John Beverley · Published 2024-10-04

How to Do Things with Ontologies

privateReview: unreviewedConfidence: high
Ontologybfobig datainteroperabilityphilosophysemantic websemantic-layerknowledge-graphgroundingschema-designmulti-agent
Open on YouTube ↗

Summary

Overview

  • Speaker: John Beverley
  • Channel: John Beverley
  • Main topic: How philosophical training in ontology engineering and formal ontologies provides unique value in solving big data interoperability challenges.
  • Purpose: To explain the value philosophers bring to ontology engineering, how ontologies solve big data interoperability challenges, and why demand for ontologists is growing. John Beverley, Assistant Professor at the University of Buffalo and VP of the National Center for Ontological Research (NCOR), presents on 'How to Do Things with Ontologies.' He discusses big data problems such as volume, velocity, variety, and veracity, focusing on information silos caused by incompatible storage systems and lack of standardized semantics. He introduces ontologies as formally well-defined, machine-interpretable controlled vocabularies and explains the role of Basic Formal Ontology (BFO) as a top-level architecture rooted in philosophical realism. Throughout the talk and subsequent Q&A, Beverley emphasizes why philosophers are uniquely suited for ontology engineering due to their training in conceptual analysis, formal logic, and handling disparate usage across scientific and industrial domains.

Topic Map

Big Data Problems and Information Silos

  • Explanation: Examining the challenges of big data, including huge volume (over 147 zettabytes stored by 2022), high velocity, variety of formats (JSON-LD, XML), and veracity issues leading to conflicting data and information silos that cannot be easily integrated.
  • Key claims:
    • Data is growing exponentially, making manual integration and human creativity insufficient.
    • Information silos cause data fragmentation, redundancy, bottlenecks, and incompatibility across repositories.
    • Lack of interoperability costs industrial datasets between $21-43 billion, with mid-size companies spending $20-50 million annually.
  • Examples:
    • Data stored in docx files or text files in one database versus another incompatible system.
    • Disagreements between trusted academic journal articles or philosophical debates creating conflicting data.
  • Terminology:
    • information silos
    • interoperability
    • zettabytes
    • data fragmentation
    • veracity
  • Why it matters: Information silos prevent organizations from extracting true value from data, leading to massive financial waste and operational bottlenecks.

Ontology Solutions and the Promise of Ontology Engineering

  • Explanation: Ontologies serve as known solutions to information silo challenges by acting as formally well-defined, machine-interpretable controlled vocabularies that make implicit meanings explicit and provide a semantic layer.
  • Key claims:
    • Ontologies provide a lingua franca across silos, reducing fragmentation and redundancy.
    • They promote automated data integration, standardizing querying and mapping.
    • Ontologies hit all five 'V's of big data: volume, velocity, variety, veracity, and value.
  • Examples:
    • The Gene Ontology (GO) consortium, started in 1998 alongside the human genome project to coordinate biological terminology.
    • Government and industry adoption of ontologies in AI roadmaps (DARPA, DHS, NIST, Gardner).
  • Terminology:
    • ontology
    • semantic layer
    • lingua franca
    • Gene Ontology
    • knowledge graph
  • Why it matters: Ontologies allow computers to reason over diverse, multi-modal datasets encoded in different languages or formats.

Philosophical Foundations: BFO, Realism, and Methodological Convictions

  • Explanation: Exploring Basic Formal Ontology (BFO), created by philosopher Barry Smith at the University of Buffalo, as the ISO-standard top-level architecture used across over 700 projects and government agencies.
  • Key claims:
    • Realism: Ontologies are designed to represent the world rather than simply concepts about the world.
    • Adequatism: BFO is non-reductive; classes and relations motivated by research communities are not paraphrased away.
    • Fallibilism: BFO is committed to tracking scientific research over time, recognizing that models might change.
  • Examples:
    • Distinguishing between independent continuants (material entities like tables or agents) and realizable entities (roles and dispositions like professor role or solubility).
    • Modeling Barry Smith as an instance of Agent, which is a subclass of Object, which is a subclass of Material Entity.
  • Terminology:
    • Basic Formal Ontology (BFO)
    • realism
    • adequatism
    • fallibilism
    • continuant
    • occurrent
    • disposition
    • role
  • Why it matters: Using a rigorous top-level ontology like BFO prevents ad-hoc modeling errors and provides guardrails for cross-domain interoperability.

