Herman Cappelen

This is the most comprehensive book ever published on philosophical methodology. A team of thirty-eight of the world's leading philosophers present original essays on various aspects of how philosophy should be and is done. The first part is devoted to broad traditions and approaches to philosophical methodology (including logical empiricism, phenomenology, and ordinary language philosophy). The entries in the second part address topics in philosophical methodology, such as intuitions, conceptual analysis, and transcendental arguments. The third part of the book is devoted to essays about the interconnections between philosophy and neighbouring fields, including those of mathematics, psychology, literature and film, and neuroscience.

Philosophy is Perfect offers a bold defence of philosophy against its pessimistic and moderate critics. Where “flagellants” see futility and “hedgers” see only piecemeal progress, Herman Cappelen advances a model of Hyper-Optimism: the claim that philosophy has been massively successful, foundational to all inquiry, and flourishing precisely because of its characteristic disagreement. The book unfolds in four parts. Part I argues that all disciplines ultimately “answer to philosophy” when their core concepts and methods are scrutinized, showing that pessimism about philosophy would undermine confidence in every branch of knowledge. Part II turns the standard critique on its head, arguing that persistent disagreement is not a defect but a structural virtue of philosophical inquiry: convergence would signal decay, not progress. Part III presents the positive thesis—that philosophy has already found the correct answers to the “Very Important Questions.” Drawing on the idea of a “nearly filled logical space,” Cappelen argues that after centuries of systematic exploration, the correct answers are already articulated and defended in the philosophical commons. Part IV addresses objections, distinguishing achievements from aims, resisting the retreat to mere “mapping,” and reframing methodological criticisms as philosophical contributions in their own right. The result is a provocative re-evaluation of philosophy’s status. Rather than a discipline in crisis, Cappelen depicts philosophy as uniquely successful: it has discovered foundational truths, thrives on principled disagreement, and sets the terms for all other forms of inquiry. Philosophy is Perfect is both a challenge to entrenched pessimism and a constructive vision of philosophy’s achievements and future.

__Insensitive Semantics_ is an overview of and contribution to the debates about how to accommodate context sensitivity within a theory of human communication, investigating the effects of context on communicative interaction and, as a corollary, what a context of utterance is and what it is to be in one._ Provides detailed and wide-ranging overviews of the central positions and arguments surrounding contextualism Addresses broad and varied aspects of the distinction between the semantic and non-semantic content of language Defends a distinctive and explanatorily powerful combination of semantic minimalism and speech act pluralism Confronts core problems which not only run to the heart of philosophy of language and linguistics, but which arise in epistemology, metaphysics, and moral philosophy as well.

But the sort of context sensitivity exhibited in such sentences does not compromise the claim that meaning determines truth conditions, since recourse to context here is directed and restricted by conventional meaning alone. Anyone who understands sentence (2) knows that its utterances are true just in case whatever object is demonstrated in the context of utterance is nice; and he also knows that any utterance of (2) says of, or expresses about, whichever object is demonstrated that it’s nice. (Similarly, anyone who understands (3) knows that any utterance of it is true just in case whoever utters it has eaten. And every utterance says of, or expresses about, the speaker that he or she has eaten.) In sum, according to the thesis that meaning determines truth conditions, (indicative) sentences divide into two classes – those with truth conditions tout..

__Insensitive Semantics_ is an overview of and contribution to the debates about how to accommodate context sensitivity within a theory of human communication, investigating the effects of context on communicative interaction and, as a corollary, what a context of utterance is and what it is to be in one._ Provides detailed and wide-ranging overviews of the central positions and arguments surrounding contextualism Addresses broad and varied aspects of the distinction between the semantic and non-semantic content of language Defends a distinctive and explanatorily powerful combination of semantic minimalism and speech act pluralism Confronts core problems which not only run to the heart of philosophy of language and linguistics, but which arise in epistemology, metaphysics, and moral philosophy as well.

