Deep neural networks have become increasingly successful in applications from biology to cosmology to social science. Trained DNNs, moreover, correspond to models that ideally allow the prediction of new phenomena. Building in part on the literature on ‘eXplainable AI’, I here argue that these models are instrumental in a sense that makes them non-explanatory, and that their automated generation is opaque in a unique way. This combination implies the possibility of an unprecedented gap between discovery and explanation: When unsupervised models (...) are successfully used in exploratory contexts, scientists face a whole new challenge in forming the concepts required for understanding underlying mechanisms. (shrink)

Several philosophical issues in connection with computer simulations rely on the assumption that results of simulations are trustworthy. Examples of these include the debate on the experimental role of computer simulations :483–496, 2009; Morrison in Philos Stud 143:33–57, 2009), the nature of computer data Computer simulations and the changing face of scientific experimentation, Cambridge Scholars Publishing, Barcelona, 2013; Humphreys, in: Durán, Arnold Computer simulations and the changing face of scientific experimentation, Cambridge Scholars Publishing, Barcelona, 2013), and the explanatory power of (...) computer simulations :277–292, 2008; Durán in Int Stud Philos Sci 31:27–45, 2017). The aim of this article is to show that these authors are right in assuming that results of computer simulations are to be trusted when computer simulations are reliable processes. After a short reconstruction of the problem of epistemic opacity, the article elaborates extensively on computational reliabilism, a specified form of process reliabilism with computer simulations located at the center. The article ends with a discussion of four sources for computational reliabilism, namely, verification and validation, robustness analysis for computer simulations, a history of successful implementations, and the role of expert knowledge in simulations. (shrink)

What does it mean to trust the results of a computer simulation? This paper argues that trust in simulations should be grounded in empirical evidence, good engineering practice, and established theoretical principles. Without these constraints, computer simulation risks becoming little more than speculation. We argue against two prominent positions in the epistemology of computer simulation and defend a conservative view that emphasizes the difference between the norms governing scientific investigation and those governing ordinary epistemic practices.

It is often claimed that scientists can obtain new knowledge about nature by running computer simulations. How is this possible? I answer this question by arguing that computer simulations are arguments. This view parallels Norton’s argument view about thought experiments. I show that computer simulations can be reconstructed as arguments that fully capture the epistemic power of the simulations. Assuming the extended mind hypothesis, I furthermore argue that running the computer simulation is to execute the reconstructing argument. I discuss some (...) objections and reject the view that computer simulations produce knowledge because they are experiments. I conclude by comparing thought experiments and computer simulations, assuming that both are arguments. (shrink)

Computer simulations and experiments share many important features. One way of explaining the similarities is to say that computer simulations just are experiments. This claim is quite popular in the literature. The aim of this paper is to argue against the claim and to develop an alternative explanation of why computer simulations resemble experiments. To this purpose, experiment is characterized in terms of an intervention on a system and of the observation of the reaction. Thus, if computer simulations are experiments, (...) either the computer hardware or the target system must be intervened on and observed. I argue against the first option using the non-observation argument, among others. The second option is excluded by e.g. the over-control argument, which stresses epistemological differences between experiments and simulations. To account for the similarities between experiments and computer simulations, I propose to say that computer simulations can model possible experiments and do in fact often do so. (shrink)

This paper argues that the difference between contemporary software intensive scientific practice and more traditional non-software intensive varieties results from the characteristically high conditionality of software. We explain why the path complexity of programs with high conditionality imposes limits on standard error correction techniques and why this matters. While it is possible, in general, to characterize the error distribution in inquiry that does not involve high conditionality, we cannot characterize the error distribution in inquiry that depends on software. Software intensive (...) science presents distinctive error and uncertainty modalities that pose new challenges for the epistemology of science. (shrink)

