Re: limitations of classification systems, fiction, lack of ontological commitment

Charles,

How great that you are inspired by this topic, and started diggin in. Yes,
you are definitely learning, and so is everyone

Great questions, I wish I had the time to give an appropriate response -
but I do not
Let me hurry to reply briefly before this thread gets buried into the
tasklist
*I look forward to see how this develops but may not be able to follow up


KR is about conceptual alignment and adherence to a worldview/model of your
choice.
in LLM what gets out of the model is asserted in the prompt.

So I am not sure what prompted these answers, without the prompt we cannot
understand and evaluate
the output.Sounds to me from the answer that the prompt injected bias.
You may also want to evaluate the output from different models then come
back to tell us
And rephrase your prompts

Also the output you get seems to prove that without ontological commitment,
LLM  outcomes may be a fantasy


I take this opportunity to remember QUINE
He is the one who brought up the notion of ontological commitment in the
first place
Quine, W. V. (1948). On What There Is. *Review of Metaphysics*, *2*(5),
21–38.

There are several passages in Quine's seminal paper 'on what there's'  that
I would like to cite as relevant to KR today
bu. I ll leave it to other participants to flesh out what is relevant to
them

I just report here the starting and concluding paragraphs

Paper begins with




*A curious thing about the ontological problem is its simplicity. It can be
put in threeAnglo-Saxon monosyllables: ‘What is there?’ It can be answered,
moreover, in aword— ‘Everything’—and everyone will accept this answer as
true. However, this ismerely to say that there is what there is. There
remains room for disagreement overcases; and so the issue has stayed alive
down the centuries.*
Ends with





*From among the various conceptual schemes best suited to these various
pursuits,one—the phenomenalistic—claims epistemological priority. Viewed
from within thephenomenalistic conceptual scheme, the ontologies of
physical objects andmathematical objects are myths. The quality of myth,
however, is relative; relative, inthis case, to the epistemological point
of view. This point of view is one among various,corresponding to one among
our various interests and purposes*

https://rintintin.colorado.edu/~vancecd/phil375/Quine.pdf




On Wed, Dec 3, 2025 at 1:00 PM CHARLES WAWERU <cmukabi@aol.com> wrote:

