The QBist and Relational Quantum Mechanics (RQM) informational readings of quantum theory have been developed across two decades without a clear position on what kind of physical system qualifies as an "agent" or "observer" for the formalism. Fuchs, Mermin, and Schack write as if the agent is a human physicist or a generic Bayesian; Rovelli writes as if any physical system can play the relational-observer role; Healey deflates the agent into an abstract Bayesian without specifying its substrate; the recent Khrennikov-Schack-Zwirn intersubjectivity exchange sharpens the question without resolving it. This paper argues that the QBism/RQM informational reading is substrate-flexible: any physical system whose input-output statistics admit characterization through quantum-probability structure with non-trivial Contextuality-by-Default (CbD) signatures resistant to simplex-embeddable ontological models is a candidate epistemic agent for the formalism. Substrate flexibility is the most coherent reading of the shared formal commitments of QBism and RQM once the agent role is separated from historically human-centered examples; the non-triviality requirement is necessary but not sufficient for agency, which additionally requires an input-output architecture capable of state-sensitive updating across measurement contexts. The thesis preserves the QBist objection to view-from-nowhere framings while removing the requirement that agents be human or conscious; it disciplines Rovelli's "any physical system" claim by indexing it to the non-triviality requirement; and it specifies what would count as evidence for or against. Engineered cortical wetware preparations (Cortical Labs CL1, DishBrain) provide a non-human, non-conscious-in-any-unambiguous-sense, controllable testbed on which the question can be empirically pursued; nothing in the argument requires attributing phenomenal consciousness or quantum-coherent biological dynamics to such systems. The paper distinguishes substrate flexibility from Pienaar's prior extension of the QBist agent (which extends the agent's senses, not its substrate), engages the neo-Everettian opposition (Wallace 2012, 2023) directly, and rejects the recent attempts (Edwards 2024, 2025) to fold classical large language models into a QBism-grounded formalism. Classical AI architectures admit simplex-embeddable models for their token-generation processes and therefore fail the non-triviality requirement; substrate flexibility is narrower, not wider, than such proposals.
The QBist and Relational Quantum Mechanics (RQM) informational readings of quantum theory have been developed across two decades without a clear position on what kind of physical system qualifies as an "agent" or "observer" for the formalism. Fuchs, Mermin, and Schack write as if the agent is a human physicist or a generic Bayesian; Rovelli writes as if any physical system can play the relational-observer role; Healey deflates the agent into an abstract Bayesian without specifying its substrate; the recent Khrennikov-Schack-Zwirn intersubjectivity exchange sharpens the question without resolving it. This paper argues that the QBism/RQM informational reading is substrate-flexible: any physical system whose input-output statistics admit characterization through quantum-probability structure with non-trivial Contextuality-by-Default (CbD) signatures resistant to simplex-embeddable ontological models is a candidate epistemic agent for the formalism. Substrate flexibility is the most coherent reading of the shared formal commitments of QBism and RQM once the agent role is separated from historically human-centered examples; the non-triviality requirement is necessary but not sufficient for agency, which additionally requires an input-output architecture capable of state-sensitive updating across measurement contexts. The thesis preserves the QBist objection to view-from-nowhere framings while removing the requirement that agents be human or conscious; it disciplines Rovelli's "any physical system" claim by indexing it to the non-triviality requirement; and it specifies what would count as evidence for or against. Engineered cortical wetware preparations (Cortical Labs CL1, DishBrain) provide a non-human, non-conscious-in-any-unambiguous-sense, controllable testbed on which the question can be empirically pursued; nothing in the argument requires attributing phenomenal consciousness or quantum-coherent biological dynamics to such systems. The paper distinguishes substrate flexibility from Pienaar's prior extension of the QBist agent (which extends the agent's senses, not its substrate), engages the neo-Everettian opposition (Wallace 2012, 2023) directly, and rejects the recent attempts (Edwards 2024, 2025) to fold classical large language models into a QBism-grounded formalism. Classical AI architectures admit simplex-embeddable models for their token-generation processes and therefore fail the non-triviality requirement; substrate flexibility is narrower, not wider, than such proposals.