We address some of the epistemological challenges highlighted by the Critical Data Studies literature by reference to some of the key debates in the philosophy of science concerning computational modeling and simulation. We provide a brief overview of these debates focusing particularly on what Paul Humphreys calls epistemic opacity. We argue that debates in Critical Data Studies and philosophy of science have neglected the problem of error management and error detection. This is an especially important feature of the epistemology of (...) Big Data. In “Error” section we explain the main characteristics of error detection and correction along with the relationship between error and path complexity in software. In this section we provide an overview of conventional statistical methods for error detection and review their limitations when faced with the high degree of conditionality inherent to modern software systems. (shrink)

For computer simulation models to usefully inform climate risk management, uncertainties in model projections must be explored and characterized. Because doing so requires running the model many times over, and because computing resources are finite, uncertainty assessment is more feasible using models that demand less computer processor time. Such models are generally simpler in the sense of being more idealized, or less realistic. So modelers face a trade-off between realism and uncertainty quantification. Seeing this trade-off for the important epistemic issue (...) that it is requires a shift in perspective from the established simplicity literature in philosophy of science. (shrink)

Many philosophical accounts of scientific models fail to distinguish between a simulation model and other forms of models. This failure is unfortunate because there are important differences pertaining to their methodology and epistemology that favor their philosophical understanding. The core claim presented here is that simulation models are rich and complex units of analysis in their own right, that they depart from known forms of scientific models in significant ways, and that a proper understanding of the type of model simulations (...) are fundamental for their philosophical assessment. I argue that simulation models can be distinguished from other forms of models by the many algorithmic structures, representation relations, and new semantic connections involved in their architecture. In this article, I reconstruct a general architecture for a simulation model, one that faithfully captures the complexities involved in most scientific research with computer simulations. Furthermore, I submit that a new methodology capable of conforming such architecture into a fully functional, computationally tractable computer simulation must be in place. I discuss this methodology—what I call recasting—and argue for its philosophical novelty. If these efforts are heading towards the right interpretation of simulation models, then one can show that computer simulations shed new light on the philosophy of science. To illustrate the potential of my interpretation of simulation models, I briefly discuss simulation-based explanations as a novel approach to questions about scientific explanation. (shrink)

Computer simulations are widely used in current scientific practice, as a tool to obtain information about various phenomena. Scientists accordingly rely on the outputs of computer simulations to make statements about the empirical world. In that sense, simulations seem to enable scientists to acquire empirical knowledge. The aim of this paper is to assess whether computer simulations actually allow for the production of empirical knowledge, and how. It provides an epistemological analysis of present-day empirical science, to which the traditional epistemological (...) categories cannot apply in any simple way. Our strategy consists in acknowledging the complexity of scientific practice, and trying to assess its rationality. Hence, while we are careful not to abstract away from the details of scientific practice, our approach is not strictly descriptive: our goal is to state in what conditions empirical science can rely on computer simulations. In order to do so, we need to adopt a renewed epistemological framework, whose categories would enable us to give a finer-grained, and better-fitted analysis of the rationality of scientific practice. (shrink)

Where should computer simulations be located on the ‘usual methodological map’ which distinguishes experiment from theory? Specifically, do simulations ultimately qualify as experiments or as thought experiments? Ever since Galison raised that question, a passionate debate has developed, pushing many issues to the forefront of discussions concerning the epistemology and methodology of computer simulation. This review article illuminates the positions in that debate, evaluates the discourse and gives an outlook on questions that have not yet been addressed.

This paper examines a constellation of ethical and editorial issues that have arisen since philosophers started to conduct, submit and publish empirical research. These issues encompass concerns over responsible authorship, fair treatment of human subjects, ethicality of experimental procedures, availability of data, unselective reporting and publishability of research findings. This study aims to assess whether the philosophical community has as yet successfully addressed such issues. To do so, the instructions for authors, submission process and published research papers of 29 main (...) journals in philosophy have been considered and analyzed. In light of the evidence reported here, it is argued that the philosophical community has as yet failed to properly tackle such issues. The paper also delivers some recommendations for authors, reviewers and editors in the field. (shrink)

