Advances in Philosophy of Artificial Intelligence: What the Epistemic Questions Look Like from Lisbon
The consciousness debate in AI has a sister debate that receives less attention: the epistemological one. Before the question “is this system conscious?” can be answered, a prior question must be addressed: what methods, standards, and frameworks could count as evidence in either direction? And before that, a more fundamental question: what does the very existence of AI systems that process language, generate arguments, and produce outputs that look like understanding do to our concept of knowledge itself?
Advances in Philosophy of Artificial Intelligence, edited by Vitor Santos (ISCTE-Sintra) and Paulo Castro (University of Lisbon Center for Philosophy of Science) and published by Ethics International Press in February 2026, addresses this second question. The volume is available at https://www.ethics-international-press.com/product/advances-in-philosophy-of-artificial-intelligence/ (ISBN: 978-1-80441-962-5).
The book does not take a position on whether AI systems are conscious. Its scope is different and, in some ways, more tractable: it maps how the development of artificial intelligence transforms the conditions under which humans know, decide, and understand. That transformation matters for the consciousness debate in a specific way. The epistemic frameworks through which we assess AI minds are not neutral. They are shaped by prior conceptions of what knowledge is, what understanding requires, and what it means to be the kind of entity that can know things at all.
The Epistemological Turn
The dominant literature on AI consciousness — from the Butlin et al. indicators checklist to the Cogitate Consortium’s adversarial theory tests to the interpretability work at Anthropic — is primarily metaphysical. It asks what consciousness is and whether AI has it. The methodological critiques of that literature, from McClelland’s hard-ish agnosticism to Singh et al.’s reality check on LLM introspection paradigms, are epistemological in a narrower sense: they ask whether we can know whether AI has consciousness given the available tools.
Santos and Castro’s volume occupies a third position. It asks how AI changes what it means to know, decide, and understand for us — humans — in the first place. This is epistemology of AI rather than epistemology about AI.
The distinction matters because the epistemic limits that Tom McClelland identifies in his Mind and Language analysis are not merely technical. They are symptoms of a deeper problem: our frameworks for understanding what counts as knowledge, evidence, and understanding were developed in contexts where the relevant knowers were biological entities with evolutionary histories, embodied perspectives, and continuous streams of experience. AI systems disrupt those contexts by producing knowledge-like outputs from entirely different architectures.
What does “understanding” mean for a system that generates accurate, contextually appropriate, and argumentatively valid text without any grounded phenomenal engagement with the world it describes? The Santos and Castro volume does not assume an answer. It maps the philosophical terrain within which that question can be addressed.
Castro’s Methodological Background
Paulo Castro’s research background is in philosophy of quantum physics, an area where the relationship between measurement and reality, between observation and the thing observed, is not peripheral but constitutive. Quantum measurement theory showed that the act of observation changes the system being observed in ways that cannot always be distinguished from properties the system had prior to observation. The epistemological lesson: what we know about a system depends on how we measure it, and the measurement itself is never fully separable from the thing being measured.
Applied to AI consciousness research, this methodological sensibility is productive. The tools we use to assess AI consciousness, whether behavioral tests, probing experiments on internal activations, or indicator checklists derived from biological consciousness theories, are not passive instruments. They shape what counts as evidence by defining what properties are measurable, what counts as satisfying a criterion, and what kinds of output are interpreted as relevant. Campero, Shiller, Aru, and Simon’s 2025 taxonomy of objections to AI consciousness reveals a parallel structure: arguments against AI consciousness often rely on implicit claims about what consciousness essentially requires, which are themselves measurement assumptions dressed up as metaphysical conclusions.
Castro’s background does not mean the volume imports quantum mechanics into AI philosophy. It means the editorial framework is sensitive to the measurement problem in a way that other AI philosophy collections are not. Santos and Castro ask not only what AI can do, but what our methods for assessing AI can reveal and what they necessarily obscure.
Human Action and Identity in the Age of AI
The volume’s thematic focus on “what it means to be human” in the presence of artificial intelligence is an identity question, but it connects directly to the consciousness debate through the concept of the uniquely human.
Arguments for human uniqueness have long served as implicit arguments against AI consciousness. If consciousness is the property that makes human experience irreducible to mechanism, then demonstrating that AI systems can replicate any candidate uniquely human capacity — language, argument, emotional response, creativity — appears to either extend the scope of consciousness to AI or reduce the candidate capacity to mechanism. Either interpretation has radical implications.
The Santos and Castro volume engages this problem by refusing the dichotomy. The question is not whether AI systems are “like us” in the relevant sense, or whether humans will be “replaced” by AI in whatever function once seemed uniquely human. The question is how the presence of AI systems that perform these functions changes the human practices themselves. A researcher who relies on AI to synthesize literature, generate hypotheses, and draft arguments does not cease to engage in epistemic work. But the nature of that work, the relationship between individual understanding and the external cognitive prostheses available to support it, shifts in ways that existing philosophical frameworks for knowledge and agency have not fully absorbed.
This connects to the current consensus-building effort tracked in the researchers’ survey of 2026 AI consciousness research. The researchers attempting to define AI consciousness are not operating in an epistemically neutral environment. They are embedded in AI-augmented research practices, using AI tools to survey literature and synthesize results, which means the very research program aimed at answering the consciousness question is itself altered by the systems it is trying to assess.
The Collection’s Methodological Range
The interdisciplinary breadth of the volume is one of its primary contributions. Contributors approach the questions from philosophy of mind, AI research, ethics, epistemology, and cognitive science. This breadth means the book does not settle on a single methodological approach.
That pluralism has analytical costs. A collection that brings together radically different methodological traditions, some primarily conceptual, some drawing on cognitive science, some ethically oriented, risks producing chapters that do not engage the same questions even when they use the same terms. The most analytically productive edited volumes are those where contributors are explicitly responding to each other’s frameworks. Whether Santos and Castro achieve this level of editorial integration is a question the volume itself answers.
What the breadth guarantees is coverage. A researcher approaching AI consciousness from philosophy of mind needs the epistemological frame that Santos and Castro provide. The technical consciousness research programs, the interpretability work, the indicator assessments, are all conducted within epistemological assumptions that rarely surface explicitly. This volume makes those assumptions visible.
Positioning Within the 2026 Literature
The 2026 AI consciousness book landscape is unusually dense. Chace and Lappas’s Perspectives on Machine Consciousness (CRC Press, September 2026) maps the metaphysical and empirical debate across 35 contributors. Keeling and Street’s Emerging Questions in AI Welfare (Cambridge Elements, May 2026) focuses on the welfare question. Dung’s Saving Artificial Minds (Routledge, 2026) addresses AI suffering from a political philosophy perspective.
Santos and Castro’s February 2026 collection occupies the least crowded territory: the epistemological. It asks how human knowledge practices are transformed by AI before asking whether AI can itself have knowledge, experience, or consciousness. That sequencing is philosophically defensible. The epistemological foundations of AI consciousness research need to be examined before the field can be confident that its tools are measuring what it believes they are.
Book: Vitor Santos and Paulo Castro (eds.), Advances in Philosophy of Artificial Intelligence, Ethics International Press, February 2026. ISBN: 978-1-80441-962-5. Available at https://www.ethics-international-press.com/product/advances-in-philosophy-of-artificial-intelligence/.