14065 research outputs found
Sort by
Uncertainties about Link Uncertainty: ML Models as Phenomenological Models
Full text linkThere is much debate regarding the epistemic potentials and limitations of machine learning (ML) models in science, and how best to use them to gain new scientific explanations and understanding. Emily Sullivan has drawn an analogy between ML models and scientific toy models, arguing that until the ‘link uncertainty’ between the model and target system has been reduced, they provide how-possibly explanations of their target phenomena. She takes this link uncertainty to be a significant hindrance to obtaining scientific understanding from ML models, a view which is commonly echoed in the literature. Yet, the exact nature of this uncertainty remains largely unexplored. In an attempt to clarify the uncertainties accompanying ML models, I reconsider the extent to which these models provide how-possibly explanations, and Sullivan’s analogy between toy models and ML models. My conclusion is that Sullivan generally overstates ML models’ role in providing explanations, thereby raising our epistemic expectations towards them beyond what is warranted. Further analysis of the representational and explanatory power of ML models also shows that what really hinders our understanding of the target systems is an uncertainty regarding the causal mechanisms mediating the informational dependencies discovered by the ML model, which I call ‘mechanism uncertainty’. From this, I argue that a better framework for understanding the epistemic role of ML models in science is to see them as phenomenological models. These are empirically grounded models accompanied by a mechanism uncertainty, rather than link uncertainty, which hinders a deeper understanding of the target phenomena
How to Explain Degenerate Mechanisms
Full text linkDegeneracy, the ability of structurally different elements to perform the same function and give rise to the same phenomenon, is believed to be ubiquitous at all levels of mechanisms in neurobiology. Given its biological salience, degeneracy has become an emerging topic in recent scientific literature. In this paper, I will present a new strategy for researchers to offer mechanistic explanations for degenerate mechanisms in the nervous system as complementary to the received new mechanist account. Specifically, I argue that, due to degeneracy, exemplar mechanistic models built using averaging techniques are sometimes beset by the ‘failure of averaging’ problem. To avoid this problem, many researchers opt for an alternative strategy – which is to offer what I call the population-based mechanistic explanations. According to this strategy, a large population of models instead of single exemplar models are generated to capture the real-life variability of biological mechanisms. By examining an example from cellular neuroscience, I offer an account of population-based mechanistic explanations and show that they differ from the ‘ordinary,’ exemplar-based mechanistic explanations by providing different explanatory information. Finally, I argue that the use of population-based mechanistic explanations has implications for the new mechanist philosophy. Specifically, I suggest that the notion ‘how-actually mechanistic models’ needs to be reassessed when degeneracy is taken into account
Rethinking mental representation through the epistemology of modeling
Full text linkProblems associated with classical symbolism need not entail the abandonment of representationalism, as some advocates of action-oriented theories of cognition have concluded. This paper argues that a better theory of mental representation can be achieved by applying the epistemology of scientific modeling to research on mental models. The resulting framework integrates action-oriented approaches to cognition with explicit, combinatorially structured representations. It is empirically far better substantiated than symbolic referentialism and aligns more closely with cognitive psychological research than radical enactivist alternatives
Particles before symmetry
Full text linkThe standard model of particle physics is usually cast in symmetry-first terms. Recently, a geometry-first picture has been proposed, in which the relevant symmetries do not appear explicitly at the ground level of ontology. In this paper I extend this approach to two central mechanisms of the standard model: spontaneous symmetry breaking and the Yukawa coupling, both essential for particles to acquire mass. These reformulations offer alternative explanations cast in purely geometric terms. For example, a particle’s quantum numbers correspond to the internal space it inhabits and to the geometric type of object it is (e.g.\ an -tensor). I argue that a symmetry-first account in terms of principal and associated bundles admits a genuine geometry-first counterpart only when the group’s representation coincides with the automorphism group of the fibre—a condition that cuts the slack tolerated by the symmetry-first view
Pseudo-approaches lead to pseudo-explanations: reply to Corlett et al.
Full text linkCorlett et al. criticise a ‘social turn’ in delusions research according to which paranoia is a result of a dysfunction in social cognition [1]. Instead, they propose that, despite appearances, paranoia is solely the result of alterations to domain-general responses to uncertainty. We appreciate the effort to find a parsimonious explanation, and we agree that domain-general processes play an important role in understanding delusions. However, we reject the characterisation of previous work by us and others and question whether the dichotomies set up by Corlett et al. are helpful
Executable Epistemology: The Structured Cognitive Loop as an Architecture of Intentional Understanding
Full text linkLarge language models exhibit intelligence without genuine epistemic understanding, revealing a fundamental philosophical gap: the absence of epistemic architecture. This paper introduces the Structured Cognitive Loop (SCL) as an executable epistemological framework for emergent intelligence.
