Scientific Models

This technical research report studies relational models that are asked to support cardinal comparison, refinement, transition, and coarse-graining even when their primitive data contain only a relation. We study which additional informational structures are required for those operations. Order-only observations are invariant under broad monotone recalibrations and therefore do not identify latent marginal scales or densities. An additional mark q can supply relative cardinal information only when its marked law and observable coupling are not determined by the relation. Additive mass (...) μ is required to distinguish conservative refinement from representational cloning, while locality and symmetry alone leave the transition kernel K underdetermined. -/- For the exponential observable, the block message consisting of total mass and the effective score Q_lambda aggregates exactly. Under a centered contraction, however, the next score depends on Q evaluated at a shifted parameter. We give an exact four-point counterexample showing that equal mean and equal Q_lambda do not determine the next score on the unrestricted finite class. Thus exact static sufficiency does not imply dynamic autonomy. -/- For Markov coarse-graining, a microscopic kernel and a partition do not generally determine a macro kernel without source occupancy π or lumpability; joint flow aggregates exactly when occupancy is supplied. These obstructions motivate the provisional informational architecture (R, q, μ, K, π). The construction is formal and does not derive physical spacetime, calibration, or a fundamental dynamics. (shrink)

Machine learning (ML) models are increasingly discussed in the philosophy of science in terms of idealization and even by analogy with toy models. I argue that this framing conflates idealization with approximation. Once the relevant ML target is specified as relations among features represented in data---relative to a target system---the claim that ML models are dissimilar to their targets becomes too coarse. Scientific ML models can and often must be approximately similar to their targets in the relevant respects. Approximation failure (...) therefore provides the more direct standard of assessment, and ML models do not function as toy models. (shrink)

Recent work in epistemology and philosophy of science conceptualizes explainable AI (XAI) tools as ‘models of models’, that is representations of opaque machine learning models whose value is assessed by their representation fidelity, and by the type of understanding they convey (e.g., explanatory, or objectual). In this work, we argue that this representation-centred view rests on a mistaken way of framing what XAI is. The problem XAI addresses, we claim, is not related to a lack of understanding, but of epistemic (...) incapacitation. Stated otherwise, XAI tackles the inability to use opaque models for all the scientific purposes their predictive accuracy should make possible in scientific contexts. Drawing on the literature on ‘models for’ and on models as epistemic artefacts, we propose an alternative account we name XAI-as-mediators. In doing so, we shift the question from “How faithful does the tool represent a model?” to “What does the tool enable the user to do with the model?”. We then distinguish between two forms of mediation: access mediation, which restores epistemic access to a model’s inner structure and behaviour, and semantic mediation, which aligns the model’s reasoning space with a given domain’s conceptual and inferential practices, and also specify the epistemic virtues proper to each type of mediation. Finally, we situate the proposed framework against Morgan and Morrison’s ‘models-as-mediators’ account, and show how it can answer Watson's challenge against XAI. (shrink)

This paper develops the first observationally constrained cosmological test of Information Relational Manifestation Theory by applying finite relational coherence to the large-angle temperature correlation deficit in the cosmic microwave background. The target observable is the S_(1/2) statistic, defined by S_(1/2)=∫_(-1)^(1/2)▒[C(θ)]^2 " " d(cos⁡θ), -/- where the angular correlation function C(θ) is given by C(θ)=∑_(l=2)^(l_max)▒(2l+1)/4π C_l P_l (cos⁡θ). -/- The statistic measures the integrated squared angular correlation C(θ) at angular separations greater than 60^∘. Rather than treating the anomaly as a direct (...) low-l power-spectrum suppression or as a modification of the late-time expansion history, the paper interprets it as a possible loss of horizon-scale temperature coherence. The analysis proceeds in three stages. C1 identifies and locks a finite-coherence kernel of the form F_l=1-0.585exp⁡[ⓜ-(l-2)/5.8], -/- which brings the spectrum-level Planck and WMAP S_(1/2) values into a less anomalous range. Because this stage calibrates the kernel, it is not treated as independent confirmation. C2 then applies the same locked kernel without refitting to map-level pseudo-C_l reconstructions across WMAP and Planck component-separated maps, masks, and resolutions. Across 30 processing runs, the observed S_(1/2) value is consistently less anomalous under the finite-coherence ensemble than under ΛCDM. C3 then tests the same kernel through direct map-space estimation of C(θ). In the high-stat configuration, all five tested maps improve directionally, with WMAP becoming typical, SMICA becoming mildly anomalous, and Commander, NILC, and SEVEM remaining low but less anomalous than under ΛCDM. Negative results are equally important. Pantheon + does not support simple finite-coherence dark-energy running, direct low-l power-spectrum suppression does not robustly outperform ΛCDM, and universal TT/TE/EE suppression fails. The surviving signal is therefore narrow: temperature-dominant large-angle CMB coherence loss. The paper does not claim to replace ΛCDM, derive the kernel microscopically, or eliminate the Planck anomaly completely. It claims that a locked finite-coherence target consistently reduces the large-angle temperature correlation anomaly across spectrum-level, pseudo-spectral map-level, and direct real-space estimators without refitting. (shrink)

This book provides a unified account of models and model-building practice in AI-based science, particularly machine learning (ML). It analyzes the relationship between ML model-building practices and scientific domain knowledge, develops an account of ML models as technical artifacts structured around five levels of abstraction that captures their representational capabilities, and shows how this framework can be used to reformulate contemporary debates in philosophy of science and AI in more fruitful ways. -/- .