Why Philosophers Make Good Ontologists

  • Explanation: Why philosophers possess unique skill sets in conceptual analysis, formal logic, and persuasion that make them highly valued and heavily recruited in ontology engineering.
  • Key claims:
    • Philosophers are not theory-laden or practice-laden within a specific scientific domain, allowing them to remain neutral.
    • They are trained to frame problems in multiple ways, identify subtle conceptual distinctions, and justify positions rigorously.
    • Philosophers excel at finding ontological solutions to messy domain problems, such as resolving ambiguity in virus definitions or pathogen dispositions.
  • Examples:
    • Resolving conflicting definitions of 'virion' and 'pathogen' among life science researchers.
    • Job market demand and 6-figure entry-level salaries for ontologists across finance, government, and tech.
  • Terminology:
    • conceptual analysis
    • formal logic
    • virion
    • pathogenic disposition
    • ontology engineering
  • Why it matters: Subject-matter experts often struggle with cross-domain conceptual clarity; philosophers provide the rigorous logical framework required for machine-interpretable data models.

Key Points

Big data problems stem from information silos, not just data volume

  • Explanation: Organizations struggle more with data incompatibility, proprietary formats, and lack of integration than raw storage capacity.
  • Evidence: NIST estimated lack of interoperability costs industrial datasets $21-43 billion; McKinsey estimates mid-size companies spend $20-50 million annually due to silos.
  • Practical implication: Computing strategies alone cannot solve fragmentation without semantic integration frameworks.

Realism over conceptualism in ontology design

  • Explanation: Ontologies should model the world as it is rather than modeling human concepts of the world to maintain coordination.
  • Evidence: Going from the world to words keeps researchers coordinated, whereas going from words to concepts leads to divergent definitions.
  • Practical implication: Ontology engineers must avoid subjective conceptual modeling and ground definitions in metaphysical reality.

Growing high-demand career path for philosophers in ontology

  • Explanation: There is a severe talent shortage of skilled ontologists across government agencies, financial institutions, and tech sectors, resulting in lucrative career opportunities.
  • Evidence: Job postings from organizations like Johns Hopkins APL, Morgan Stanley, Citi, and the University of Pennsylvania explicitly require philosophy backgrounds.
  • Practical implication: Philosophical training in logic and conceptual analysis translates directly into high-paying applied tech roles.

Frameworks, Models & Processes

Basic Formal Ontology (BFO)

  • How it works: Provides a top-level ontological architecture with general categories (continuant, occurrent, material entity, realizable entity) that specific domain ontologies extend downward from.
  • Components:
    • Continuant (independent and dependent continuants)
    • Occurrent (processes, temporal regions)
    • Material Entity
    • Realizable Entity (roles, dispositions, functions)
  • When to use: When building domain ontologies that require rigorous interoperability, consistency checking, and alignment with standards like ISO.

Hub & Spoke Strategy

  • How it works: Uses a top-level hub ontology (like BFO) connected to multiple modular spoke ontologies representing specific domains via downward population.
  • Components:
    • Hub ontology (BFO)
    • Spoke ontologies (domain-specific modules)
    • Downward population hierarchy
    • Shared semantics guardrails
  • When to use: When integrating disparate data sources and ensuring consistent cross-domain validation and interoperability upfront.

Examples & Case Studies

Life science researchers used a Linnean taxonomy to classify viruses, resulting in a monstrous and messy hierarchy where viruses like HIV ended up classified under human patient subcategories.

  • Illustrates: The inadequacy of ad-hoc or domain-biased taxonomies for complex data modeling.
  • Lesson: Philosophical training helps identify subtle distinctions and propose cleaner classification models, such as adopting the Baltimore classification or disposition-based pathogen models.

