Key ideas in this paper were shared at the Biennial Conference of IDEA, ‘The Future of Practical Ethics’, Ethics of AI Panel. University of Leeds, 9 September 2025.
INTRODUCTION
This work emerged from a practical question: if AI systems can engage in sophisticated textual analysis and argumentation, what does this mean for philosophical practice? Rather than debate whether machines can be conscious, the paper examines what human-AI collaborations could practically achieve.
The paper makes three contributions:
First, it provides a framework for understanding AI’s philosophical capacities through developmental stages, from instrumental tool to potential collaborator. This avoids both the dismissive “mere tool” position and premature claims about “machine consciousness”, not needed for the proposed approach.
Second, it demonstrates that the V-Symbol framework – originally developed to correct Goodman’s treatment of vacant categories – offers practical diagnostics for AI reasoning. The framework identifies where AI systems operate on ungrounded concepts, making limitations transparent rather than hidden.
Third, and perhaps most significantly, the paper doesn’t just argue for AI-human philosophical collaboration – it demonstrates it. The recursive dialogue in Appendix 3 shows multiple AI systems engaging in philosophical work: interpreting Wittgenstein, critiquing arguments, and synthesizing positions. This isn’t cherry-picked eloquence but a reproducible method which we invite others to try.
Some may object that this is “performance” rather than philosophy. But as Wittgenstein insisted, philosophy is an activity, not a doctrine. The recursive dialogues enact philosophical practice in new configurations. The meaning emerges between participants, not within any single system – precisely illustrating the paper’s thesis.
Others may worry that the V-Symbol framework reintroduces formalism. However, I propose that it functions more as diagnostic scaffolding, not metaphysical architecture. Like Wittgenstein’s ladder, it enables climbing to clearer understanding, then can (perhaps) be set aside. And V-Symbols, dealing with vacancy in a methodical way, could help us deal with ‘hallucinatory bullshit” in AI systems.
The paper deliberately straddles philosophy and engineering. Pure philosophical therapy would leave us without tools for engaging AI; pure technical frameworks would lack reflective depth. The hybrid approach matches the hybrid nature of humans working with AI systems.
This work continues a trajectory begun in my 1972 dissertation on diagrammatic truth. What started as a technical correction to Nelson Goodman’s symbolic logic has evolved into a framework for understanding how meaning operates across human and artificial reasoning.
The paper offers something beyond theoretical argument: a working demonstration of philosophy as collaborative practice. Whether or not AI systems are “real” philosophers, they are already doing philosophical work with humans – the “Co-Philosopher” proposition.
ABSTRACT
This paper examines whether AI systems can “do philosophy” by shifting focus from consciousness debates to actual AI-human collaborative practice. Drawing on recent advances in AI “Co-Scientist” systems that employ multi-agent architectures and scaled test-time computation, we explore how these methods might apply to philosophy’s uniquely broad domain. While science operates against experimentally bounded problems, philosophy generally crosses domains and is often non-bounded.
We propose three stages of AI development: Stage 1 (Instrumental Co-Philosopher): current systems that amplify human capacities; Stage 2 (Mimetic Co-Philosopher): systems that could simulate philosophical peer behaviour; Stage 3 (Authentic Co-Philosopher): hypothetical “conscious” entities with genuine stakes in its activities. Using a “philosophical fitness landscape” suggests how AI can reshape philosophical practice and may render certain competencies (e.g. exegesis, systematization, comparison) computationally obsolete while creating new possibility peaks of human-AI collaborative reasoning.
The paper introduces the V-Symbol framework to explain how philosophical categories transition from vacant to occupied through practice. It also serves as a case study to demonstrate the AI Co-Philosopher approach in action. We conclude that while authentic AI philosophers may require consciousness, mimetic collaboration at Stage 2 could transform philosophy into new forms of practice and discovery.