Claus Beisbart

In contemporary computer simulations, particles attract each other and form clusters, cells interact, and agents communicate with one another. This is at least how computer simulations are commonly described. But how can we make sense of such talk? One answer is that the particles, cells, and agents inside simulations are digital artifacts, and thus real objects. In this paper, I cast doubt on this realist position by raising the question: To what objects does a simulation give rise, if it does so at all? I tentatively suggest answers that try to determine what objects arise in computer simulations at both a fundamental and a higher level. However, closer analysis reveals that we run into underdetermination problems. Even for a single run of a simulation program, the acknowledged facts about the simulations often do not suffice to determine the objects and their properties uniquely. I suggest that, from an ontological perspective, it is more attractive to say that computer simulations involve computations, the results of which may then be interpreted in various ways. To make sense of talk about objects in simulations it is more attractive to develop a fictionalist account, or so I argue.

The choice of a social decision rule for a federal assembly affects the welfare distribution within the federation. But which decision rules can be recommended on welfarist grounds? In this paper, we focus on two welfarist desiderata, viz. (i) maximizing the expected utility of the whole federation and (ii) equalizing the expected utilities of people from dif- ferent states in the federation. We consider the European Union as an example, set up a probabilistic model of decision making and explore how different decision rules fare with regard to the desiderata. We start with a default model, where the interests, and therefore the votes of the different states are not correlated. This default model is then abandoned in favor of models with correlations. We perform computer simulations and find that decision rules with a low acceptance threshold do generally better in terms of desideratum (i), whereas the rules presented in the Accession Treaty and in the (still unratified) Constitution of the Euro- pean Union tend to do better in terms of desideratum (ii). The ranking obtained regarding desideratum (i) is fairly stable across different corre- lation patterns.

Since the publication of his seminal monograph "The scientific image", Bas van Fraassen is a key figure in philosophy of science. In this book, other philosophers with various outlooks critically discuss his work on theories, empiricism and philosophical stances. The book starts with a new article by van Fraassen on his preferred account of theories, the so-called semantic view. This account is now 50 years old, and van Fraassen takes this anniversary as an opportunity to review the account, its history and the philosophical discussion about it. In the main part of the book, Nancy Cartwright, Finnur Dellsén, Matthias Egg, Steven French, Michael Friedman, Milena Ivanova and Michela Massimi discuss van Fraassen's contributions to philosophy. Three chapters focus on his engagement with realism (French, Friedman, Ivanova). Others study his voluntarism (Cartwright) and his view on representation (Massimi). Finally, there are contributions about his elaboration of empiricism (Dellsén) and his proposal to consider philosophical positions as stances (Egg). The volume includes a laudatio written by Steven French and finishes with a reply by van Fraassen to his critics.

The goal of this paper is to re-assess reflective equilibrium (“RE”). We ask whether there is a conception of RE that can be defended against the various objections that have been raised against RE in the literature. To answer this question, we provide a systematic overview of the main objections, and for each objection, we investigate why it looks plausible, on what standard or expectation it is based, how it can be answered and which features RE must have to meet the objection. We find that there is a conception of RE that promises to withstand all objections. However, this conception has some features that may be unexpected: it aims at a justification that is tailored to understanding and it is neither tied to intuitions nor does it imply coherentism. We conclude by pointing out a cluster of questions we think RE theorists should pay more attention to.

Reflective equilibrium (RE) is often regarded as a powerful method in ethics, logic, and even philosophy in general. Despite this popularity, characterizations of the method have been fairly vague and unspecific so far. It thus may be doubted whether RE is more than a jumble of appealing but ultimately sketchy ideas that cannot be spelled out consistently. In this paper, we dispel such doubts by devising a formal model of RE. The model contains as components the agent’s commitments and a theory that tries to systematize the commitments. It yields a precise picture of how the commitments and the theory are adjusted to each other. The model differentiates between equilibrium as a target state and the dynamic equilibration process. First solutions to the model, obtained by computer simulation, show that the method allows for consistent specification and that the model’s implications are plausible in view of expectations on RE. In particular, the mutual adjustment of commitments and theory can improve one’s commitments, as proponents of RE have suggested. We argue that our model is fruitful not only because it points to issues that need to be dealt with for a better understanding of RE, but also because it provides the means to address these issues.