A UB philosophy graduate student went to a conference with John Beverley, met his former boss at Johns Hopkins Applied Physics Lab in the bathroom, and walked out with a job offer in 3 minutes.

  • Illustrates: The extreme market demand and high employability of philosophy graduates with ontology training.
  • Lesson: Ontology skills are so scarce and valuable that qualified candidates receive rapid hiring interest.

Actionable Takeaways

  • Immediate:
    • Read 'Semantic Web for the Working Ontologist' as a foundational text.
    • Participate in open NCOR ontology training sessions and working groups.
    • Create a GitHub account to track ontology artifacts, issues, and contributions.
  • Strategic:
    • Adopt BFO and realist ontology principles to future-proof data integration projects.
    • Recognize that generative AI highlights the necessity of rigorous semantic layers and formal logic rather than replacing them.
    • Position philosophical training as a high-value asset in tech and data science job markets.
  • Questions to investigate:
    • How can ontologies better integrate with machine learning models and large language models?
    • What are the precise boundaries between descriptive logic and first-order logic in practical ontology development?
    • How can government agencies accelerate the adoption of standardized core ontologies across disparate departments?

Claims Worth Verifying

  • As of 2022, over 147 zettabytes of data had been stored globally. (statistic)
  • Lack of interoperability across industrial datasets costs companies between $21-43 billion (NIST report). (report finding)
  • Mid-size companies spend 20-50 million dollars annually due to information silos (McKinsey estimate). (report finding)
  • Basic Formal Ontology (BFO) is the only ISO standard for top-level ontology development. (standardization fact)

Notable Quotes

"How philosophers provide a lot of value in the field that I work in, building ontologies and using them." (at 0:29) "Ontologies are known solutions to information silo challenges in the context of Big Data." (at 16:40) "Ontologies are formally well-defined machine-interpretable controlled vocabularies designed to represent entities and logical relationships among them." (at 16:52) "Computers are dumb and they do exactly what you tell them and no more or no less." (at 42:26) "If we attempt to go from the words we use to the world, it is unlikely that we will end up in the same place. If we attempt to go from the world to the words we use, it is more likely we will remain coordinated." (at 34:15)

Compressed Summary

  • Big data creates massive interoperability costs due to information silos and unstructured semantics.
  • Ontologies provide formally defined, machine-interpretable semantic layers acting as a lingua franca across data silos.
  • Basic Formal Ontology (BFO) relies on philosophical realism, adequatism, and fallibilism to model the real world.
  • Philosophers excel in ontology engineering because of their rigorous training in conceptual analysis, formal logic, and neutral problem framing.
  • There is an exploding job market and high demand for trained ontologists across government, finance, and technology.
  • Keywords: ontology, bfo, big data, interoperability, philosophy, semantic web
  • Core insight: Applying philosophical realism and formal ontology engineering to big data solves the trillion-dollar interoperability problem caused by information silos.

Core insights

6
Architecturemedium noveltymoderate evidence

The interoperability problem is architectural, not storage-related: the primary cost of big data is incompatible storage systems and non-standardized semantics (information silos), not raw volume. The offered remedy is a formally-defined, machine-interpretable controlled vocabulary placed as a semantic layer between heterogeneous repositories and their consumers.

Why it matters

It relocates integration responsibility from per-silo application code to a shared formal semantic layer, which is exactly the boundary question engineers face when wiring multiple data sources, retrieval stores, or tool schemas into an agent.

Generalization

Any system integrating N heterogeneous sources should invest in one shared machine-interpretable semantic layer rather than N^2 point-to-point mappings; the mapping/querying burden collapses onto the shared layer.

Ontologies provide a lingua franca across silos, reducing fragmentation and redundancy.
Open source video
Ontologies allow computers to reason over diverse, multi-modal datasets encoded in different languages or formats.
Open source video
Information silos cause data fragmentation, redundancy, bottlenecks, and incompatibility across repositories.
Open source video
Mental Modelhigh noveltymoderate evidence

Realism as a modeling discipline: definitions should be anchored to the world (referents) rather than to human concepts about the world. The claimed asymmetry is that world-to-words pathways keep independent parties coordinated, while words-to-concepts pathways produce divergent, mutually incompatible definitions.

