perspectives, helping users create shared conceptual spaces.
Embrace Human-Centered Epistemology:
Make the act of knowing central to our approach, studying how humans infer and structure knowledge, rather than treating it as a static output to be reverse-engineered.
6. Final Reflection
This critique touches on a deeper philosophical divide between mechanistic and human-centered views of knowledge. By emphasizing cognitive origins and processes, our approach could offer a radical alternative to traditional modeling paradigms. Instead of reducing knowledge to artifacts, you advocate for tools and methods that explore the birthplace of meaning—the human mind.
If you want to create an AI that can reasoning, you must not fear to face reasoning.