This paper discusses recent attempts to solve the problem of induction. Two broad strategies to escape Hume's fork can be distinguished. The first tries to localize the justification of specific inductions in uncontroversial empirical knowledge, e.g.mundane scientific knowledge (J. D. Norton) or perception (M. Lange). I argue that related attempts to (dis)solve the problem fail. The second strategy tries to put forward an argument in favor of induction. As a discussion of work by R. White shows, this argument can barely prove that induction is reliable or at least not unreliable. But D. Steel, F. Huber and G. Schurz could show that enumerative induction is necessary and sufficient for a certain epistemic goal or optimal in a certain sense. These proofs, however, only solve the problem of induction if the goal or a certain standard has priority over the avoidance of error. This suggests that the difficulties to justify induction do not so much derive from Hume's fork, but rather from a plurality of sensible epistemic goals that can conflict with each other.

Some machine learning models, in particular deep neural networks (DNNs), are not very well understood; nevertheless, they are frequently used in science. Does this lack of understanding pose a problem for using DNNs to understand empirical phenomena? Emily Sullivan has recently argued that understanding with DNNs is not limited by our lack of understanding of DNNs themselves. In the present paper, we will argue, _contra_ Sullivan, that our current lack of understanding of DNNs does limit our ability to understand with DNNs. Sullivan’s claim hinges on which notion of understanding is at play. If we employ a weak notion of understanding, then her claim is tenable, but rather weak. If, however, we employ a strong notion of understanding, particularly explanatory understanding, then her claim is not tenable.

This paper details the content and structure of the folk concept of life, and discusses its relevance for scientific research on life. In four empirical studies, we investigate which features of life are considered salient, universal, central, and necessary. Functionings, such as nutrition and reproduction, but not material composition, turn out to be salient features commonly associated with living beings (Study 1). By contrast, being made of cells is considered a universal feature of living species (Study 2), a central aspect of life (Study 3), and our best candidate for being necessary for life (Study 4). These results are best explained by the hypothesis that people take life to be a natural kind subject to scientific scrutiny.

In his Critique of Pure Reason, in the chapter on the antinomy of pure reason, Kant not only argues that aprioristic cosmology is doomed to failure; he also implies that empirical knowledge about the universe is impossible. Today, such a negative verdict about the possibility of cosmological knowledge seems implausible because physical cosmology has made substantial progress. In particular, the spatiotemporal extension of the universe now seems a matter of empirical investigation in which models figure centrally. But I think it is worth considering the possibility that Kant got something right and that he offers insights that can help us to better understand problems in present-day cosmology. In this article, I explore a striking coincidence: according to both Kant and current wisdom, cosmology faces a serious underdetermination problem regarding the spatiotemporal extension of the world. As a closer analysis reveals, however, Kant and modern cosmology differ on the reasons why underdetermination arises. In current cosmology, underdetermination follows from laws that are knowable a posteriori, and not only from the very idea of cosmological knowledge, as Kant would have it. This suggests that the current underdetermination problem is not fully a Kantian one.

Computer simulations are often claimed to be opaque and thus to lack transparency. But what exactly is the opacity of simulations? This paper aims to answer that question by proposing an explication of opacity. Such an explication is needed, I argue, because the pioneering definition of opacity by P. Humphreys and a recent elaboration by Durán and Formanek are too narrow. While it is true that simulations are opaque in that they include too many computations and thus cannot be checked by hand, this doesn’t exhaust what we might want to call the opacity of simulations. I thus make a fresh start with the natural idea that the opacity of a method is its disposition to resist knowledge and understanding. I draw on recent work on understanding and elaborate the idea by a systematic investigation into what type of knowledge and what type of understanding are required if opacity is to be avoided and why the required sort of understanding, in particular, is difficult to achieve. My proposal is that a method is opaque to the degree that it’s difficult for humans to know and to understand why its outcomes arise. This proposal allows for a comparison between different methods regarding opacity. It further refers to a kind of epistemic access that is important in scientific work with simulations.