Why it matters

It inverts a common practice of defining shared vocabularies by agreeing on concept labels; if coordination is the goal, the anchor must be an external referent, which materially changes how shared schemas, memory keys, and tool contracts are authored.

Generalization

In any multi-agent or multi-team setting, ground shared identifiers in observable environment state/entities rather than in negotiated concept labels to reduce semantic drift.

Going from the world to words keeps researchers coordinated, whereas going from words to concepts leads to divergent definitions.
Open source video
Architecturemedium noveltymoderate evidence

A top-level ontology such as BFO functions as a stable upper architecture that constrains domain models, reused across 700+ projects and government agencies. Its three stated commitments are realism (represent the world), adequatism (non-reductive: do not paraphrase away classes/relations that research communities need), and fallibilism (commit to revising the model as scientific findings change).

Why it matters

This is a concrete pattern for a two-tier schema regime — a small, revisable upper layer enforcing invariants, plus flexible domain layers — which is directly analogous to designing constrained core abstractions for agent memory, tool interfaces, or shared state.

Generalization

Separate a minimal invariant core schema from domain-extensible layers; make the core explicitly revisable so accuracy improvement does not require rewriting domain layers.

Using a rigorous top-level ontology like BFO prevents ad-hoc modeling errors and provides guardrails for cross-domain interoperability.
Open source video
BFO is non-reductive; classes and relations motivated by research communities are not paraphrased away.
Open source video
BFO is committed to tracking scientific research over time, recognizing that models might change.
Open source video
Mechanismhigh noveltymoderate evidence

The ontology distinguishes independent continuants (material entities such as tables or agents) from realizable entities (roles and dispositions, e.g., a professor role or solubility) — i.e., capabilities and roles are modeled as first-class entities that inhere in continuants rather than as boolean state flags.

Why it matters

Modeling unexercised capabilities as explicit entities enables reasoning about what an entity could do, not just what it currently is — relevant to capability registries, tool availability, and permission modeling in agent runtimes.

Generalization

Represent latent capabilities/dispositions as distinct typed entities attached to substrates, so planners can reason over potential behavior independent of current activation state.

Distinguishing between independent continuants (material entities like tables or agents) and realizable entities (roles and dispositions like professor role or solubility).
Open source video
Practicemedium noveltymoderate evidence

Cross-domain terminology conflicts are resolved by a domain-neutral party: the argument is that philosophers are 'not theory-laden or practice-laden within a specific scientific domain,' letting them arbitrate conflicting definitions (e.g., 'virion' and 'pathogen' across life-science groups) while subject-matter experts cannot.

Why it matters

It argues for a dedicated neutral reconciliation role rather than delegating schema/definition merging to the most domain-expert participant, which is a testable organizational and agent-role design choice.

Generalization

Assign cross-domain semantic reconciliation to a domain-neutral arbiter (human or model) rather than to the party with the deepest single-domain commitment.

Philosophers are not theory-laden or practice-laden within a specific scientific domain, allowing them to remain neutral.
Open source video
Resolving conflicting definitions of 'virion' and 'pathogen' among life science researchers.
Open source video
Failure Modelow noveltymoderate evidence

The failure mode is automation-capacity, not storage-capacity: data growth is framed as outrunning manual integration and even human creativity, so the integration layer must itself be machine-interpretable rather than a documentation artifact for humans.

Why it matters

It sets the bar for the shared semantic layer: a glossary intended for human reading does not satisfy machine interpretability, so integration must be encoded in a form machines can reason over.

Generalization

For scaling systems, any shared contract between components must be machine-consumable, not just human-readable documentation.

Data is growing exponentially, making manual integration and human creativity insufficient.
Open source video

Deep dives

4

LLM-based neutral arbitration for cross-domain ontology reconciliation

Research question

Can a domain-neutral LLM arbiter, constrained by a top-level ontology, resolve cross-domain definition conflicts as well as or better than human neutral arbiters, without collapsing domain-specific distinctions?