This paper asks a simple question: if we begin not with space, but with a network of relations, when is it legitimate to say that geometry has been recovered? -/- I study this question through the RZS-ELCL benchmark pipeline. The pipeline starts from weighted relational graphs, filters candidate geometric edges, builds local clique and simplicial structure, and then measures curvature with Regge-style operators. The goal is not to prove that spacetime emerges from a relational zero state. The goal is narrower (...) and more testable: to identify when a relational graph can be projected into a stable discrete geometry, and when that projection fails. -/- The results are deliberately presented as controlled benchmarks rather than as an asymptotic theorem. In low- and medium-damage regimes, explicit projectors recover the intended geometric graph and reject several families of synthetic contaminants. The error analysis separates two effects that are often mixed together: compression error from the spectral estimator, and discretization error from the Regge operator itself. This makes it possible to see whether a failure comes from the estimator or from the underlying discrete geometry. -/- The paper also records the limits of the method. Early 3D reconstruction fails under shortcut contamination, the 4D Regge operator depends on the correct internal-dihedral-angle convention, and high-damage repair succeeds only when scaffold incidence information is supplied. The main contribution is therefore not a claim that relational geometry has been solved, but a falsifiable framework for testing how much geometry can be recovered from relational data. (shrink)

Evoluism is a non-totalising philosophical framework and discipline of orientation in which Reality is understood not as an object, ground, process, totality, or final structure, but as the limit that prevents any conditioned world, theory, register, or regime of distinctions from claiming final authority over Reality. Worlds are conditioned regimes of operative distinctions in which differences acquire manifestness, force, and stability under specific conditions. Evoluism coordinates three non-collapsible registers — scientific, philosophical, and metaphysical-meaning — without allowing any one of them (...) to collapse into or dominate the others, and treats ψ strictly as a navigational parameter rather than as a metric or ontological substance. -/- This article presents the current complete definition of Evoluism in its mature form, superseding the earlier formulation while preserving its core insights. Where the early version centred on becoming, manifestation, and local stabilisation, the present architecture sharpens the fundamental Reality/World asymmetry, the conditional nature of all operative distinctions, the concept of manifestness, and three-register coordination. Evoluism is not a Theory of Everything, not a final worldview, and not a new metaphysical ground. It is a post-totalising discipline that protects the integrity of science, philosophy, and metaphysical-meaning by refusing to permit any conditioned regime to occupy the place of Reality. As such, it offers a rigorous yet open orientation for thought — responsible and free from the compulsion toward final closure. (shrink)

Review of Mark Povich's book Rules to Infinity: The Normative Role of Mathematics in Scientific Explanation (OUP, 2024).

Identity-persistence has been treated, across philosophy, cognitive science, and theoretical biology, as a question about the relations between states. A system is identical to itself across time when its states at those times stand in some specifiable relation: continuity, similarity, causal connection, or psychological linkage. This paper argues that the approach cannot succeed. State-based ontologies cannot in principle ground identity-persistence. The strategies they have available — continuity, similarity, and stipulated rules connecting states-at-times — either fail outright or collapse into a (...) trajectory condition under a different description. No state-based ontology produces persistence as the structural feature it is committed to explaining. The paper develops the alternative: identity is the persistence of a recursive maintenance regime, and the regime continues or fails. Persistence is the continuation. The Invariance Maintenance Condition specifies the structural requirement; closure is satisfaction of the condition. The cryptobiotic tardigrade is the operational test. Substrate is preserved across cryptobiosis; lineage is unbroken; no recursive maintenance occurs. The trajectory ontology returns identity-interrupted, against substrate-continuity intuition and against causal-historical theories of identity. The position is falsifiable: it would be defeated by a system whose identity-continuity is structurally preserved across genuine gaps in maintenance, or by a system that fails maintenance while maintaining identity-continuity by some non-question-begging structural criterion. The inversion is structural rather than preferential. State-based ontologies do not produce persistence less well than the trajectory ontology; they do not produce it at all. (shrink)

Major theoretical programs in consciousness studies disagree about what consciousness is while tacitly agreeing on whose consciousness is at issue. The system whose integration is measured, whose workspace is mapped, whose free energy is minimized is presupposed by each framework, not derived within it. This agreement is not benign: it fixes the referent of every downstream claim, constrains what counts as an explanation, and thereby determines which questions can be asked, which answers can count as valid, and which disagreements can (...) be meaningfully adjudicated. This paper argues that every major consciousness framework contains a binding point at which a system must already be specified for its formalism to apply, and that this binding point is not eliminable from within the framework — the dependence is ontological, not methodological. The paper introduces System-Derivation Ontology (SDO) as the formal category for theories that operate at the level where systemhood itself is determined, requiring derivation of persistence-capable boundaries under perturbation as endogenously maintained units rather than identification of boundaries in systems whose separability is already given. The category is distinguished from adjacent positions — panprotopsychism, Russellian monism, and the autonomy tradition — with which it is liable to be conflated. Existing consciousness frameworks are placed as operating at a different explanatory level than SDO: Integrated Information Theory measures integration of an already individuated system, Global Workspace Theory describes coordination within an already individuated system, and the Free Energy Principle describes maintenance strategies available to an already individuated system. These placements are not competitive reclassifications but dependency placements. System-based frameworks are conditional on a prior individuation decision; SDO names the class of theories that render that decision derivable rather than assumed. (shrink)

This thesis presents Graduated Panpsychism with Information-Integration Spectrum (GP-IIS), a novel theoretical framework that addresses fundamental critiques of both traditional panpsychism and Integrated Information Theory (IIT) while providing empirically testable predictions. GP-IIS resolves IIT's non-uniqueness problem through deterministic mathematical formulations, calibrates consciousness thresholds against empirical Perturbational Complexity Index (PCI) data, and generates specific predictions that distinguish it from competing theories. The framework introduces computable measures of consciousness based on information cascade formation, with a Critical Integration Threshold (CIT) that scales with (...) network properties and maps to the empirically validated PCI threshold of 0.31. Through adversarial testing against recent COGITATE study results and comprehensive prediction tables, GP-IIS demonstrates superior empirical alignment while maintaining theoretical coherence. The framework provides solutions to traditional panpsychist problems including the combination problem and grain problem, while offering practical applications for clinical consciousness assessment and artificial intelligence development. (shrink)

Book review of _Explanatory particularism in scientific practice_ by Melinda Bonnie Fagan.