The QBist and Relational Quantum Mechanics (RQM) informational readings of quantum theory have long lacked a clear account of what kind of physical system qualifies as an “agent” or “observer.” Fuchs, Mermin, and Schack often treat the agent as a human physicist or Bayesian decision-maker; Rovelli suggests that any physical system may serve as a relational observer; Healey abstracts the agent into a Bayesian role without specifying its substrate; and the Khrennikov-Schack-Zwirn exchange sharpens, but does not resolve, the intersubjectivity problem. This paper argues that the QBism/RQM informational reading is best understood as substrate-flexible: any physical system whose input-output statistics exhibit quantum-probability structure, including non-trivial Contextuality-by-Default (CbD) signatures resistant to simplex-embeddable ontological models, is a candidate epistemic agent. This condition is necessary but not sufficient: agency also requires state-sensitive updating across measurement contexts. The argument preserves QBism’s rejection of view-from-nowhere objectivity while removing the assumption that agents must be human or conscious, and disciplines Rovelli’s “any physical system” claim by tying observerhood to measurable non-triviality. Engineered cortical wetware systems such as CL1 and DishBrain offer controllable, non-human, and not clearly conscious testbeds for evaluating this claim, without requiring phenomenal consciousness or quantum-coherent biological dynamics. The paper distinguishes substrate flexibility from Pienaar’s extension of the QBist agent, which expands the agent’s senses rather than its substrate; engages neo-Everettian objections; and rejects Edwards’s attempts to fold classical large language models into QBist formalism. Classical AI token-generation systems remain simplex-embeddable and therefore fail the non-triviality requirement. Substrate flexibility is thus narrower than proposals that include current AI systems.
This article explores the relationship between Kantian epistemology and QBism, a modern interpretation of quantum mechanics. QBism reconceives scientific objects not as independent entities but as phenomena synthesized through the observer's active role. It emphasizes the interdependence of knowledge and observation, rejecting the notion of a pre-existent, absolute reality. Much like Immanuel Kant's assertion that the thing-in-itself lies outside human cognition, QBism denies ontological status to the unobservable or unexperienced, grounding objectivity in intersubjectivity: the shared cognitive structures of human agents. Within this Critical framework, I address an objection raised by Carlo Rovelli, who critiques QBism for anchoring reality to the observer, claiming it risks falling into idealism by positioning the subject outside nature. Rovelli defends a relational view, situating the observer within the world as describable by the same physical laws that apply to all objects. In response, I argue that the QBist observer represents a universal human subject, not a single determinate agent and that intersubjective agreement ensures objectivity without collapsing into solipsism.
Abstract This paper develops an agent-centric account of measurement that treats the preferred-basis problem as fundamentally perspectival. On this view, the system–apparatus–environment decomposition and the observables that are apt to become classically robust are determined by the physical constitution and epistemic constraints of an embodied class of agents. Decoherence then stabilises those agent-specified observables, yielding facts that are stable for us without positing an absolute, observer-independent basis. On this picture, ‘measurements’ are public not because they are metaphysically privileged, but because agents like us share the relevant sensorimotor and operational structure. I motivate this account through a discussion of two recent no-go results for relational quantum mechanics (RQM) [1, 2], and a subsequent response [3]: my aim is not to defend RQM per se, but to refine the relational insight with a principled account of basis selection rooted in embodiment. I provide a phenomenological gloss, drawing on body-schema considerations, to argue that quantum mechanics is best understood as an idiosyncratically human description of interactions with the physical world—a structurally constrained, agent-indexed framework within which classicality emerges.
