My research combines philosophical with historical methods and centers on the following topics (see PhD dissertation abstract below):
        Prediction. There is a long-standing tradition in the philosophy of science that views a theory making novel predictions superior to theories that simply explain. Historical studies however have failed to unearth any positive evidence in support of this view. In response to this, temporal novelty has been rejected by some in favour of so-called use-novel predictions: evidence should be regarded as novel if the theory has not been purposefully designed to account for it. Yet there are several conceptual problems with the use-novel account and I don’t think the historical record really supports the idea of use-novelty anyway. Rather, I’m sympathetic to the view that explanations count as least as much, if not even more than predictions in the appraisal of theories. If true, this would of course raise a number of issues for the current philosophy of science. For instance, replies to the pessimistic meta induction assume that only those theories which make novel predictions should be considered in the argument. But if there are no grounds for this preference, then the PMI is a much stronger argument than it is already.
        Explanation. Intuitively, scientific explanations that are not true cannot be scientific explanations proper. Especially causal-mechanistic accounts of explanation explicitly make the contrary assumption. Without diverting into a van-Fraassian pragmatic account of explanation, I believe that mechanistic accounts should allow for non-referring entities and their envisaged interactions to be explanatory (see dissertation abstract, below).  
        Data reliability. Data reliability is absolutely fundamental for scientific knowledge. If data are not reliable, they cannot be used for eliminating false theories and for confirming true ones. But how do we know that the results produced by our experiments are trustworthy? Philosophers of science have not a lot to say about this question. The standard answer, however, has it that scientists need to perform error checks, repeat experiments, ‘calibrate’ the results gained with one experimental technique or instrument with another, and so on. I believe that the few extant accounts on data reliability have severe shortcomings. In my current project (see below), I’m seeking to show that the role of theories guiding data reliability judgments has been underestimated hitherto.  

Current project, abstract.
In the philosophy of science one generally assumes that the issue of theory-confirmation can clearly be separated from the issue of data reliability. This project is going to try to undermine this view by means of the notion of ‘theory-driven data reliability judgements’, which says that theories with certain properties significantly influence decisions made by scientists about whether experimental data are to be deemed reliable or not. The project will try to corroborate this thesis by means of case studies from the history of science. In particular, it will be argued that the postulation and acceptance of undetected error sources lends crucial support to the main thesis of this project. The project will furthermore inquire whether the following consequences can be drawn from the main thesis and the historical case studies: 1. theoretical virtues like elegance, simplicity, and unifying power of explanations weigh much heavier in theory-choice than generally believed; 2. not the empirical adequacy (or truth) of theories should be regarded as the normative first aim of science, but rather their maximal explanatory power; 3. a well-known major conceptual problem of scientific realism is aggravated by the considerations of this project.

PhD thesis, abstract.
In this thesis, I investigate the role unobservables play in scientific explanations, the naturalness of explanations, and the fertility (or developmental potential) of theories. 
     Realist accounts (in particular, the causal-mechanistic account) require unobservables to be real for them to fulfill an explanatory function in theories. Antirealist accounts do not assign any particular role to unobservables and marginalize the explanatory power of theories, their naturalness, and their developmental potential as merely of pragmatic interest. The position I put forth here criticizes both of these extremes as unsatisfactory. In contrast to antirealist accounts, I take the role of unobservables for the explanatory power of theories, their naturalness, and their developmental potential seriously, but in contrast to realist accounts, my position refrains from committing to the reality of unobservables. 
    In order to demarcate my account sufficiently from realist accounts of unobservables in explanations, I consider—among other things—the views realists hold about our capacity to access unobservables epistemically (before they can figure in our explanations) and criticize these views not on metaphysical grounds—as the antirealist would do—but rather on 'practical' ones. For this purpose, I discuss various prominent historical ‘discoveries’ of unobservables and show that inference procedures proposed by realists do not adequately describe these cases. 
    Despite my critique of realist accounts of unobservables, I try to articulate a position, in which unobservables play a positive part in explanations without one having to presuppose their reality. I also point out that, in some important cases, the postulated unobservables are also ‘visual’ entities: we cannot imagine them without visualizing them. I call these entities Imaginary-Constitutives. I show that these entities not only play an important role in scientific explanations, but also that they figure in important ways in natural and ‘fertile’ theories.