We give an overview of the practice of developing and using complex climate models, as seen from experiences in a major climate modelling center and through participation in the Coupled Model Intercomparison Project. We discuss the construction and calibration of models; their evaluation, especially through use of out-of-sample tests; and their exploitation in multi-model ensembles to identify biases and make predictions. We stress that adequacy or utility of climate models is best assessed via their skill against more naïve predictions. The (...) framework we use for making inferences about reality using simulations is naturally Bayesian, and has many points of contact with more familiar examples of scientific epistemology. While the use of complex simulations in science is a development that changes much in how science is done in practice, we argue that the concepts being applied fit very much into traditional practices of the scientific method, albeit those more often associated with laboratory work. (shrink)

Scientificity is essentially methodology. The use of information technology as methodological instruments in science has been increasing for decades, this raises the question: Does this transform science? This question is the subject of the Special Issue in Minds and Machines “The epistemological significance of methods in computer simulation and machine learning”. We show that there is a technological change in this area that has three methodological and epistemic consequences: methodological opacity, reproducibility issues, and altered forms of justification.

In this paper, we argue for the centrality of prediction in the use of computational models in science. We focus on the consequences of the irreversibility of computational models and on the conditional or ceteris paribus, nature of the kinds of their predictions. By irreversibility, we mean the fact that computational models can generally arrive at the same state via many possible sequences of previous states. Thus, while in the natural world, it is generally assumed that physical states have a (...) unique history, representations of those states in a computational model will usually be compatible with more than one possible history in the model. We describe some of the challenges involved in prediction and retrodiction in computational models while arguing that prediction is an essential feature of non-arbitrary decision making. Furthermore, we contend that the non-predictive virtues of computational models are dependent to a significant degree on the predictive success of the models in question. (shrink)

Personalised medicine has been discussed as a medical paradigm shift that will improve health while reducing inefficiency and waste. At the same time, it raises new practical, regulatory, and ethical challenges. In this paper, we examine PM strategies epistemologically in order to develop capacities to address these challenges, focusing on a recently proposed strategy for developing patient-specific models from induced pluripotent stem cells so as to make individualised treatment predictions. We compare this strategy to two main PM strategies—stratified medicine and (...) computational models. Drawing on epistemological work in the philosophy of medicine, we explain why these two methods, while powerful, are neither truly personalised nor, epistemologically speaking, novel strategies. Both are forms of correlational black box. We then argue that the iPSC models would count as a new kind of black box. They would not rely entirely on mechanistic knowledge, and they would utilise correlational evidence in a different way from other strategies—a way that would enable personalised predictions. In arguing that the iPSC models would present a novel method of gaining evidence for clinical practice, we provide an epistemic analysis that can help to inform the practical, regulatory, and ethical challenges of developing an iPSC system. (shrink)

A chronicled approach to the notion of computer simulations shows that there are two predominant interpretations in the specialized literature. According to the first interpretation, computer simulations are techniques for finding the set of solutions to a mathematical model. I call this first interpretation the problem-solving technique viewpoint. In its second interpretation, computer simulations are considered to describe patterns of behavior of a target system. I call this second interpretation the description of patterns of behavior viewpoint of computer simulations. This (...) article explores these two interpretations of computer simulations from three different angles. First, I collect a series of definitions of computer simulation from the historical record. I track back definitions to the early 1960s and show how each viewpoint shares similar interpretations of computer simulations—ultimately clustering into the two viewpoints aforementioned. This reconstruction also includes the most recent literature. Second, I unpack the philosophical assumptions behind each viewpoint, with a special emphasis on their differences. Third, I discuss the philosophical implications of each viewpoint in the context of the recent discussion on the logic of scientific explanation for computer simulations. (shrink)