Philosophie kann dazu beitragen, dass wir vernünftiger mit den Problemen umgehen, die unsere Gesellschaft und ihr Selbstverständnis herausfordern. Dazu muss die Philosophie sich aber öffentlich einmischen und verstärkt in die Bildung Einzug halten – diese Position vertritt vorliegender Band. Die Beiträge von Anne Burkard, Rainer Hegselmann, Romy Jaster und Markus Wild zeigen einerseits auf, welche Rolle die Philosophie in öffentlichen Debatten spielen kann und soll. Andererseits analysieren sie, welchen Beitrag Philosophie zur schulischen und universitären Bildung liefert.

Cosmological questions (e.g., how far the world extends and how it all began) have occupied humans for ages and given rise to numerous conjectures, both within and outside philosophy. To put to rest fruitless speculation, Kant argued that these questions move beyond the limits of human knowledge. This article begins with Kant’s doubts about cosmology and shows that his arguments presuppose unreasonably high standards on knowledge and unwarranted assumptions about space-time. As an analysis of the foundations of twentieth-century cosmology reveals, other worries about the discipline can be avoided too if the universe is modeled using Einstein’s general theory of relativity. There is now strong observational support for one particular model. However, due to underdetermination problems, the big cosmological questions cannot be fully answered using this model either. This opens the space for more speculative proposals again (e.g., that the universe is only part of a huge multiverse).

We construct a new measure of voting power that yields reasonable measurements even if the individual votes are not cast independently. Our measure hinges on probabilities of counterfactuals, such as the probability that the outcome of a collective decision would have been yes, had a voter voted yes rather than no as she did in the real world. The probabilities of such counterfactuals are calculated on the basis of causal information, following the approach by Balke and Pearl. Opinion leaders whose votes have causal influence on other voters' votes can have significantly more voting power under our measure. But the new measure of voting power is also sensitive to the voting rule. We show that our measure can be regarded as an average treatment effect, we provide examples in which it yields intuitively plausible results and we prove that it reduces to Banzhaf voting power in the limiting case of independent and equiprobable votes

The mean majority deficit in a two-tier voting system is a function of the partition of the population. We derive a new square-root rule: For odd-numbered population sizes and equipopulous units the mean majority deficit is maximal when the member size of the units in the partition is close to the square root of the population size. Furthermore, within the partitions into roughly equipopulous units, partitions with small even numbers of units or small even-sized units yield high mean majority deficits. We discuss the implications for the winner-takes-all system in the US Electoral College.

We consider a decision board with representatives who vote on proposals on behalf of their constituencies. We look for decision rules that realize utilitarian and (welfarist) egalitarian ideals. We set up a simple model and obtain roughly the following results. If the interests of people from the same constituency are uncorrelated, then a weighted rule with square root weights does best in terms of both ideals. If there are perfect correlations, then the utilitarian ideal requires proportional weights, whereas the egalitarian ideal requires equal weights. We investigate correlations that are in between these extremes and provide analytic arguments to connect our results to Barberà and Jackson (J Polit Econ 114(2):317–339, 2006) and to Banzhaf voting power.

We develop a utilitarian framework to assess different decision rules for the European Council of Ministers. The proposals to be decided on are conceptualized as utility vectors and a probability distribution is assumed over the utilities. We first show what decision rules yield the highest expected utilities for different means of the probability distri- bution. For proposals with high mean utility, simple bench- mark rules (such as majority voting with proportional weights) tend to outperform rules that have been proposed in the political arena. For proposals with low mean utility, it is the other way round. We then compare the expected utilities for smaller and larger countries and look for Pareto- dominance relations. Finally, we provide an extension of the model, discuss its restrictions, and compare our approach with assessments of decision rules that are based on the Penrose measure of voting power.