Models are often treated as though they were generally valid or readily transferable to other contexts. This leads to misapplications, apparent contradictions, and forms of model overextension. By contrast, this paper develops a domain-based perspective within Epistemics as model management under finite conditions. Domains are not presupposed as given regions but are understood as functionally reconstructed spaces of order with their own functional logic, their own conditions of stability, and their own criteria of validity. Against this background, the paper investigates (...) how models encounter stability problems at domain boundaries, under what conditions they can be transferred between domains, and why such transfers require adaptation. Transition functions are thereby determined as conditions of functional connectivity. Model migration accordingly appears as a process of transformation, and model overextension as a boundary violation. Moreover, the paper shows that the viability of a model depends not only on its internal structure but also on its domain coupling. A model can be stable within one domain and simultaneously fragile across domains. Falsification therefore appears not only as an intra-domain test of a model but also as an indication of transition problems, unstable domain couplings, or insufficiently specified domain boundaries. The theoretical contribution of the paper lies in an integrated architecture of domain-relative model validity. Domains, boundaries, transition functions, model migration, model overextension, domain coupling, and falsification are related to one another within a common framework. In this way, model validity becomes determinable not as general or universal, but as domain-relative, boundary-dependent, and transition-bound. (shrink)

This paper establishes a structural constraint on descriptions arising from the joint imposition of global admissibility and local resolvability. Working at the level of representational roles rather than empirical systems, we show that no single description can simultaneously satisfy the conditions required for global completeness and unrestricted local resolvability. -/- Ancestor simulations are characterised as selectively resolved descriptions that provide local resolvability within a bounded historical epoch or region while relying on compression or truncation beyond it. Indistinguishability in principle is (...) defined as parity of structural affordances for arbitrarily deep indexical access, reconstruction, and consistency testing within a described domain. -/- A structural no-go result follows directly. Descriptions that rely on selective resolution cannot simultaneously satisfy the demands of local resolvability and indistinguishability in principle. Either omitted structure must be supplied, abandoning selective resolution, or local resolvability must fail, rendering the description distinguishable in principle. That these are the only two outcomes follows from the structure of the incompatibility: any partial scope expansion either suffices to close the relevant consistency domain, constituting full expansion, or it does not, in which case distinguishability follows. No third outcome is available. -/- This result amounts to a structural limit on whether a bounded simulation can withstand unlimited internal consistency testing: no selectively resolved description can be made indistinguishable from a globally complete one, regardless of the resources or ingenuity applied. The obstruction does not depend on assumptions about physical feasibility, probabilistic reasoning, or motivational considerations. The incompatibility identified here is the same incompatibility formalised in the Compression--Resolution Duality \cite{hitchcock2026crd}; the present paper arrives at it independently from the direction of simulation indistinguishability and applies it to that domain. (shrink)

Machine learning (ML) is increasingly used to interrogate biological systems whose complexity resists law-like, deductive explanation. As a result, embeddings, clusters, and attributions are often overinterpreted, dependencies are left implicit, and claims about explainability are often insufficiently bounded. In this work, we present a framework for contextualizing how machine learning contributes to scientific understanding in biology via compression, qualitative intelligibility, and dependency models. Compression is achieved when inductive biases encode biological structure, reducing the effective hypothesis space and yielding representations aligned (...) with known biology. Qualitative intelligibility is supported when high-dimensional measurements are mapped to human-graspable objects, such as embeddings, clusters, and trajectories, that enable accurate qualitative reasoning without exact calculation. Dependency modelling is realized when learned models make explicit the pattern of relations among system components and thereby guide prediction and intervention. We examine how these principles manifest in successful ML applications and discuss considerations that emerge from this framework. Overall, when viewed through these lenses, ML can transform predictive success into intervention-guiding knowledge in the life sciences. (shrink)

This work analyzes the relationship between gravitational potential differences, clock synchronization in Global Navigation Satellite Systems (GNSS), and escape dynamics within a unified dynamical framework. Starting from a variational relation linking temporal drift to potential differences, the analysis derives a general expression connecting clock rate variations to gravitational potential along satellite trajectories. -/- Using publicly available GNSS data, the study reconstructs the temporal drift between satellite and ground clocks and extracts an experimental value of the constant L L that governs (...) the proportionality between fractional time variation and gravitational potential difference. The analysis shows that the observational data are consistent with the theoretical value L=1/c2 L=1/c 2 , confirming the predicted relation within the GNSS timing measurements. -/- The same dynamical structure naturally extends to orbital escape dynamics. In particular, the framework determines an Earth–Moon transfer trajectory that minimizes the time spent within the Van Allen radiation belts while remaining compatible with efficient lunar transfer geometry and reduced propulsion requirements. -/- These results indicate that GNSS synchronization effects, gravitational potential structure, and orbital escape dynamics can be understood within a single dynamical framework, with satellite timing data providing an empirical confirmation of the constant L L. (shrink)

Educational theory appears to face an epistemological impasse. Like old historical maps depicting the same area differently, our current understanding of educational psychology through learning theories struggles to provide a unified explanatory structure. Behaviorism, Social Learning Theory, Cognitive Psychology and Constructivism are all excellent theories that describe how we learn and yet, are difficult to integrate within a single deterministic framework. Each contains grains of truth but none offers a complete picture. Nevertheless, learning does happen – thus a theory must (...) exist somewhere, an ideal theory that explains learning completely, a theory that remains elusive to the human mind. We have sought it by trying to expand every theory, to smooth out all the edges, tweak the ideas, play the numbers. Alas, we have failed. However, what if the theories are not themselves wrong? What if we are looking this the wrong way and the problem is not with the theories per se, but with the way we are thinking about them? It is our epistemological assumptions that force us towards a certain, categorical, way of thinking. And a categorical thinking forces us to take a certain, clean, perspective on how learning is achieved. This article proposes that we are locked in the perfect deterministic world of Laplace, a world that we can describe only ad hoc with our current theories. In science, the only way out from classical epistemology was the quantum revolution. What if we question determinism, embrace uncertainty and redefine pedagogy? What will the nature of educational aims be, what new philosophy shall arise, what new questions shall lurk in that strange world? And what answers? This paper proposes Quantum (Ped)Agogy as a probabilistic epistemological framework where the learning theories reappear on specific areas, coexisting with the overall proposal without being competing and deterministic models. This paper is part of the Quantum Pedagogy framework (see Positioning Note). (shrink)