This paper examines a constellation of ethical and editorial issues that have arisen since philosophers started to conduct, submit and publish empirical research. These issues encompass concerns over responsible authorship, fair treatment of human subjects, ethicality of experimental procedures, availability of data, unselective reporting and publishability of research findings. This study aims to assess whether the philosophical community has as yet successfully addressed such issues. To do so, the instructions for authors, submission process and published research papers of 29 main (...) journals in philosophy have been considered and analyzed. In light of the evidence reported here, it is argued that the philosophical community has as yet failed to properly tackle such issues. The paper also delivers some recommendations for authors, reviewers and editors in the field. (shrink)

ABSTRACT This paper examines consequences of the computer revolution in mathematics. By comparing its repercussions with those of conceptual developments that unfolded in the nineteenth century, I argue that the key epistemological lesson to draw from the two transformative periods is that effective and successful mathematical practices in science result from integrating the computational and conceptual styles of mathematics, and not that one of the two styles of mathematical reasoning is superior. Finally, I show that the methodology deployed by applied (...) mathematicians in modern scientific computing is a paradigmatic instance of this key lesson. (shrink)

The epistemology of computer simulations has become a mainstream topic in the philosophy of technology. Within this large area, significant differences hold between the various types of models and simulation technologies. Agent-based and multi-agent systems simulations introduce a specific constraint on the types of agents and systems modelled. We argue that such difference is crucial and that simulation for the artificial sciences requires the formulation of its own specific epistemological principles. We present a minimally committed epistemology which relies on the (...) methodological principles of the Philosophy of Information and requires weak assumptions on the usability of the simulation and the controllability of the model. We use these principles to provide a new definition of simulation for the context of interest. (shrink)

Modelling and simulations have gained a leading position in contemporary attempts to describe, explain, and quantitatively predict the human brain's operations. Computer models are highly sophisticated tools developed to achieve an integrated knowledge of the brain with the aim of overcoming the actual fragmentation resulting from different neuroscientific approaches. In this paper we investigate plausibility of simulation technologies for emulation of consciousness and the potential clinical impact of large-scale brain simulation on the assessment and care of disorders of consciousness, e.g. (...) Coma, Vegetative State/Unresponsive Wakefulness Syndrome, Minimally Conscious State.Notwithstanding their technical limitations, we suggest that simulation technologies may offer new solutions to old practical problems, particularly in clinical contexts. We take DOCs as an illustrative case, arguing that the simulation of neural correlates of consciousness is potentially useful for improving treatments of patients with DOCs. (shrink)

The pervasive use of computer simulations in the sciences brings novel epistemological issues discussed in the philosophy of science literature since about a decade. Evolutionary biology strongly relies on such simulations, and in relation to it there exists a research program (Artificial Life) that mainly studies simulations themselves. This paper addresses the specificity of computer simulations in evolutionary biology, in the context (described in Sect. 1) of a set of questions about their scope as explanations, the nature of validation processes (...) and the relation between simulations and true experiments or mathematical models. After making distinctions, especially between a weak use where simulations test hypotheses about the world, and a strong use where they allow one to explore sets of evolutionary dynamics not necessarily extant in our world, I argue in Sect. 2 that (weak) simulations are likely to represent in virtue of the fact that they instantiate specific features of causal processes that may be isomorphic to features of some causal processes in the world, though the latter are always intertwined with a myriad of different processes and hence unlikely to be directly manipulated and studied. I therefore argue that these simulations are merely able to provide candidate explanations for real patterns. Section 3 ends up by placing strong and weak simulations in Levins’ triangle, that conceives of simulations as devices trying to fulfil one or two among three incompatible epistemic values (precision, realism, genericity). (shrink)

The Epistemology Of Computer Simulation has developed as an epistemological and methodological analysis of simulative sciences using quantitative computational models to represent and predict empirical phenomena of interest. In this paper, Executable Cell Biology and Agent-Based Modelling are examined to show how one may take advantage of qualitative computational models to evaluate reachability properties of reactive systems. In contrast to the thesis, advanced by EOCS, that computational models are not adequate representations of the simulated empirical systems, it is shown how (...) the representational adequacy of qualitative models is essential to evaluate reachability properties. Justification theory, if not playing an essential role in EOCS, is exhibited to be involved in the process of advancing and corroborating model-based hypotheses about empirical systems in ECB and ABM. Finally, the practice of evaluating model-based hypothesis by testing the simulated systems is shown to constitute an argument in favour of the thesis that computer simulations in ECB and ABM can be put on a par with scientific experiments. (shrink)