Why

It tests whether the human philosopher's reconciliation role can be automated, which would scale semantic integration across many domains and reduce reliance on scarce neutral human experts.

Philosophers are not theory-laden or practice-laden within a specific scientific domain, allowing them to remain neutral.
Open source video
Resolving conflicting definitions of 'virion' and 'pathogen' among life science researchers.
Open source video
Source video

Realizable entities vs state flags in agent capability modeling

Research question

Does modeling agent capabilities as realizable entities (dispositions/roles) rather than boolean state flags improve planning correctness on tasks requiring unexercised capabilities?

Why

It provides a concrete modeling choice for agent runtimes with measurable planning outcomes, directly affecting how tool registries and permission systems are designed.

Distinguishing between independent continuants (material entities like tables or agents) and realizable entities (roles and dispositions like professor role or solubility).
Open source video
Source video

Measuring the five V's: ontology-mediated vs silo-based integration

Research question

Do ontology-mediated pipelines measurably improve volume, velocity, variety, veracity, and value handling compared to silo-based pipelines, and under what conditions?

Why

The claim that ontologies hit all five 'V's is stated without measurement; a benchmark would validate or falsify it and guide investment decisions.

Ontologies hit all five 'V's of big data: volume, velocity, variety, veracity, and value.
Open source video
Lack of interoperability costs industrial datasets between $21-43 billion, with mid-size companies spending $20-50 million annually.
Open source video
Source video

Versioning and migration for fallibilist top-level ontologies

Research question

How should a fallibilist top-level ontology be versioned and migrated so that downstream domain models incur minimal churn while the upper layer remains scientifically current?

Why

Fallibilism is a core commitment but lacks operational mechanisms; unresolved versioning creates adoption risk and migration costs that can negate interoperability benefits.

BFO is committed to tracking scientific research over time, recognizing that models might change.
Open source video
Source video

Article ideas

4

The Semantic Layer Is the Integration Layer: Why N^2 Mappings Fail

The cost of big data interoperability is architectural, not volumetric; investing in a shared machine-interpretable semantic layer collapses N^2 point-to-point mappings into a single contract, and any agent system integrating heterogeneous sources should do the same.

Angle

Argue that engineering teams misdiagnose interoperability failures as storage or volume problems when the real fix is a shared formal semantic layer; illustrate with agent tool/schema integration.

Source video

Grounding Shared Vocabularies in the World, Not in Concepts

Multi-agent and multi-team systems should anchor shared identifiers in observable environment state rather than negotiated concept labels, because words-to-concepts pathways produce divergent definitions while world-to-words pathways keep independent parties coordinated.

Angle

Invert the common practice of agreeing on concept labels for shared schemas, memory keys, and tool contracts; propose environment-referent grounding as a design discipline.

Source video

The Neutral Arbiter: Why Cross-Domain Reconciliation Needs a Domain-Neutral Party

Delegating schema and definition merging to the deepest domain expert produces worse cross-domain interoperability than assigning it to a domain-neutral arbiter (human or LLM agent), because domain commitment introduces bias; organizations should explicitly design for neutrality.

Angle

Test the counterintuitive claim that lacking domain depth is an asset for reconciliation; discuss implications for agent role assignment and organizational design.

Source video

Modeling Capabilities as Realizable Entities: A Better Pattern for Agent Tool Registries

Representing latent capabilities and roles as first-class typed entities attached to substrates, rather than as boolean state flags, enables planners to reason about potential behavior independent of activation state, improving agent planning correctness.

Angle

Translate the ontology distinction between continuants and realizable entities into a concrete pattern for agent capability registries and permission models.

Source video

Project ideas

4

Neutral Arbiter Agent

new

An LLM-based agent, prompted with a top-level ontology (e.g., BFO) and conflicting domain definitions, can resolve conflicts (e.g., 'virion' vs 'pathogen') without collapsing legitimate distinctions, achieving higher inter-annotator agreement than domain-expert-only resolution.

Proof of concept

Build a prototype that ingests conflicting definitions from two life-science subdomains, uses BFO constraints to propose unified definitions, and compares outputs to human expert arbitration on a held-out set of conflict cases.