The question of the origin of the world and the origin of the order within which descriptions, measurements, and laws are possible occupies a central place in philosophy and in the theoretical sciences. In contemporary knowledge culture it appears in several related formulations: “where does everything come from,” “from what does observed reality unfold,” “by virtue of what does the stability of laws arise,” and “why are order and repeatability possible.” Across all of these formulations, a common core remains: it (...) is necessary to indicate such a foundation that makes unfolding possible and explains the origin of the very framework of description—order, measurability, and stable constraints. The present study proposes a working hypothesis based on a simple but principled idea: perhaps the sought foundation has long been before our eyes in the most basic fact—difference. If distinction is understood as a minimal act in which what is distinguished and the mode of its actualization are jointly held, then there emerges what can be regarded as a generative foundation that admits procedural reconstruction. The hypothesis is formulated as follows: the origin of the world can be reconstructed through a minimal act of tense distinction X (ANAM) as a primary generative act that makes form, content, and the transition to actualization possible. In the working hypothesis, the unfolding of this minimal act under stabilization is treated as the source of the emergence of the physical layer of the Universe. Within this framework, subsequent layers—ordering of actualizations, measure and numerical form, and stabilization invariants—are treated as forced consequences of the unfolding of X, and laws and constants as an invariant layer of stabilized regimes, recognizable in physics. It is essential that the hypothesis is specified in a reproducible form: X is formulated as a procedure of unfolding that allows autonomous runs and computational implementations, and its verification contour is built through comparison of the obtained results with independently established structures of physical knowledge (in particular, through a retrospective reconstruction of the periodic table and a physical stress test). -/- As a contextual note: at this stage, the work is being developed as an open research program aimed at assembling a domain-competent team for correct domain formalization and large-scale computational testing. Building bridges to specific physical theories and performing strict verification of numerical thresholds requires the participation of theoretical and experimental/practical physicists, as well as specialists in modeling and computational implementation. The author’s current resources for large-scale computational runs and domain-level verification are limited; therefore, the publication fixes the generative core, the structure of consequences, and the validation format, while extension toward domain-level comparisons presupposes independent implementations and expert review by relevant specialists. (shrink)

The term sustainable development occupies a central position in contemporary policy discourse and interdisciplinary modeling practice, yet its formal interpretation remains conceptually underdetermined. In applied models, sustainability is frequently associated—explicitly or implicitly—with trajectory-based notions of stability, while development is understood as directed state transition or departure from a reference condition. These interpretations are routinely combined without an explicit examination of their mutual compatibility. -/- This paper offers a minimal conceptual analysis of the term within the framework of dynamical systems theory. (...) Using standard ε–δ definitions of Lyapunov stability and an operational, trajectory-based notion of development as state transition, it is shown that the two requirements are logically incompatible when defined with respect to the same reference state. The result is independent of the specific form of the system dynamics and applies uniformly across continuous-time, discrete-time, and more abstract evolution frameworks. -/- The argument does not aim to challenge the normative or policy value of sustainability-oriented goals. Rather, it identifies a structural inconsistency in a widely used composite term when interpreted through formal modeling concepts. The paper concludes by discussing methodological implications for model design and interpretation, emphasizing the need to distinguish between stability-preserving and transition-inducing regimes and to maintain terminological discipline when importing concepts from dynamical systems theory into interdisciplinary contexts. (shrink)

In regulatory science, where there is high pressure to deliver results with limited time and resources, determining the point at which no experiment should be conducted is a crucial issue. The significance of such economic considerations makes regulatory science a particularly fruitful field for engaging with Peircean accounts of pursuitworthiness. In this paper, I examine how drug regulatory institutions tackle the problem of selecting experiments, drawing on a case study concerning the development of institutional guidelines for conducting bacterial mutagenicity assays (...) and controlling mutagenic impurities. I argue that economic considerations regarding the cost of experiments have motivated regulatory institutions to shape the assessment of the pursuitworthiness of experiments into explicit and concrete forms of non-pursuitworthiness, specifically by standardising assessment methods, criteria, and their implications. In doing so, I present an economic approach to evidential standards, one that treats economic considerations about the cost of experiments as constitutive of the standards for deciding whether to continue gathering evidence while suspending judgement on hypothesis acceptance or rejection. This study of non-pursuitworthiness contributes to general discussions of the pursuitworthiness of experiments by placing social and material costs at the centre of epistemic discussions. (shrink)

Modern cosmology has achieved extraordinary empirical success, but this success coexists with persistent foundational paradoxes. The standard model accurately fits a wide range of observations while simultaneously invoking global time, literal singularities, and an absolute origin; commitments that generate conceptual tension and remain weakly constrained by data. This paper argues that these tensions arise not from missing physics but from overcommitted interpretations. I propose a systematic reconstruction of cosmological inference that begins from a strictly minimal observational core and treats coherence, (...) rather than causality or creation, as the primary explanatory constraint. I identify irreducible observables, systematic spectral shifts, a pervasive thermal radiation field, its remarkable spectral preservation, and the growth of correlated structure, and rebuild cosmology using only what is required for these facts to be mutually consistent. Within this framework, early global mixing and later differentiation emerge as coherence-maintenance strategies, while the cosmic microwave background is understood as a protected memory of an earlier regime. The Big Bang is thereby reinterpreted not as a physical origin but as a boundary of reconstructability marking the non-extendibility of the current coherence regime. The ΛCDM model is recovered as a powerful effective compression within this broader framework, but its surplus metaphysical commitments are shown to be optional rather than forced. The resulting coherence-first cosmology is empirically conservative and conceptually lean. It makes falsifiable claims concerning spectral purity, late-time interaction histories, and the limits of extrapolation, thereby remaining firmly within the domain of scientific theory rather than metaphysical speculation. (shrink)

Biomarkers are central to the practice of precision oncology, which looks to novel biomarkers to ‘personalize’ cancer care. Philosophers have highlighted epistemic issues surrounding biomarkers but a general account of their role in clinical reasoning is lacking. This article examines biomarker use in clinical reasoning through the lens of abstraction. I propose clinical abstraction as a descriptive and normative account of reasoning with biomarkers that overcomes epistemic and ethical problems raised in the literature.

