Conceptual engineering for the AI age

The concepts your
AI runs on are
probably broken.

Most AI projects fail not because of bad algorithms, but because organizations are working with incoherent concepts — 'fairness,' 'transparency,' 'accountability' — that don't do the job they're supposed to do. We find the defects and replace them.

Kevin Scharp Professor of Philosophy, UIUC  ·  Director, Arché Philosophical Research Centre (St Andrews, 2016–22)  ·  Edited 3-volume Springer collection on conceptual engineering
Philosophical concept diagram by Kevin Scharp
A philosophical system diagram — from kevinscharp.com/diagrams

"Philosophy as the the study of defective concepts."

— Kevin Scharp

01 — Background

Philosophy at the frontier of AI

Kevin Scharp is one of the world's leading authorities on conceptual engineering — the rigorous methodology for identifying defective concepts and replacing them with ones that actually work.

His academic work on the philosophy of AI has been published in top journals, including Philosophical Review and Synthese, and he is editing a three-volume collection on conceptual engineering for Springer Press. He directed the Arché Philosophical Research Centre at the University of St Andrews for five years and currently holds a position at the University of Twente's Ethics of Socially Disruptive Technology project.

Beyond academia, he has applied his expertise to real AI systems — including algorithmic derivatives trading — and creates philosophy content (video, podcast, blog) reaching a broad audience.

Education BA Mathematics, Washington University  ·  PhD Philosophy, University of Pittsburgh
Positions UIUC  ·  University of St Andrews  ·  University of Twente  ·  Ohio State University
Research Philosophy of AI  ·  Conceptual Engineering  ·  Ethics  ·  Philosophy of Language  ·  Truth Theory

02 — The method

Conceptual engineering.
Not strategy consulting.

Most AI consultancies offer implementation advice: how to build faster, cheaper, at scale. Axiom Advisory does something different. It brings the tools of conceptual engineering — the philosophical methodology Kevin helped build — to the most pressing challenges in AI.

The question isn't just how to build AI. It's whether the concepts you're building on are even coherent enough to guide the work.

01

Diagnose the defect

Identify which concepts in your AI strategy or governance framework are incoherent, inconsistent, or doing work they can't actually do. Fairness, accountability, explainability — all of these are often used without a clear, workable definition.

02

Map the conceptual landscape

Chart the relationships between your key concepts — where they overlap, conflict, or leave gaps. This gives you an exact picture of what you're actually working with, which is often very different from what you think you're working with.

03

Design replacement concepts

Using the methodology of conceptual engineering, develop new concepts that do the work you need them to do — without the incoherence. These are built to specification, not approximated from existing intuition.

04

Integrate and implement

Work with your team to integrate the new concepts into your AI strategy, governance, or product documentation. Ensure the conceptual shift actually sticks and improves decision-making.

"The aim of philosophy is to show the fly the way out of the fly bottle."
— Ludwig Wittgenstein, as cited by Scharp

03 — Engagements

Where conceptual engineering
changes outcomes

AI Governance Audits

For organizations deploying AI systems that affect individuals — hiring, lending, healthcare, criminal justice. We audit the conceptual framework underlying your governance policy and identify where the concepts are incoherent, inconsistent, or legally vulnerable.

Typical output: a written report identifying specific conceptual defects, a set of replacement concepts, and a roadmap for implementation. 8–12 week engagement.

Concept Definition Projects

For AI product teams that have hit a conceptual wall — when 'fairness' or 'transparency' or 'explainability' mean different things to different stakeholders and nobody can agree on a definition. We produce a clear, workable, philosophically rigorous definition that your team can actually use.

Typical output: a concept specification document with definitional analysis, scope conditions, and test cases. 4–8 week engagement.

AI Ethics Strategy

For organizations that want to move beyond a checklist approach to AI ethics. We help you build a genuine conceptual foundation for your AI ethics practice — one that can withstand scrutiny from regulators, auditors, and the public.

Typical output: a strategic framework document and a series of facilitated workshops. 3–6 month engagement.

Conceptual Due Diligence

For investors and acquirers evaluating AI companies. Before you commit capital to an AI business, understand the conceptual foundations of their product claims. Are their key concepts coherent? Do they do the work the product claims depend on?

Typical output: a written due diligence report. 2–4 week engagement.

ISO 42001 advisory — Aug 2026 EU AI Act readiness

Now booking for the Aug 2026 EU AI Act readiness window. An 8–12 week ISO/IEC 42001 advisory engagement that builds a defensible AI management system on the conceptual foundation your governance already needs.

→ Visit the ISO 42001 advisory page

EU AI Act advisory — Aug 2026 Annex III readiness

Now booking for the Aug 2026 Annex III enforcement window. An 8–12 week EU AI Act advisory engagement that overlays the AIMS foundation with the five conceptual gaps the Act assumes but never defines.

→ Visit the EU AI Act advisory page

04 — Key publications

The intellectual foundation

OUP 2013

Replacing Truth

Oxford University Press

Scharp's paradigm work in conceptual engineering: the concept of truth is inconsistent, and should be replaced for certain theoretical purposes with a team of better concepts that avoid the liar paradox and related problems.

This same methodology — diagnosing defective concepts and engineering replacements — is what Axiom Advisory applies to AI contexts.

OUP 2019

Semantics for Reasons

Oxford University Press (with Bryan Weaver)

Weds conceptual engineering with the philosophy of language, developing a detailed semantics for reason locutions and showing how the concept of 'reason' is structured — with implications for ethics, epistemology, and rationality.

Synthese, 2021 "Conceptual Engineering for Truth: Aletheic Properties and New Aletheic Concepts"

The technical core of Scharp's approach to conceptual engineering — how AI techniques (including machine learning) can be used to systematically identify replacement concepts for defective ones, and how experimental philosophy validates which replacements actually work.

Oxford University Press, 2020 "Philosophy as the Study of Defective Concepts" — in Conceptual Engineering and Conceptual Ethics

The most accessible introduction to Scharp's framework: many philosophical problems stem from defective concepts, not from the world being confusing. Philosophy should actively replace those concepts rather than merely analyze them. The chapter that defines his intellectual mission.

"This is the most fundamental transformation in social life of our era — akin to those brought on by the Enlightenment, the rise of the natural sciences, the religious wars of the 17th century, or the industrial revolution."
— Kevin Scharp, on AI's societal impact

Organizations navigating the AI transition face a choice: work with the defective concepts they've inherited — or build the conceptual foundations that the next era of technology actually requires.

Axiom Advisory exists to help organizations make that choice well.