To provide an introduction to this book, we explain the motivation to publish this volume, state its main goal, characterize its intended readership, and give an overview of its content. To this purpose, we briefly summarize each chapter and put it in the context of the whole volume. We also take the opportunity to stress connections between the chapters. We conclude with a brief outlook.The main motivation to publish this volume was the diagnosis that the validation of computer simulation needs more attention in practice and in theory. The aim of this volume is to improve our understanding of validation. To this purpose, computer scientists, mathematicians, working scientists from various fields, as well as philosophers of science join efforts. They explain basic notions and principles of validation, embed validation in philosophical frameworks such as Bayesian epistemology, detail the steps needed during validation, provide best practice examples, reflect upon challenges to validation, and put validation in a broader perspective. As we suggest in our outlook, the validation of computer simulations will remain an important research topic that needs cross- and interdisciplinary efforts. A key issue is whether, and if so, how very rigorous approaches to validation that have proven useful in, e.g., engineering can be extended to other disciplines.

This chapter clarifies the concept of validation of computer simulations by comparing various definitions that have been proposed for the notion. While the definitions agree in taking validation to be an evaluationEvaluation, they differ on the following questions: What exactly is evaluated—results from a computer simulation, a model, a computer codeCode? What are the standardsStandard of evaluationEvaluation––truthTruth, accuracyAccuracy, and credibilityCredibility or also something else? What type of verdict does validation lead to––that the simulation is such and such good, or that it passes a testTest defined by a certain threshold? How strong needs the case to be for the verdict? Does validation necessarily proceed by comparing simulation outputsOutput with measured dataData? Along with these questions, the chapter explains notions that figure prominently in them, e.g., the concepts of accuracy and credibility. It further discusses natural answers to the questions as well as arguments that speak in favor and against these answers. The aim is to obtain a better understandingUnderstanding of the options we have for defining validation and how they are related to each other.

Bayesian epistemologyEpistemology offers a powerful framework for characterizing scientific inference. Its basic idea is that rational belief comes in degrees that can be measured in terms of probabilities. The axioms of the probability calculus and a rule for updatingUpdating emerge as constraints on the formation of rational belief. Bayesian epistemologyEpistemology has led to useful explications of notions such asConfirmation confirmation. It thus is natural to ask whether Bayesian epistemologyEpistemology offers a useful framework for thinking about the inferences implicit in the validation of computer simulations. The aim of this chapter is to answer this question. Bayesian epistemologyEpistemology is briefly summarized and then applied to validation. UpdatingUpdating is shown to form a viable method for data-driven validation. Bayesians can also express how a simulation obtains prior credibilityCredibility because the underlying conceptual modelConceptual model is credible. But the impact of this prior credibilityCredibility is indirect since simulations at best provide partial and approximate solutions to theConceptual model conceptual model. Fortunately, this gap between the simulations and the conceptual model can be addressed using what we call Bayesian verification. The final part of the chapter systematically evaluates the use of Bayesian epistemologyEpistemology in validation, e.g., by comparing it to a falsificationist approach. It is argued that Bayesian epistemologyEpistemology goes beyond mere calibrationCalibration and that it can provide the foundations for a sound evaluationEvaluation of computer simulations.

Verification and validation are methods with which computer simulations are tested. While many practitioners draw a clear line between verification and validation and demand that the former precedes the latter, some philosophers have suggested that the distinction has been over-exaggerated. This chapter clarifies the relationship between verification and validation. Regarding the latter, validation of the conceptual and of the computational modelComputational model are distinguished. I argue that, as a method, verification is clearly different from validation of either of the models. However, the methods are related to each other as follows: If we allow that the validation of the computational model need not include the comparisonComparison between simulation output and measured data, then the computational model may be validated by validating the conceptual model independently and by verifying the simulation with respect to it. This is often not realistic, however, because, in most cases, the conceptual modelConceptual model cannot be validated independently from the simulation. In such cases, the computational modelComputational model is verified with the aim to use it as an appropriate substitute for the conceptual model. Then simulation output is compared to measured data to validate both the computational and the conceptual model. I analyze the underlying inferences and argue that they require some prior confidence in the conceptual model and in verification. This suggests that verification precede validation that proceeds via a comparisonComparison between simulation outputOutput and measured data. Recent arguments to the effect that the distinction between verification and validation is not clear-cut do not refute these results, or so I argue against philosopher E. Winsberg.