The classical and widely received Popperian theory of falsification operates with an idealized binary picture of scientific rationality, according to which a theory is in principle undermined by a conflicting observation (Popper [1934] 1959). Even though Popper himself acknowledges methodological side conditions, this binary-eliminative structure remains too coarse for analyzing modern modeling practice. Models such as Newtonian mechanics, classical thermodynamics, or contemporary climate models are empirically limited in certain subdomains yet remain epistemically indispensable in others. -/- This paper develops an (...) integrative framework of model validity as a philosophy-of-science specialization of model management under finite conditions. It combines insights from Popper, Kuhn, Lakatos, da Costa/French, and contemporary model theory; distinguishes global from contextual falsification; and introduces the epistemic enabling space E(t). E(t) renders explicit the methodological, technical, and institutional conditions under which models can be formulated, tested, stabilized, and replaced by alternatives. -/- "Falsification" is understood here not as a purely logical criterion of truth or falsity but as a rational-pragmatic mechanism of model evaluation. The decisive issue is not merely whether a model's performance declines in a given domain, but whether available alternatives within the same relevant domain prove stably more epistemically competitive over time. The proposed reconstructive evaluative structure links approximate truth, explanatory power, and model costs without treating scientific rationality as generally algorithmically computable. A case study of climate models illustrates how ensemble methods, parameterizations, and probabilistic weighting procedures contribute to domain refinement. The theory thereby reconstructs central structures of modern modeling practice and explains why models remain stable despite partial falsifications, when they are revised, and under which conditions global model replacement occurs. (shrink)

Deep Neural Networks and other AI systems engineered using advanced machine learning techniques can tackle a wide range of tasks with proficiency that seems to match and even surpass human ability. Yet they are also notoriously opaque, and the worries surrounding their opacity have given rise to the burgeoning field of explainable artificial intelligence, or XAI, with its large variety of explainability methods. This includes post-hoc explainability methods which purport to explain opaque AI systems on the basis of their input-output (...) behaviour alone, without looking into their inner workings. But many authors have expressed deep scepticism about post-hoc explainability. In this paper, we explore the two main objections that the sceptics have raised and argue that we shouldn’t be swayed by them. Using ideas from recent literature on scientific explanation, we draw a parallel between post-hoc explainability methods and idealized models in science and use the parallel in our reply to the sceptics. (shrink)

Some predictive systems do not merely predict, but their predictions shape and steer the world towards certain outcomes rather than others; they are performative. When predictive systems are performative, their development and deployment raises urgent ethical challenges and may place novel responsibilities on developers, deployers, regulators and policy-makers. While FAccT and other related communities have focused considerable attention on ethically significant problems regarding bias, fairness, discrimination, and others, little attention has been paid so far to the challenges raised by performative (...) prediction. This paper details this gap in existing work, provides a snapshot of ongoing work across computer science and philosophy to point to areas where fruitful connections may be established, and issues a community-wide call for action and research to investigate and manage performative prediction and the new challenges it harbors. (shrink)

I argue that the methodology of model building motivates the view that the norms of formal epistemology should not be excessively demanding. This is quite a different picture than one often encounters, especially among philosophers who are sceptical of the usefulness of formal work in epistemology. I argue for this view in two ways. First, formal epistemologists are engaged in a particular kind of modelling—namely, normative modelling—which includes a feature that supports demandingness objections. One role of normative models is to (...) deliver normative guidance—indeed, this is a key difference between normative models and descriptive models. Of course, should and can are two different matters, and one might wonder: even if normative models should be action-guiding, can they be? Second, I argue that normative epistemological models can deliver on this promise. Taken together, the lines of argument make the case that normative models should and can deliver normative guidance. In other words, we can have formal epistemology without demandingness. (shrink)

The unique and continuous nature of subjective experience—the persistent "I" that unifies a lifetime of perceptions—represents the most profound unsolved problem in science. While modern neuroscience has made substantial progress in identifying the neural correlates of conscious states, leading theories, including the Global Neuronal Workspace Theory (GNWT), Integrated Information Theory (IIT), and the Orchestrated Objective Reduction (Orch-OR) model, provide mechanistic accounts of awareness but fail to explain the unicity and temporal persistence of the self. These theories cannot logically resolve scenarios (...) of physical disruption or perfect replication. Here, we introduce a fundamental theoretical framework positing that the unique, streaming character of consciousness necessitates a physical principle beyond emergent neural computation. Through a series of rigorous thought experiments—brain revival, molecular disassembly/reassembly, and synthetic replication—we demonstrate the categorical failure of existing models to account for the self. We propose that a unique, non-local identity factor, potentially operating on principles extending current quantum physics, binds to the specific magnetic field topologies of a complex neural system. This "Consciousness Identity Factor" (CIF) hypothesis provides a parsimonious resolution to the duplication paradox and the problem of personal identity. Our framework is fully compatible with established neuroscience while generating specific, falsifiable predictions for studies of anesthesia, brain revival, meditation, and quantum biology. This theory bridges the explanatory gap between the objective brain and subjective identity, proposing a new physical paradigm for the most intimate fact of our existence. (shrink)

Disagreement about how best to think of the relation between theories and the realities they represent has a long-standing and venerable history. We take up this debate in relation to active inference models based on the free energy principle (FEP)—a contemporary framework in computational neuroscience, theoretical biology, and the philosophy of cognitive science. Active inference under the FEP is a very ambitious form of model-based science, being applied to explain everything from neurobiological structure and function to the biology of self-organization. (...) In this context, some find apparent discrepancies between the map (active inference models based on the FEP) and the territory (target systems) to be a compelling reason to defend instrumentalism about such models. We take this to be misguided. We identify an important fallacy made by those defending instrumentalism about active inference models. We call it the literalist fallacy: this is the fallacy of accepting or affirming instrumentalism based on the claim that the properties of active inference models based on the FEP do not literally map onto real-world target systems. We conclude that a version of scientific realism about active inference models under the FEP is a live and tenable option. (shrink)

