The Modular Structure of Physical Theories
Olivier Darrigol

Far from being homogenous wholes, all advanced theories contain a variety of modules that are themselves theories with different domains of application. There is strong historical evidence that this modularity plays an essential role in the construction, application, comparison, and communication of theories. The purposes of this talk are to define, to illustrate, and to justify this notion.

Probabilistic Unawareness
Mikaël Cozic

The modeling of unawareness has become a major topic in formal epistemology. Convincing models of unawareness are now available for full beliefs in the framework of epistemic logic. Our aim is to investigate the modeling of unawareness in the probabilistic model of partial beliefs. We will show how the use of impossible states allow to model unawareness in the probabilistic case.

Models and Fiction
Roman Frigg

Recent philosophical debate has focussed on the relationship between models and their target systems. This paper is concerned with the nature of the model itself. Most models are not physical objects. This raises important questions. What sort of entity are models, what is truth in a model, and how do we learn about models? In this paper I argue that models shares important aspects in common with literary fiction, and that therefore theories of fiction can be brought to bear on these questions. In particular, I argue that the pretence account as developed by Walton (1990) has the resources to answer these questions. I introduce this account, outline the answers that it offers, and develop a general picture of scientific modelling.

Criteria of Significance as a Tool in Discussions about Reduction
Sebastian Lutz

Within logical empiricism, empirical significance and reduction both started out as the explicit definability of all terms. For reduction, the demand for explicit definitions was weakened, for example, to reduction sentences, meaning postulates, and interpretative systems. For empirical significance, the demand was weakened to, among others, verifiability, falsifiability, and Ayer's trivialized criteria. I relate the different criteria to each other, making a distinction between syntactic and semantic criteria of significance and reduction as well as between criteria for terms and for sentences. To relate the criteria to contemporary notions of reduction, I elucidate the relation between the syntactic and the semantic view of theories and argue for an applicability of model theory to scientific models and theories.

Science without Models: Resampling Methods in Statistical Inference
Jan Sprenger

Statistical inference and data anaylsis have conquered more and more terrain in the empirical science. Therefore, the inferential function of statistical models becomes an increasingly important object of scrutiny. Parametric statistical modeling is traditionally considered to be an indispensable tool in data analysis, but in a variety of cases where the underlying process is ill understood, explicit modeling assumptions lead the entire inference astray. This is especially salient in sciences where most models have tentative character, e.g. in climatology. Resampling methods are simulation-based techniques of data-based inference that are able to replace specific modeling assumptions. They are generally applicable, yield surprisingly accurate predictions and often outperform parametric statistical models. Their main drawback consists in the associated computational costs, but this problem has been solved by the advent of fast and cheap computers. Philosophers of science should realize that direct learning from data often provides a full-value alternative to explicit model-building.

On the Justification and Formation of Definitions: The Case of Deterministic Chaos
Charlotte Werndl

Mathematicians typically have good reasons for studying a definition, and these reasons often reveal how this definition was formed. Furthermore, finding a proper definition is often regarded as a considerable advance in mathematical knowledge. These considerations motivate the following general questions: how are definitions in mathematics justified and formed? And is this justification and formation of definitions reasonable? I will look at these questions for the case study of notions of deterministic chaos, which are part of ergodic theory and topological dynamics. Whenever possible I generalise and generally make claims about the justification and formation of definitions in mathematics. Lakatos coined the term 'proof-generated definition', which is a definition that is formed in order to be able to prove a specific conjecture. However, apart from this idea, there is little philosophical reflection on the actual practice of how definitions are justified and formed in mathematics. Next to proof-generation I will identify other reasonable kinds of justification and formation of definitions, some of which have not been identified before. Moreover, I will discuss the interrelationships between the different kinds of justifying and forming definitions. These results show that Lakatos's strong emphasis on proof-generated definitions is not warranted. In nearly all cases various types of justification and formation of definitions can be expected to play a role. Finally, I argue that having a critical look at the arguments that are, often implicitly, involved in justifying and generating notions of chaos and assessing their validity leads to a deep understanding of chaos; without this analysis the understanding of chaos would not be as deep.