Unlike traditional AI research that asks "what is intelligence?" (ontological), SCL asks "under what conditions does cognition emerge?" (epistemological). Situated within contemporary philosophy of mind and cognitive phenomenology, this framework bridges conceptual philosophy and implementable cognition. Drawing on process philosophy, enactive cognition, and extended mind theory, we reconceptualize intelligence not as a possessed property but as a performed process—a continuous loop of judgment, memory, control, action, and regulation.
SCL makes three interrelated contributions. First, it operationalizes philosophical insights into computationally interpretable structures, enabling what we term "executable epistemology"—philosophy as structural experiment. Second, it demonstrates that functional separation within cognitive architecture yields more coherent and interpretable behavior than monolithic prompt-based approaches, with empirical support from controlled agent evaluations. Third, it redefines the measure of intelligence: not representational accuracy but the capacity to reconstruct one's own epistemic state through intentional understanding.
This framework has implications across philosophy of mind, epistemology, and artificial intelligence. For philosophy of mind, it offers a new mode of engagement where theories of cognition can be enacted and tested. For AI, it grounds behavioral intelligence in epistemic structure rather than statistical regularity. For epistemology, it suggests that knowledge is best understood not as truth-possession but as continuous structural reconstruction within a phenomenologically coherent loop.
We situate SCL within debates on cognitive phenomenology, emergence, normativity, and intentionality, arguing that genuine progress requires not larger models but architectures that structurally realize cognitive science principles
Unification and Surprise: On the Confirmatory Reach of Unification
Full text linkThere is no doubt that a theory that is unified has a certain appeal. Scientific practice in fundamental physics relies heavily on it. But is a unified theory more likely to be empirically adequate than a non-unified theory? Myrvold has pointed out that, on a Bayesian account, only a specific form of unification, which he calls mutual information unification, can have confirmatory value. In this paper, we argue that Myrvold’s analysis suffers from an overly narrow understanding of what counts as evidence. If one frames evidence in a way that includes observations beyond the theory’s intended domain, one finds a much richer and more interesting perspective on the connection between unification and theory confirmation. By adopting this strategy, we give a Bayesian account of unification that (i) goes beyond mutual information unification to include other cases of unification, and (ii) gives a crucial role to the element of surprise in the discovery of a unified theory. We illus- trate the explanatory strength of this account with some cases from fundamental physics and other disciplines
Representational Interference and the Limits of Abstract Representation
Full text linkThis paper introduces the Representational Uncertainty Principle (RUP) as a structural account of the limits of representational precision. We argue that as representations become more narrowly defined—by fixing more internal structure—they constrain the integration of perceptual and contextual cues. This often suppresses representational flexibility: the capacity to draw on multiple situational cues to stabilize meaning. When this flexibility is reduced, representational diffraction becomes more prominent: a structural phenomenon in which aspects of a situation are subsumed under a representation that deviates from the expected or standard framing, resulting in ambiguity or tension. Drawing on a structural analogy with quantum mechanics, we treat interference and diffraction as complementary manifestations of how representational content is formed. This framework explains why overly precise representations often fail in contexts that demand sensitivity to subtle variations. We support this account through examples of conceptual ambiguity and apparent contradiction, and by developing a framework that distinguishes between the structuring role of the representational vehicle and the dynamic process of integration that gives rise to content. The RUP thus highlights a structural tension between abstraction, context sensitivity, and the need for orientation within experience
A Single Authoritative List of the World's Species: background and roadmap
No full textEven though (some) species lists are in a state of disorder, they are extensively used by scientists, policymakers and international organizations. This chapter considers two ways of alleviating the problems that taxonomic disorder causes for these users of taxonomy: a 'guided approach' to taxonomic knowledge, and a 'facilitating approach' to taxonomic knowledge. On the guided approach, specialists construct a single authoritative list that all users can use; on the facilitating approach, users have to navigate the complexity of taxonomic knowledge themselves, but have a layer of meta-data to help them do this. We argue that currently the facilitating approach is in place but does not (yet) function optimally. We show that pursuing the guided approach to complement the current facilitating approach can alleviate the limitations of the latter, and suggest how this can be done
Dialogue Concerning the Two Chief Views on Spacetime Rigidity
Full text linkDuring their afternoon walks about the piazza, Sagredo and Salviati discuss the significance of spacetime rigidity: the non-existence of distinct isometries that restrict to the same isometry on an open subset of their domain. Their conversations clarify a number of foundational and philosophical results regarding general relativity