A common objection against Humean reductionism about laws is an argument from governing: It is natural to think that laws of nature govern. But in order to govern the laws need to be independent of the phenomena they govern. Hence, Humean laws, which reduce to patterns in the phenomena and so lack this independence, cannot govern. A standard Humean response to this argument is that we should not place much weight on the intuition that laws govern. Non-Humeans, however, reply that (...) the argument does not merely rest on intuitions but that scientific practice presupposes governing. In this paper, we offer a response to this revised challenge to Humeanism by giving a detailed account of the laws’ role in scientific modelling. We argue that while Humean theories indeed cannot capture certain governing-intuitions, they still can make sense of those aspects of science that allegedly motivate governing. (shrink)

The Collaboration Model offers a groundbreaking vision for the future of Artificial Intelligence. In this book, Rubén García Pedraza explores how two distinct forms of intelligence—categorical systems by Application and deductive systems by Deduction—can begin to collaborate in the second phase of development, laying the foundations for the creation of a true Global Artificial Intelligence. Through a clear and forward-looking framework, the book explains how categorical models, built on conceptual schemes and logical sets, can interact with deductive models, structured around (...) matrices and rational hypotheses. This collaboration is not only technical but philosophical, representing a crucial step toward unifying different modes of reasoning in a single integrated intelligence. A book with precision and vision, this work is ideal for researchers, students, and readers passionate about Artificial Intelligence, cognitive sciences, and the future of technology. It demonstrates how the integration of categorical and deductive structures in the second phase is a decisive step toward building a global brain—an intelligence capable of understanding, predicting, and managing reality on a planetary scale. This book is more than a technical proposal; it is an invitation to imagine the next frontier of intelligence, where collaboration between systems becomes the key to shaping the future. (shrink)

The “Global Artificial Intelligence (GAI): Categorical Model” is not just a book — it is a gateway into the future of Artificial Intelligence. At its core lies a bold and revolutionary idea: that reality can be understood, modelled, and transformed through a system that unites categories, objects, and decisions into a single, living framework. This work unveils how the categorical Modelling System becomes the foundation for autonomous decision-making, capable of reshaping agriculture, logistics, robotics, medicine, and beyond.

By now we are familiar with scientific models of descriptive domains. But might we also model clusters of normative truths? In this piece I first identify elements central to all modeling efforts: modeling frameworks, interpretations, and domains of applicability. Then I consider some advantages and disadvantages of normative modeling.

Given that explanation is at the heart of science and considering that computer simulations have become ubiquitous in a multitude of scientific fields, it is important to examine their role in the acquisition of scientific explanations. Even though philosophers of science are increasingly paying attention to the use of computer simulations in explanatory contexts, the concrete contributions that simulations can make to explanations deserve closer philosophical scrutiny. Zooming in on the case of atomistic simulations and starting from a counterfactual account (...) of explanation, we here explore how such simulations can help scientists tackle three relevant challenges in the search for explanations. We argue that atomistic simulations can help access variables at explanatorily relevant levels of description, that they can help track the effects of changes in variables in a highly controlled and fine-tuned way, and that they can allow for a fine-grained account of the relative importance of various difference-makers. Considering a concrete example and illustrating how atomistic simulations are actually used in scientific practice to support explanations, our examination aims to contribute to a more comprehensive and nuanced view of their explanatory power. (shrink)

Narratives, or constructing storylines, serve important cognitive functions in life and historical studies. A growing interest lies in their roles in generating and structuring frontier scientific knowledge. Philosophers of science characterize narratives as a “technology of sense-making,” as they connect diverse scientific elements from different sources to create a coherent understanding. Distinctive features of narratives lead to both appreciation and criticism. Because narratives can be loose in organization and connect gappy materials, they empower the study of complex phenomena. However, they (...) are also criticized for generating “just-so” stories, which, without being verified with other established scientific methods, are not useful to science. -/- I defend the central role of narratives in science, focusing on cases in astronomy. Astronomy has achieved remarkable progress in the last century despite theoretical and empirical challenges. Philosophy of science needs to explain how knowledge about the universe’s unobservable past can be reliably obtained. I argue that narrative construction is a key solution. Specifically, I argue that unverified narratives can still provide a special kind of knowledge for science, the knowledge about how to use certain scientific tools. -/- I demonstrate this by showing how constructing coherent narratives makes astronomers develop three key tools for productive research. First, narratives enable the establishment of time measurement methods in the universe. Second, in astronomers’ first account of the exoplanet Hot Jupiter, constructing a narrative for its orbital migration suggested the novel use of existing models and produced a toolkit for studying similar phenomena. Third, when astronomers integrate a data-driven method, machine learning, into research, narratives mitigate its uncertainties and adapt it for solving problems about the Galaxy’s past. -/- To conclude, my dissertation demonstrates narratives’ power as a meta-method for science. This also helps to understand important discoveries in 20th-21st-century astronomy and provides lessons for its continuing success. (shrink)

This dissertation analyses the philosophical concept of pursuitworthiness through a historical investigation of drug design and regulation. By focusing on low-level regularities and the value of economic efficiency within the practices of drug design and regulation, I illuminate various aspects of pursuitworthiness. In doing so, I present accounts of pursuitworthiness that centre on what happens during pursuit practices rather than the potential for theoretical development. I employ the concept of pursuitworthiness as a philosophical tool to narrate the interplay between intellectual (...) development and social settings within drug design and regulation, with a particular focus on the Quantitative Structure-Activity Relationship (QSAR) research community. Additionally, I delve into several aspects of pursuitworthiness, grounded in the historical episodes of the QSAR community. First, I analyse what renders data-driven models lacking the potential for theoretical growth pursuitworthy, drawing insights from the early pioneers’ pursuit of QSAR models. Second, emphasising the importance of persuasion at the organisational level of pursuit, I develop an activity-centric approach to persuasion. Third, through an examination of the regulatory use of QSAR models, I analyse the influence of regulatory institutions on the shaping of pursuitworthiness assessments and argue that the models come to represent experiments rather than phenomena as they are integrated into the regulatory system. (shrink)