> Dear Paola, in my need to contribute albeit  in a small way. I did a
> little AI  search on ‘ Is there anything sits above knowledge
> representation for AI?’
> Here is output ..
>
> In AI, knowledge representation is a crucial aspect, but there are
> higher-level concepts that encompass it. Some of these include:
>
> - *Artificial General Intelligence (AGI)*: The ultimate goal of AI
> research, aiming to create intelligent systems that can perform any
> intellectual task.
> - *Cognitive Architectures*: Frameworks that model human cognition,
> integrating knowledge representation, reasoning, and learning.
> - *Symbolic Reasoning*: Using logical representations to reason and make
> decisions.
> - *Meta-Reasoning*: The ability of AI systems to reason about their own
> reasoning processes.
>
> These areas explore how AI systems can effectively utilize and manage
> knowledge.
> I continued asking ‘then generate hierarchical levels’.
>
> Here's a hierarchical representation of AI concepts related to knowledge
> representation:
>
> 1. *Artificial General Intelligence (AGI)*
> 2. *Cognitive Architectures*
> 3. *Meta-Reasoning*
> 4. *Symbolic Reasoning*
> 5. *Knowledge Representation*
> 6. *Machine Learning*
> 7. *Data Processing*
>
> This hierarchy reflects the progression from raw data to more abstract,
> intelligent processing and reasoning capabilities.
> Then I requested references :
>
> For the hierarchical levels of AI concepts, here's a general outline based
> on common AI research areas:
>
> 1. *Data Processing*: Basic data handling and computation.
> 2. *Machine Learning*: Systems learn from data.
> 3. *Knowledge Representation*: Structuring knowledge for AI use.
> 4. *Symbolic Reasoning*: Using logic for decision-making.
> 5. *Meta-Reasoning*: Reasoning about reasoning processes.
> 6. *Cognitive Architectures*: Frameworks modeling cognition.
> 7. *Artificial General Intelligence (AGI)*: Human-like general
> intelligence.
>
> Sources:
> - Russell & Norvig, "Artificial Intelligence: A Modern Approach"
> - Laird, "The Soar Cognitive Architecture"
> - Goertzel, "Artificial General Intelligence"
>
> Final I asked about the ‘concept and conceptual layer’.
>
> Concepts and conceptual layers are crucial in AI knowledge representation.
> Here's a possible hierarchy:
>
> 1. *Percepts*: Raw sensory data
> 2. *Concepts*: Abstract representations of objects, events, or ideas
> 3. *Categories*: Groupings of related concepts
> 4. *Relationships*: Connections between concepts (e.g., causal, spatial,
> temporal)
> 5. *Theories*: Complex models explaining phenomena
> 6. *Frameworks*: High-level structures organizing theories and concepts
>
> Conceptual layers can include:
> - *Subsymbolic*: Distributed representations (e.g., neural networks)
> - *Symbolic*: Explicit, interpretable representations (e.g., logic,
> ontologies)
> - *Conceptual Spaces*: Geometric representations of concepts (e.g., vector
> spaces)
>
> These layers help organize knowledge and facilitate reasoning, learning,
> and abstraction in AI systems.
> I am not sure where this information sits in the scheme of things.
> Kindly let me know if I am learning or wallowing in confusion .
> Charles
>
> On 30 Nov 2025, at 04:30, Paola Di Maio <paoladimaio10@gmail.com> wrote:
>
> Some participants come to this list to learn about KR, and thus, about the
> world
> Other may come to impose their views of the world
> I only share some thoughts in the hope to inspire newcomers to the
> discussions to be skeptical of the reductionist views, especially
> when they are fictional
>
> The metaphor  of ' finger pointing at the moon  may be useful to explain
> how maths relates to the real world
> *moon=object in the real world,  finger=pointer to an object
>
> Lack of ontological commitment in mathematics does *not reduce its
> usefulness*.It allows mathematics to serve as a *symbolic, structural, or
> fictional framework* that organizes knowledge, supports reasoning, and
> aids scientific modeling, *without asserting that numbers, sets, or
> functions exist as real entities*.
>
> *Just some side notes for the record  *no problem if some participants
> have different views!*
>
> 1. The limitations of classification systems are well understood in
> science !
> All classification systems have representational limitations—structural,
> cultural, and epistemic constraints that prevent them from perfectly
> capturing the complexity of real-world subjects, and are sometimes
> misaligned
>
> Subject classification systems simplify and distort the vast complexity of
> knowledge. Their limitations stem from:
>
>    -
>
>    Structural constraints (hierarchies, reductionism)
>    - Cultural and historical biases
>    -
>
>    Linguistic and epistemic factors
>    -
>
>    The ever-changing nature of knowledge
>
>
> 2. *Ontology captures and represents 'what exists*'  *
> Ontic categories describe what exist
>
>
> 3. *MORE ON Lack of Ontological Commitment of Mathematics*
> *Fictionalism* Mathematics is akin to a story: numbers, sets, and
> functions are characters or constructs in a narrative.
>
> Statements like “2+2=4” are “true” within the story, but there is no
> metaphysical commitment to numbers actually existing.
>
> Hartry Field’s Science Without Numbers demonstrates how physics can be
> formulated nominalistically, showing mathematics is dispensable to physical
> ontology.
>
> *Nominalism* Mathematics is a linguistic or conceptual system, describing
> patterns, relations, or structures without positing entities.
>
