Justifying (some) thought experiments

Thought experiments are often criticised, e.g. for being intuition pumps (Dennett 1995), for excessive simplicity (Goodin 1982, 8-12), and for unreliability (Machery 2017, 45-148). In a paper I gave at ECPR 2016 in Prague (where I co-organised the first ever panels in this nascent section, with Keith Dowding), I addressed these important concerns with a new methodology for thought experiments. However, whenever I present this paper, most questions involve whether we should be doing thought experiments in the first place. We thus need a more foundational defence of thought experiments which explains that they are necessary and desirable (even though many actual thought experiments are silly or flawed). I start by presenting a new typology of thought experiments, by combining and amending the frameworks of Norton (1991, 131) and Gendler (2000, 25-7). One advantage of this new typology is to highlight *conceptual* thought experiments, e.g. for testing definitions. Conceptual thought experiments are mostly ignored in the methodological literature on thought experiments (though cf. Brendel 2004, 102-3). Yet they are incredibly widespread in political theory and beyond (e.g. Feinberg 1985, 10-49 on what counts as harm or offence). Conceptual thought experiments are vital because they help us test general claims, e.g. ‘Corruption is the misuse of public office for private gain’. Such general claims need to be tested, with actual or hypothetical examples. Another part of my argument is to reject the apparent difference between actual and hypothetical examples. ‘Actual’ examples always leave out some variables for the sake of simplicity, and even then we regularly have to amend the actual example by tweaking the variables – hence rendering it hypothetical. We could hardly test definitions without conceptual thought experiments. The same applies to *normative* thought experiments, which are similarly vital, because they help us test general normative claims, e.g. ‘Do as you like provided you do not harm others’. Such general claims also needed testing, with actual or hypothetical examples, and once again, even actual examples are likely to become hypothetical very quickly. The question of whether our judgements/intuitions are reliable remains a major concern, one which I cannot fully address in the present paper. But I do tackle it in two surprising ways. First, I show that in conceptual experiments we often *want* to pump (linguistic) intuitions to the surface. Consider the standard definition of corruption – the misuse of public office for private gain. This excludes the misuse of public office for *party* gain. Yet most of us would feel that such cases are also corrupt, e.g. politicians accepting money for their parties in exchange for policy influence. Second, sometimes we pump an intuition to the surface precisely to show that it is *wrong* - another underappreciated form of thought experiment (e.g. Hobbes’s man on a sinking ship, who *does* act freely, Hobbes then shows). In short, the paper accepts many problems with thought experiments but justifies some of them as necessary and desirable.