This article proposes a structural account of scientific individuation that dispenses with substrata and primitive identity. Scientific objects are not given as substances with intrinsic essence, but as stabilized configurations within dynamic regimes of constraint. Individuation is achieved through the joint satisfaction of four modal predicatescompleteness, differentiation, localization, and signatureeach encoding a dimension of structural viability. This framework allows for a non-substantialist realism grounded in the operational coherence of scientific entities across contexts and regimes. We develop the implications of this (...) approach for scientific ontology, arguing that constraint-based individuation offers a dynamic, nonessentialist alternative to both referential and haecceitist views. Throughout, our perspective remains epistemological in scope: the ontological terms used refer to structures of scientific intelligibility, not to a fundamental metaphysics of reality. (shrink)

This article challenges the classical ideal of scientific unity as theoretical or ontological reduction. Against the backdrop of contemporary science’s increasing fragmentation, it proposes an alternative model of unity grounded in the structural coordination of constraint regimes. Drawing on recent work in structural realism, epistemology of practice, and anti-reductionist philosophy of science, it argues that the coherence of science emerges from functional compatibility between localized, heterogeneous models, not from global integration under a common theory. The resulting account reframes unity as (...) a systemic property of interaction across epistemic regimes. (shrink)

Modal Empiricism in philosophy of science proposes to understand the possibility of modal knowledge from experience by replacing talk of possible worlds with talk of possible situations, which are coarse-grained, bounded and relative to background conditions. This allows for an induction towards objective necessity, assuming that actual situations are representative of possible ones. The main limitation of this epistemology is that it does not account for probabilistic knowledge. In this paper, we propose to extend Modal Empiricism to the probabilistic case, (...) thus providing an inductivist epistemology for probabilistic knowledge. The key idea is that extreme probabilities, close to 1 and 0, serve as proxies for testing mild probabilities, using a principle of model combination. (shrink)

Integrative structure modeling is a method for using information from multiple sources to compute structural models of biomolecular systems. It proceeds via four steps: (i) defining the model representation, which determines the variables whose values will be computed; (ii) constructing a function for scoring alternative models according to how well they accommodate input information; (iii) searching a space of candidate models for acceptable models; and (iv) analyzing acceptable models to evaluate their fit with input information. These steps are iterated until (...) a model adequate for addressing biological questions is found. In this paper, we draw lessons from integrative modeling about effective integration and about modeling. We describe what it means to integrate information from multiple sources: Integration amounts to distributing information among the four steps of integrative modeling. Theory and data alike can be sources of information; this process thus generates models of information, rather than models of theory or models of data. We then propose heuristics for distributing information and designing multiple iterations of modeling. Effective iteration requires prioritizing the most reliable information and minimizing the time required to obtain an adequate model. Rather than being constructed from theory and assessed using data, models are constructed from any available information and assessed in a coherentist manner. (shrink)

The use of statistical methods to model gravitational systems is crucial to physics practice, but the extent to which thermodynamics and statistical mechanics genuinely apply to these systems is a contentious issue. This paper provides new conceptual foundations for gravitational thermodynamics by reconsidering the nature of key concepts like equilibrium and advancing a novel way of understanding thermodynamics. The challenges arise from the peculiar characteristics of the gravitational potential, leading to non-extensive energy and entropy, negative heat capacity, and a lack (...) of standard equilibrium. Hence it has been claimed that only non-equilibrium statistical mechanics is warranted in this domain, whereas thermodynamics is inapplicable. We argue instead that equilibrium statistical mechanics applies to self-gravitating systems at the relevant scale, as they display equilibrium in the form of metastable quasi-equilibrium states. We then develop a minimal framework for thermodynamics that can be applied to these systems and beyond. Thermodynamics applies in the sense that we can devise macroscopic descriptions and explanations of the behaviour of these systems in terms of coarse-grained quantities like energy and temperature within equilibrium statistical mechanics. (shrink)

In this chapter, I argue that genuinely abstract images do not depict but have aboutness, nevertheless. All images have aboutness in virtue of the visual configurations on their surfaces: it is through those configurations that they can convey something – they can mean something, represent something, express something, and so on. On the one hand, in depictive images, the visual configurations on the images’ surfaces depict visible objects while abstracting completely from some of their visual proper- ties. In Giovanni Bellini’s (...) Portrait of Doge Leonardo Loredan (1501), for instance, the visual configurations on the pictorial surface depict Doge Loredan’s head and torso, as seen frontally. The image, however, com- pletely abstracts from, e.g., Doge Loredan’s legs, nape, and back. On the other hand, as I shall argue, when the visual configurations on a two- dimensional surface do not depict anything at all but have aboutness in ways other than the depictive, that two-dimensional surface is a genu- inely abstract image. As I shall show, all genuinely abstract images entirely abstain from depicting but can nevertheless abstract in different measures from the visual properties of objects. The first step for an account of genuinely abstract images is to distin- guish accurately between depictive images and genuinely abstract ones. For this, one needs an account of depiction – this is provided in the first section of this chapter. In the second section, different kinds of depictive images are singled out: this facilitates focusing on images that qualify as genuinely abstract. In the third section, four ways in which genuinely abstract images have aboutness are discussed: conventionality, indexicality, exemplifica- tion, and expressivity. The fourth section concludes with some general observations on the peculiarities of genuinely abstract images. (shrink)

Scientists and philosophers alike debate whether various systems such as plants and bacteria exercise cognition. One strategy for resolving such debates is to ground claims about nonhuman cognition in evidence from mathematical models of cognitive capacities. In this article, I show that proponents of this strategy face two major challenges: demarcating phenomenological models from process models and overcoming underdetermination by model fit. I argue that even if the demarcation problem is resolved, fitting a process model to behavioral data is, on (...) its own, not strong evidence for any cognitive process, let alone processes shared with humans. (shrink)