> Mathematical objects are seen as placeholders or names, not actual beings.
>
> * Formalism*
>
> Mathematics consists of symbol manipulation according to rules.
>
> Truth is internal to the formal system, not dependent on entities existing
> in reality.
>
> There is no ontological claim beyond the consistency of the formal
> structure.
>
> ________________________________
>
> *Implications of Lack of Ontological Commitment*
>
> Philosophical: Avoids metaphysical debates over the existence of abstract
> objects.
>
> Scientific: Shows that mathematics can be used as a tool for modeling,
> explanation, and prediction without assuming mathematical objects exist.
>
> Epistemic: Shifts focus from discovering “real” entities to understanding
> structures, patterns, and relations.
>
> Practical: Emphasizes that mathematical work is justified by utility,
> coherence, and explanatory power rather than ontological truth.
> ------------------------------
>
>
> *MORE LIMITATIONS OF CLASSIFICATION SYSTEMS*
>
> 1. Reductionism
>
> Classification systems force complex, multifaceted subjects into
> predefined, discrete categories.
>
> Real-world topics often span multiple domains.
>
> Example: “Climate change” involves science, politics, economics,
> ethics—but often must be placed in one dominant category.
>
> Limitation: Nuanced or interdisciplinary knowledge becomes oversimplified.
>
> ________________________________
>
> 2. Rigid Hierarchies
>
> Most classification systems are hierarchical (trees), assuming that
> knowledge can be arranged from general → specific.
> But many fields do not follow clean hierarchies.
>
> Consequences:
>
> Relationships between subjects that are lateral, cyclical, or network-like
> are lost.
>
> Some topics fit multiple parent categories but must be assigned only one.
>
> ________________________________
>
> 3. Cultural Bias and Eurocentrism
>
> Many widely used systems were created in Western institutions during
> specific historical periods.
> Thus they often reflect:
>
> Western cultural priorities
>
> Colonial perspectives
>
> Christian or Euro-American worldviews
>
> Gendered assumptions
>
> Examples:
>
> Dewey Decimal once grouped non-Christian religions as a single minor
> section.
>
> Indigenous knowledge systems do not map neatly onto Western
> categorizations.
>
> ________________________________
>
> 4. Static Categories in a Dynamic Knowledge Landscape
>
> Knowledge evolves, but classification schemes update slowly.
>
> Limitations:
>
> Emerging fields (e.g., AI ethics, quantum biology) lack appropriate
> categories.
>
> Outdated terminology persists long after it becomes obsolete.
>
> ________________________________
>
> 5. Ambiguity and Boundary Problems
>
> Subjects don’t always have sharp boundaries.
>
> “Digital humanities,” “bioinformatics,” “neuroeconomics”—these hybrid
> fields strain rigid category structures.
>
> Result: Misclassification or forced placement into inadequate categories.
>
> ________________________________
>
> 6. Language-Based Constraints
>
> Classification systems often depend on the language in which they were
> originally created.
>
> Concepts with no direct translation get misrepresented.
>
> Polysemous words (one term, many meanings) complicate categorization.
>
> ________________________________
>
> 7. Ethical and Social Framework Limitations
>
> Some subjects carry social or moral implications the system fails to
> handle gracefully.
>
> Examples:
>
> LGBTQ+ topics historically hidden or marginalized
>
> Mental health categories shaped by outdated frameworks
>
> Stigmatizing terminology baked into classification labels
>
> ________________________________
>
> 8. Practical Space Constraints
>
> Especially in library systems:
>
> Only a finite number of codes or shelf spaces exist.
>
> Broad areas get subdivided excessively; others receive disproportionately
> little granularity.
>
> Outcome: Arbitrary compression or over-expansion.
>
> ________________________________
>
> 9. Authority and Gatekeeping
>
> Classification presumes that experts can definitively decide how knowledge
> should be structured.
>
> But:
>
> Some knowledge systems (e.g., community knowledge or oral traditions)
> resist systematization.
>
> Marginalized groups often have limited influence over classification
> design.
>
> ________________________________
>
> 10. Interoperability Problems
>
> Different systems don’t align cleanly.
>
> Translating between Dewey, LCC, MeSH, or scientific taxonomies can distort
> meaning.
>
> Metadata loss occurs during crosswalks (mapping between classification
> systems).
>
>
>
>
> However, if it helps, a reminder that it is what is generally accepted,
>
>
> 1. maths is  type of KR
> 2. is not NL KR *which is what we use in LLM
>
> Subsumption
> Subsumption is a key concept in knowledge representation, ontology design,
> and logic-based AI. It describes a “is-a” hierarchical relationship where
> one concept is more general and another is more specific.
> mathematics *is* a knowledge representation *although it may be
> understood or defined in other ways because  it provides:
>
>    -
>
>    Formal symbols (numbers, variables, operators)
>    -
>
>    Structured syntax (equations, functions, relations)
>    -
>
>    Precise semantics (well-defined meanings)
>    -
>
>    Inference rules (logical deduction, proof)
>
> and much more not related to what we are discussing here
>
>
>  Other views may also exist, in the vast universe of discourse, that may
> or may not contribute to the discussions in hand.
> .
>
>
>