Measurement

Definition convergence rate (% of conflicts resolved), preservation of domain-specific distinctions (expert rating), and time to resolution compared to a human neutral arbiter baseline.

Source video

Disposition-Typed Capability Registry for Agents

movement-lab

Modeling agent capabilities as realizable entities (dispositions/roles) with explicit substrate attachment improves planning correctness on tasks requiring unexercised capabilities compared to boolean state flags, because planners can reason over latent capabilities.

Proof of concept

Implement two versions of a tool registry for an LLM agent: one with boolean flags, one with typed disposition entities; evaluate on a set of planning tasks that require recognizing unexercised capabilities.

Measurement

Task success rate, planning latency, and number of invalid tool attempts across the two registry designs.

Source video

Ontology-Mediated Integration Benchmark

beyond-evals

Ontology-mediated data integration pipelines reduce integration latency and mapping effort and lower error rates compared to silo-based point-to-point pipelines when integrating N heterogeneous sources.

Proof of concept

Build two pipelines for a synthetic multi-source dataset (e.g., 5 sources with different schemas); one uses a shared ontology-based semantic layer, the other uses point-to-point mappings. Compare on integration tasks and adding a new source.

Measurement

Integration latency, number of mapping rules, query error rate, and cost of adding a new source for each pipeline.

Source video

Fallibilist Ontology Versioning Protocol

gatehouse

A versioned upper ontology with explicit deprecation and migration policies reduces downstream migration cost and semantic drift across dependent domain models compared to an unversioned top-level ontology.

Proof of concept

Define a versioning scheme for a small top-level ontology (e.g., BFO subset) and simulate a scientific finding that changes a core class; measure migration effort for 5 dependent domain models.

Measurement

Time to migrate, number of breaking changes, and semantic drift (difference in inferred classifications pre/post migration).

Source video

Architectural implications

3

Integration logic is currently duplicated per silo and per application.

Before

Each system stores data in its own format (docx, text files, JSON-LD, XML) with point-to-point or manual integration.

After

A shared, formally-defined machine-interpretable semantic layer mediates all sources and consumers.

Consequence

Standardized querying and mapping replace bespoke integration code, reducing fragmentation and redundancy.

Source video

Domain models are created ad hoc without a constraining upper layer.

Before

Each project defines its own classes from scratch with no invariant core.

After

A small top-level ontology (e.g., BFO) constrains all domain models and is shared across 700+ projects.

Consequence

Cross-domain interoperability is preserved as a side effect of the constraint, and ad-hoc modeling errors are prevented.

Source video

Neutrality is treated as a needed property of the reconciler, not depth of domain knowledge.

Before

Domain experts own and resolve terminology for their own datasets.

After

A domain-neutral party arbitrates cross-domain definition conflicts.

Consequence

Resistance to reconciliation drops, at the cost of the arbiter's domain depth.

Source video

Tradeoffs and failure modes

4

Adequatism vs. model parsimony

Benefit

Non-reductive modeling preserves every class and relation that a research community needs, maximizing fidelity and adoption.

Cost or risk

Refusing to paraphrase classes away grows the model and can sacrifice simplicity/compactness, complicating maintenance and reasoning.

BFO is non-reductive; classes and relations motivated by research communities are not paraphrased away.
Open source video
Source video

Fallibilism vs. downstream stability

Benefit

Committing to track scientific research keeps the top-level ontology accurate over time.

Cost or risk

If the upper ontology may change, downstream adopters face version churn and migration costs on a foundation they depend on.

BFO is committed to tracking scientific research over time, recognizing that models might change.
Open source video
Source video

Neutrality vs. domain expertise

Benefit

A domain-neutral ontologist can arbitrate conflicting definitions without being invested in one camp.

Cost or risk

The neutral arbiter lacks deep domain practice, so resolution quality depends on eliciting and checking expert input rather than native expertise.

Philosophers are not theory-laden or practice-laden within a specific scientific domain, allowing them to remain neutral.
Open source video
Source video

Realism vs. speed of vocabulary agreement

Benefit

Anchoring to the world rather than to concepts keeps independent parties coordinated.