It has been argued that the Duhem problem is renewed with computational models since model assumptions having a representational aim and computational assumptions cannot be tested in isolation. In particular, while the Verification and Validation methodology is supposed to prevent such holism, Winsberg argues that verification and validation cannot be separated in practice. Morrison replies that Winsberg overstates the entanglement between the steps. The paper aims at arbitrating these two positions, by stressing their respective validity in relation to domains of (...) application. It importantly argues for an increasing use of formal methods in verification, that makes disentanglement possible. (shrink)

Mathematical models are a well established tool in most natural sciences. Although models have been neglected by the philosophy of science for a long time, their epistemological status as a link between theory and reality is now fairly well understood. However, regarding the epistemological status of mathematical models in the social sciences, there still exists a considerable unclarity. In my paper I argue that this results from specific challenges that mathematical models and especially computer simulations face in the social sciences. (...) The most important difference between the social sciences and the natural sciences with respect to modeling is that in the social sciences powerful and well confirmed background theories (like Newtonian mechanics, quantum mechanics or the theory of relativity in physics) do not exist in the social sciences. Therefore, an epistemology of models that is formed on the role model of physics may not be appropriate for the social sciences. I discuss the challenges that modeling faces in the social sciences and point out their epistemological consequences. The most important consequences are that greater emphasis must be placed on empirical validation than on theoretical validation and that the relevance of purely theoretical simulations is strongly limited. (shrink)

This paper evaluates Nancy Cartwright’s critique of economic models. Cartwright argues that economics fails to build relevant “nomological machines” able to isolate capacities. In this paper, I contend that many economic models are not used as nomological machines. I give some evidence for this claim and build on an inferential and pragmatic approach to economic modeling. Modeling in economics responds to peculiar inferential norms where a “good” model is essentially a model that enhances our knowledge about possible worlds. As a (...) consequence, models and experiments are very different knowledge-producing devices, at least in economics. (shrink)

This is my personal homepage. Find my philosophical papers under "Philosophy".

If well-designed, the results of a Randomised Clinical Trial can justify a causal claim between treatment and effect in the study population; however, additional information might be needed to carry over this result to another population. RCTs have been criticized exactly on grounds of failing to provide this sort of information Evidence, inference and enquiry. Oxford University Press, New York, 2011), as well as to black-box important details regarding the mechanisms underpinning the causal law instantiated by the RCT result. On (...) the other side, so-called In Silico Clinical Trials face the same criticisms addressed against standard modelling and simulation techniques, and cannot be equated to experiments Philosophy of molecular medicine: foundational issues in research and practice, Routledge, New York, 2017; Parker in Synthese 169:483–496, 2009; Parke in Philos Sci 81:516–536, 2014; Diez Roux in Am J Epidemiol 181:100–102, 2015 and related discussions in Frigg and Reiss in Synthese 169:593–613, 2009; Winsberg in Synthese 169:575–592, 2009; Beisbart and Norton in Int Stud Philos Sci 26:403–422, 2012). We undertake a formal analysis of both methods in order to identify their distinct contribution to causal inference in the clinical setting. Britton et al.’s study :E2098–E2105, 2013) on the impact of ion current variability on cardiac electrophysiology is used for illustrative purposes. We deduce that, by predicting variability through interpolation, ISCTs aid with problems regarding extrapolation of RCTs results, and therefore in assessing their external validity. Furthermore, ISCTs can be said to encode “thick” causal knowledge —as opposed to “thin” difference-making information inferred from RCTs. Hence, ISCTs and RCTs cannot replace one another but rather, they are complementary in that the former provide information about the determinants of variability of causal effects, while the latter can, under certain conditions, establish causality in the first place. (shrink)

