Amir Konigsberg

In 1950, Alan Turing proposed replacing the question "Can machines think?" with a behavioral test: if a machine's outputs are indistinguishable from those of a thinking being, the question of whether it truly thinks can be set aside. This paper argues that Turing's move was not only a pragmatic simplification but also an epistemological commitment, a decision about what kind of evidence counts as relevant to intelligence attribution, and that this commitment has quietly constrained AI research for seven decades. We trace how Turing's behavioral epistemology became embedded in the field's evaluative infrastructure, rendering the field structurally unable to pose a class of questions about process, mechanism, and internal organization that cognitive psychology, neuroscience, and related disciplines learned to ask. We draw a parallel to the behaviorist-to-cognitivist transition in psychology: just as psychology's commitment to studying only observable behavior prevented it from asking productive questions about internal mental processes until that commitment was abandoned, AI's commitment to behavioral evaluation prevents it from distinguishing between systems that achieve identical outputs through fundamentally different computational processes, a distinction on which intelligence attribution depends. We argue that the field requires an epistemological transition comparable to the cognitive revolution: not an abandonment of behavioral evidence, but a recognition that behavioral evidence alone is insufficient for the construct claims the field wishes to make. We articulate what a post-behaviorist epistemology for AI would involve and identify the specific questions it would make askable that the field currently has no way to ask.

In 1950, Alan Turing proposed replacing the question "Can machines think?" with a behavioral test: if a machine's outputs are indistinguishable from those of a thinking being, the question of whether it truly thinks can be set aside. This paper argues that Turing's move was not only a pragmatic simplification but also an epistemological commitment, a decision about what kind of evidence counts as relevant to intelligence attribution, and that this commitment has constrained AI research for seven decades. We trace how Turing's behavioral epistemology became embedded in the field's evaluative infrastructure, rendering the field structurally unable to pose a class of questions about process, mechanism, and internal organization that cognitive psychology, neuroscience, and related disciplines learned to ask. We draw a parallel to the behaviorist-to-cognitivist transition in psychology: just as psychology's commitment to studying only observable behavior prevented it from asking productive questions about internal mental processes until that commitment was abandoned, AI's commitment to behavioral evaluation prevents it from distinguishing between systems that achieve identical outputs through fundamentally different computational processes, a distinction on which intelligence attribution depends. We argue that the field requires an epistemological transition comparable to the cognitive revolution: not an abandonment of behavioral evidence, but a recognition that behavioral evidence alone is insufficient for the construct claims the field wishes to make. We articulate what a post-behaviorist epistemology for AI would involve and identify the specific questions it would make askable that the field currently has no way to ask.

The rapid integration of large language models into everyday cognitive tasks has created a need for conceptual frameworks adequate to the cognitive consequences of delegating thinking to AI systems. Existing constructs in psychology and also in epistemology, including critical thinking, metacognition, intellectual autonomy, and epistemic agency, each address related phenomena but none adequately captures the specific capacity threatened by habitual AI-assisted cognition, which I define as the ability to remain the genuine author of one's own understanding. This paper introduces cognitive sovereignty as a distinct construct, defined as the capacity to (a) notice when one's thinking is being displaced, (b) maintain a meaningful connection to how one's beliefs and judgments are formed, and (c) distinguish between genuine reasoning and the subjective impression of having reasoned. I trace the concept's philosophical lineage, engage with the extended mind objection, differentiate cognitive sovereignty from adjacent constructs through systematic comparison, and present a growing body of empirical evidence that motivates the construct. The paper argues that cognitive sovereignty names a phenomenon that existing constructs individually fail to capture and that its articulation is a prerequisite for empirical research on AI's impact on human thinking.

Aumann's Agreement Theorem (1976) establishes that two Bayesian rational agents with common priors and common knowledge of each other's posterior beliefs cannot agree to disagree. Their posteriors must coincide. This paper applies Aumann's framework to AI agents built on large language models (LLMs), a domain in which the theorem's conditions appear, at first glance, to be unusually well satisfied. LLMs trained on overlapping data are often assumed to share something like common priors, and in multi-agent protocols their outputs are shared between participants. Yet LLM-based agents routinely produce divergent outputs on identical inputs, and multi-agent systems built from them are increasingly deployed in debate, deliberation, and consensus protocols that implicitly treat this divergence as epistemically meaningful. We argue that Aumann's theorem fails to apply to these agents not because the prior or rationality conditions are violated in the familiar ways they are violated for humans, but for a more fundamental reason: LLMs do not possess beliefs in the sense the theorem requires. Their outputs are samples from conditional probability distributions over token sequences, not reports of posterior probabilities conditioned on private information. We formalize the distinction between genuine disagreement, which carries epistemic content because it signals the existence of unshared evidence, and what we term pseudo-disagreement, which has the surface form of disagreement but arises from stochastic variation in generation processes that lack epistemic states. We show formally that pseudo-disagreement does not satisfy the informational conditions that make genuine disagreement epistemically valuable, and we trace the implications for multi-agent debate protocols, consensus methods, LLM-as-judge paradigms, and the broader practice of treating AI outputs as bearing on questions of truth. Our analysis applies specifically to autoregressive language models and the multi-agent systems built from them; AI systems with fundamentally different architectures, such as those maintaining explicit world models or calibrated Bayesian uncertainty estimates, may require separate treatment.

Rational aesthetic deference becomes apparent when one person’s aesthetic belief gives another person a reason to move his own aesthetic belief in the direction of the other person. It occurs when one person’s aesthetic belief gives another person a normative reason to move your belief in the direction of mine, on epistemic grounds. In such a case, what the first person believes also provides a justification for the second person’s aesthetic belief. This kind of justification is an indirect justification because it is based on reasons that merit deferring to someone else’s judgment, rather than on reasons that support..

Contributors to the recent disagreement debate have sought to provide a uniform response to cases in which epistemic peers disagree about the epistemic import of a shared body of evidence, no matter what kind of evidence they are disagreeing about. The varied cases addressed in the literature have included examples of disagreement about restaurant bills, court verdicts, weather forecasting, chess, morality, religious beliefs, and even disagreements about philosophical disagreements. The equal treatment of these varied cases has motivated the search for a uniform response to peer disagreement wherever it is encountered. In this article I challenge this prevalent approach in the literature. I grant the notion of epistemic peer and accept that being a peer may amount to the same thing in different domains; nonetheless I contend that different domains appear to call for different responses to disagreement. I argue that the appropriate response to finding out about a disagreement with a peer is different in different domains.

In this paper I present a criticism of Sarah Moss‘ recent proposal to use scoring rules as a means of reaching epistemic compromise in disagreements between epistemic peers that have encountered conflict. The problem I have with Moss‘ proposal is twofold. Firstly, it appears to involve a double counting of epistemic value. Secondly, it isn‘t clear whether the notion of epistemic value that Moss appeals to actually involves the type of value that would be acceptable and unproblematic to regard as epistemic.