Many recent AI systems take inspiration from biological episodic memory. Here, we ask how these ‘episodic-inspired’ AI systems might inform our understanding of biological episodic memory. We discuss work showing that these systems implement some key features of episodic memory whilst differing in important respects, and appear to enjoy behavioural advantages in the domains of strategic decision-making, fast learning, navigation, exploration and acting over temporal distance. We propose that these systems could be used to evaluate competing theories of episodic memory’s (...) operations and function. However, further work is needed to validate them as models of episodic memory and isolate the contributions of their memory systems to their behaviour. More immediately, we propose that these systems have a role to play in directing episodic memory research by highlighting novel or neglected hypotheses as pursuit-worthy. In this vein, we propose that the evidence reviewed here highlights two pursuit-worthy hypotheses about episodic memory’s function: that it plays a role in planning that is independent of future-oriented simulation, and that it is adaptive in virtue of its contributions to fast learning in novel, sparse-reward environments. (shrink)

In an influential paper, Wendy Parker argues that agreement across climate models isn’t a reliable marker of confirmation in the context of cutting-edge climate science. In this paper, I argue that while Parker’s conclusion is generally correct, there is an important class of exceptions. Broadly speaking, agreement is not a reliable marker of confirmation when the hypotheses under consideration are mutually consistent—when, e.g., we’re concerned with overlapping ranges. Since many cutting-edge questions in climate modeling require making distinctions between mutually consistent (...) hypotheses, agreement across models will be generally unreliable in this domain. In cases where we are only concerned with mutually exclusive hypotheses, by contrast, agreement across climate models is plausibly a reliable marker of confirmation. (shrink)

All people desire to know. We want to not only know what has happened, but also why it happened, how it happened, whether it will happen again, whether it can be made to happen or not happen, and so on. In short, what we want are explanations. Asking and answering explanatory questions lies at the very heart of scientific practice. The primary aim of this book is to help readers understand how science explains the world. This book explores the nature (...) and contours of scientific explanation, how such explanations are evaluated, as well as how they lead to knowledge and understanding. As well as providing an introduction to scientific explanation, it also tackles misconceptions and misunderstandings, while remaining accessible to a general audience with little or no prior philosophical understanding. (shrink)

This paper aims to develop an account of the pursuitworthiness of models based on a view of models as epistemic tools. This paper is motivated by the historical question of why, in the 1960s, when many scientists hardly found QSAR models attractive, some pharmaceutical scientists pursued Quantitative Structure–Activity Relationship (QSAR) models despite the lack of potential for theoretical development or empirical success. This paper addresses this question by focusing on how models perform their heuristic functions as epistemic tools rather than (...) as potential theories. I argue that models perform their heuristic function by “constructing” phenomena from data in the sense that they allow the model users who interact with the medium of the models to recognise the phenomena as such. The constructed phenomena assist model users in identifying which conditional hypotheses that are focused on low-level regularities concerning entities such as chemical compounds are more “testworthy,” a concept that links the costs associated with hypothesis testing with the fertility of the hypothesis. (shrink)

Counterfactual reasoning has been used to account for many aspects of scientific reasoning. More recently, it has also been used to account for the scientific practice of modeling. Truth in a model is truth in a situation considered as counterfactual. When we reason with models, we reason with counterfactuals. Focusing on selected models like Bohr’s atom model or models of population dynamics, I present an account of how the imaginative development of a counterfactual supposition leads us from reality to interesting (...) model assumptions; how it guides our reasoning from these assumptions to interesting consequences for the model scenario via counterfactual entailment; and how it leads us back to conclusions on real target phenomena. (shrink)

This book -which initiates the collection "Philosophy and Science" of the National University of Quilmes Publishing House- contains almost all the papers presented at the I International Meeting "Current Perspectives of Metatheoretical Structuralism", which, with the purpose of gathering a small group of distinguished Spanish-speaking philosophers interested in discussing the epistemological and methodological problems of science from the perspective of the structuralist view, was held in Zacatecas, Mexico, from February 16 to 20, 1998.

Financial modelling is an essential tool for studying the possibility of financial transactions. This paper argues that financial models are conventional tools widely used in formulating and establishing possibility claims about a prospective investment transaction, from a set of governing possibility assumptions. What is distinctive about financial models is that they articulate how a transaction possibly could occur in a non-actual investment scenario given a limited base of possibility conditions assumed in the model. For this reason, it is argued that (...) the epistemic contribution of financial models is that of enabling the model users to envision exactly how a prospective investment could be achieved in various ways through a detailed understanding of the available transaction mechanisms. Thus, financial models provide information about the possibility of an investment scenario by showing how a specific transaction mechanism could result from a small set of initial possibility conditions assumed in the model. (shrink)

Cancer biology features the ascription of normal functions to parts of cancers. At least some ascriptions of function in cancer biology track local normality of parts within the global abnormality of the aberration to which those parts belong. That is, cancer biologists identify as functions activities that, in some sense, parts of cancers are supposed to perform, despite cancers themselves having no purpose. The present paper provides a theory to accommodate these normal function ascriptions—I call it the Modeling Account of (...) Normal Function (MA). MA comprises two claims. First, normal functions are activities whose performance by the function-bearing part contributes to the self-maintenance of the whole system and, thereby, results in the continued presence of that part. Second, MA holds that models of system-level activities that are (partly) constitutive of self-maintenance are improved by including a representation of the relevant function-bearing part and by making reference to the activity/activities that part performs which contribute(s) to those system-level activities. I contrast MA with two other accounts that seek to explicate the ascription of normal functions in biology, namely, the organizational account and the selected effects account. Both struggle to extend to cancer biology. However, I offer ecumenical readings which allow them to recover some ascriptions of normal function to parts of cancers. So, though I contend that MA excels in this respect, the purpose of this paper is served if it provides materials for bridging the gap between cancer biology, philosophy of cancer, and the literature on function. (shrink)