Verification and validation of computer codes and models used in simulations are two aspects of the scientific practice of high importance that recently have been discussed widely by philosophers of science. While verification is predominantly associated with the correctness of the way a model is represented by a computer code or algorithm, validation more often refers to the model’s relation to the real world and its intended use. Because complex simulations are generally opaque to a practitioner, the Duhem problem can (...) arise with verification and validation due to their entanglement; such an entanglement makes it impossible to distinguish whether a coding error or the model’s general inadequacy to its target should be blamed in the case of a failure. I argue that a clear distinction between computer modeling and simulation has to be made to disentangle verification and validation. Drawing on that distinction, I suggest to associate modeling with verification and simulation, which shares common epistemic strategies with experimentation, with validation. To explain the reasons for their entanglement in practice, I propose a Weberian ideal–typical model of modeling and simulation as roles in practice. I examine an approach to mitigate the Duhem problem for verification and validation that is generally applicable in practice and is based on differences in epistemic strategies and scopes. Based on this analysis, I suggest two strategies to increase the reliability of simulation results, namely, avoiding alterations of verified models at the validation stage as well as performing simulations of the same target system using two or more different models. In response to Winsberg’s claim that verification and validation are entangled I argue that deploying the methodology proposed in this work it is possible to mitigate inseparability of V&V in many if not all domains where modeling and simulation are used. (shrink)

This paper combines naturalized metaphysics and a philosophical reflection on a recently evolving interdisciplinary branch of quantum chemistry, ab initio molecular dynamics. Bridging the gaps among chemistry, physics, and computer science, this cutting-edge research field explores the structure and dynamics of complex molecular many-body systems through computer simulations. These simulations are allegedly crafted solely by the laws of fundamental physics, and are explicitly designed to capture nature as closely as possible. The models and algorithms employed, however, involve many approximations and (...) significant degrees of idealization of their target systems. Therefore, for philosophers of science the pivotal question of whether relying only on the fundamental laws of physics supports a reductionist or realist stance arises. One conceivable answer to this question is that the irreducible approximations and idealizations support rather anti-realist positions. After reviewing an influential attitude in the philosophy of computer simulations and the debate concerning scientific realism, I offer a fair interpretation of such ab initio modelling in quantum chemistry within a naturalistic metaphysical framework that gives rise to a specific type of ontic structural realism. (shrink)

Here we enrich Paul Weirich’s thesis holding that a simulation model can create knowledge in the form of causal explanations. We sustain the validity of exporting results from the model to the modelized world in virtue of the similarity between model and world, which is analyzable in terms of partial identity of structure, eliminating the superficial similarity that repeats empirical results by adjusting data via calibration. The structure of relations rescues from the world critical results to analyze such similarity, as (...) so certain properties which condition the kind of relations held between the elements of the model representing entities of the world. (shrink)

Numerical simulations have come to be widely used in scientific work. Like experiments, simulations generate large quantities of numbers that require analysis and constant concern with uncertainty and error. How do simulationists convince themselves, and others, about the credibility of output? The present analysis reconstructs the perspectives related to performing numerical simulations, in general, and the situations in which simulationists deal with uncertain output, in particular. Starting from a distinction between idealized and realistic simulations, the paper presents the principal methods (...) of evaluation in relation to these practices and how different audiences expect different methods. One major challenge in interpreting output data is to distinguish between “real” and “numerical” effects. Within the practice of idealized simulations, simulationists hold the underlying model accountable for results that manifest “real” effects, but because “numerical” and “real” effects cannot be distinguished on the basis of what they derive from, attempted causal explanations are rather justifications for their conclusions. At the same time, simulationists’ explanations are part and parcel of their contradictory perspectives, according to which they believe in simulations largely due to the underlying model, while painfully recognizing everything they have to add to make computations doable on the basis of this model. (shrink)