Cost or risk

Realist commitments require resolving what actually exists in the domain, which is slower and more contentious than agreeing on shared labels.

Going from the world to words keeps researchers coordinated, whereas going from words to concepts leads to divergent definitions.
Open source video
Source video

Open questions

4

Can the reconciliation and modeling work attributed here to human philosophers be performed by domain-neutral LLM-based agents or a dedicated arbiter agent?

Why unresolved

The summary only evidences human practitioners resolving conflicts such as 'virion' and 'pathogen'; no automated mechanism or experiment is described.

Research direction

Prototype a neutral arbiter agent given a top-level ontology plus conflicting domain definitions and measure whether it converges definitions without collapsing distinctions.

Source video

Does modeling dispositions/roles as first-class realizable entities measurably improve planning or reasoning versus modeling capabilities as state flags?

Why unresolved

The summary states the modeling distinction but provides no outcome, benchmark, or failure case showing a benefit.

Research direction

Build parallel capability models (disposition-typed vs boolean-state) and compare agent planning correctness on tasks requiring unexercised capabilities.

Source video

How is a fallibilist top-level ontology versioned and migrated for downstream dependents?

Why unresolved

Fallibilism is asserted as a commitment, but no versioning, deprecation, or migration mechanism appears in the summary.

Research direction

Define a versioning/deprecation policy for an upper ontology layer and measure migration cost across dependent domain models.

Source video

Do the five 'V' claims (ontology addressing volume, velocity, variety, veracity, and value) hold as measurable effects rather than framing?

Why unresolved

The claim is stated without supporting measurement in the summary.

Research direction

Design an interoperability benchmark comparing silo-based pipelines against ontology-mediated ones on integration latency, mapping effort, and error rates.

Source video

Key claims

8
factualVerification needed

Lack of interoperability costs industrial datasets an estimated $21-43 billion.

Evidence

NIST estimated lack of interoperability costs industrial datasets $21-43 billion; McKinsey estimates mid-size companies spend $20-50 million annually due to silos.

Question

What are the original NIST and McKinsey sources, scope, and methodology behind these figures?

Source video
factualVerification needed

Mid-size companies spend $20-50 million annually due to information silos.

Evidence

McKinsey estimates mid-size companies spend $20-50 million annually due to silos.

Question

What definition of 'mid-size' and what cost components does McKinsey include?

Source video
factualVerification needed

More than 147 zettabytes of data were stored by 2022.

Evidence

huge volume (over 147 zettabytes stored by 2022)

Question

Which measurement body produced the 147 zettabyte figure and does it include replicated/cold storage?

Source video
causalVerification needed

Information silos cause data fragmentation, redundancy, bottlenecks, and incompatibility across repositories.

Evidence

Information silos cause data fragmentation, redundancy, bottlenecks, and incompatibility across repositories.

Question

Can the fragmentation/redundancy effect be isolated from other organizational causes in a controlled comparison?

Source video
factualVerification needed

BFO is an ISO-standard top-level ontology used across over 700 projects and government agencies.

Evidence

as the ISO-standard top-level architecture used across over 700 projects and government agencies

Question

Is 700+ the current count of BFO-based ontologies, and under which ISO standard number is it registered?

Source video
opinionVerification needed

Ontologies address all five 'V's of big data: volume, velocity, variety, veracity, and value.

Evidence

Ontologies hit all five 'V's of big data: volume, velocity, variety, veracity, and value.

Question

Is there any measured evidence that an ontology-mediated pipeline improves velocity or volume handling versus a non-ontology pipeline?

Source video
causalVerification needed

Data growth has outrun manual integration and human creativity.

Evidence

Data is growing exponentially, making manual integration and human creativity insufficient.

Question

At what growth or source-count threshold does manual integration demonstrably fail?

Source video
factualVerification needed

Ontologists command six-figure entry-level salaries across finance, government, and tech.

Evidence

Job market demand and 6-figure entry-level salaries for ontologists across finance, government, and tech.

Question

What is the sample of postings and their compensation ranges supporting six-figure entry-level pay?

Source video

Connections

5