__Insensitive Semantics_ is an overview of and contribution to the debates about how to accommodate context sensitivity within a theory of human communication, investigating the effects of context on communicative interaction and, as a corollary, what a context of utterance is and what it is to be in one._ Provides detailed and wide-ranging overviews of the central positions and arguments surrounding contextualism Addresses broad and varied aspects of the distinction between the semantic and non-semantic content of language Defends a distinctive and explanatorily powerful combination of semantic minimalism and speech act pluralism Confronts core problems which not only run to the heart of philosophy of language and linguistics, but which arise in epistemology, metaphysics, and moral philosophy as well.

The development of externalism—the thesis that mental and linguistic content is determined by factors outside the agent—was a landmark achievement in 20th-century philosophy. Yet, contemporary debates about Large Language Models (LLMs) are dominated by a resurgent, often unreflective, internalism that seeks to locate meaning "under the hood" in model weights and architecture. This paper challenges this internalist paradigm. First, it extends classic externalist frameworks (from Kripke, Putnam, Evans, etc.) to non-human systems like LLMs through a proposed method of "de-anthropocentric abstraction." Second, it introduces and defends hyperexternalism: the radical thesis that external factors determine not only the content of an LLM's outputs, but also its structural and procedural properties. We argue that whether an LLM is genuinely following a rule (e.g., modus ponens), whether its representations are compositional, or whether an internal state functions as a belief versus a supposition, are all facts grounded in the model's training history, functional role, and relational embedding. If we are right, purely internalist approaches, such as Mechanistic Interpretability, are philosophically misguided. The Hyper-externalist manifesto provide a new framework for understanding AI and outlines a research program that reframes interpretability as an external, relational, and dynamic problem.

ABSTRACT This article challenges conventional boundaries between human and artificial cognition by examining introspective capabilities in large language models (LLMs). Although humans have traditionally been considered unique in their ability to reflect on their own mental states, we argue that LLMs may not only possess genuine introspective abilities but potentially excel at them compared to humans. We discuss five objections to machine introspection: (1) the lack of direct routes to self‐knowledge in training data, (2) the conflict between static knowledge and dynamic mental states, (3) the distorting effects of reinforcement learning on self‐reports, (4) LLMs own denials of inner experience, and (5) arguments that LLMs simply mimic language without understanding. We think all these arguments fail and that there are deep parallels between human and machine introspection. Most provocatively, we propose that LLMs superior processing capabilities and pattern recognition may enable them to develop more sophisticated theories of mind than humans possess, potentially making them more reliable introspectors than their creators. If we are right, this has significant implications for artificial intelligence (AI) alignment, transparency, and our understanding of the nature of AI.

This paper presents an argument that certain AI safety measures, rather than mitigating existential risk, may instead exacerbate it. Under certain key assumptions - the inevitability of AI failure, the expected correlation between an AI system's power at the point of failure and the severity of the resulting harm, and the tendency of safety measures to enable AI systems to become more powerful before failing - safety efforts have negative expected utility. The paper examines three response strategies: Optimism, Mitigation, and Holism. Each faces challenges stemming from intrinsic features of the AI safety landscape that we term Bottlenecking, the Perfection Barrier, and Equilibrium Fluctuation. The surprising robustness of the argument forces a reexamination of core assumptions around AI safety and points to several avenues for further research.

Since the release of ChatGPT, there has been a lot of debate about whether AI systems pose an existential risk to humanity. This paper develops a general framework for thinking about the existential risk of AI systems. We analyze a two-premise argument that AI systems pose a threat to humanity. Premise one: AI systems will become extremely powerful. Premise two: if AI systems become extremely powerful, they will destroy humanity. We use these two premises to construct a taxonomy of ‘survival stories’, in which humanity survives into the far future. In each survival story, one of the two premises fails. Either scientific barriers prevent AI systems from becoming extremely powerful; or humanity bans research into AI systems, thereby preventing them from becoming extremely powerful; or extremely powerful AI systems do not destroy humanity, because their goals prevent them from doing so; or extremely powerful AI systems do not destroy humanity, because we can reliably detect and disable systems that have the goal of doing so. We argue that different survival stories face different challenges. We also argue that different survival stories motivate different responses to the threats from AI. Finally, we use our taxonomy to produce rough estimates of ‘P(doom)’, the probability that humanity will be destroyed by AI.