Abstract The approach to quantum mechanics that we present in this book aligns fairly well with the QBist position in quantum foundations, which takes quantum mechanics as describing the interaction of an agent with the world: A theory mainly offers a way for the agent to manage her expectations about the outcomes of her interactions, not a picture of what nature is really like. We introduce QBism briefly and situate the framework within a range of options for integrating agency and quantum mechanics. We agree with QBism on many fundamental issues, but we do not subscribe to QBism wholesale. We agree with QBism’s pragmatic outlook that takes scientific theories to be tools for agents to use, and with the QBist goal of developing a participatory realism. But we are critical of the particular idealisations that QBism in its present form subscribes to. Most importantly, we are critical of the structural choice not to resolve the agent nor the world in any detail. In our view, scientific agents learn by modelling. Models are tools that agents can use and develop in dealing with the world. Guided by concrete examples, we describe a scientific modelling framework that allows for loops of three different degrees of difficulty and complexity. These range from quantum state updating, which neatly fits the QBist approach of working within a fixed probability space, via a broader class of updates including so-called Hamiltonian learning, up to the most difficult cases in which true scientific creativity is at play. All three levels of difficulty need to be acknowledged for a full picture of real scientific agency. At the end of the chapter, we describe the project of modelling scientific agency itself, which leads to the topics of Chaps. 5 and 6 .
An autoregressive language model, read at the token-commit step, is a near-fully-inspectable classical bounded-inference substrate: the competing-route distribution over candidate tokens — the live propensity before it resolves into one committed answer — is read directly from the output logits. Used as a measurement model, it lets each structural signature of quantum measurement be tested one by one, and the genuinely-quantum residue isolated by its absence. The substrate reproduces the phenomenology of quantum measurement from classical self-conditioning — back-action, non-commutativity (order effects), entropy-collapse on commitment, a substantial Busemeyer–Wang QQ-structure (order-effect difference q not distinguishable from the parameter-free quantum value 0; pooled across eight de-saturated base models over five vendors and four architectures, N=96, q=-0.003, 95% CI [-0.025, +0.019]), interference, einselection, an Aharonov-style weak-versus-strong asymmetry, and measurement-mode dependence — but not the quantum core: genuine contextuality is robustly absent (an analog Contextuality-by-Default cyclic-4 analysis gives CNT < 0 in every model and scenario), and no phase/coherence is demonstrated. A controllable analog Bell-loophole testbed makes each premise's price visible within one substrate (S_analog ~ 2.00 with all loopholes closed; ~ 2.8 on opening locality or measurement-independence), and a within-substrate timeline reads the propensity resolving across network depth, including a maximally-contested case whose outcome is never created. The order-effect incompatibility is shown to be a measured, correctly-signed classical-complementarity dial that falls as the readout partition refines. Positioning. The contribution is the demarcation itself, with statistics, made possible by an inspectability no veiled (physical or psychological) substrate allows. It instantiates beim Graben & Atmanspacher's classical complementarity and the Busemeyer–Wang / Pothos quantum-cognition program empirically; it makes no claim about physical reality. Companion papers (Friction Theory series): Paper 1 — Friction as the cost of probabilistic computation (10.5281/zenodo.20012654); Paper 10 — Race all the way down/up (10.5281/zenodo.20014567); Paper 21 — LLMs as a measurement model for bounded-decision cognition (10.5281/zenodo.20562416); Paper 28 — Dread without a dreader (10.5281/zenodo.20690747).Paper 18 in the Friction Theory paper-series. Submitted as a careful, honestly-scoped empirical demarcation (target venue: Entropy, Quantum Information / quantum-cognition & foundations-of-measurement). v2 (August 2026) — prior-art revision. The paper describes competition resolving to a commitment and cited none of the tradition that formalised it. Luce (1959) for the choice axiom the softmax implements, Ratcliff (1978) for evidence accumulated to a bound, and Usher and McClelland (2001) for the leaky competing accumulator are now credited at the point where the reading is made, with the limit stated: the estimators here are computed on a language model's output distribution, and the parallel is why the measurement was worth making rather than evidence of a shared mechanism. Editorial pass. Earlier versions remain in the version history.