I’m no expert in philosophical methodology, but I’d like to lay out some thoughts here about three methods: thought experiments, experimental philosophy, and case studies. So this is less a definitive statement to be defended than an invitation to discussion: less an article than a blog post, in other words.
X-phil has made a big splash with its critique of armchair philosophy and its use of thought experiments, the whole bestiary of brains-in-a-vat, zombies, Swampman, et al.. Why rest with philosopher’s “intuitions” about the matters of life, and especially, why rest with their intuitions about other people’s intuitions, the X-phils ask. As far as I can tell, the X-phil critique claims the armchair approach is wrapped up in very questionable assumptions about normalcy and normativity, i.e., that any one given philosopher’s intuitions map onto that of “the rational man,” that is, the way most people do think, or the way they should think. Why not, the X-phil proponents say, instead design experiments that test how people do in fact think, and thereby settle the empirical question, and then, once we settle the normative question, we can test the discrepancy and come up with suggestions to improve rational performance, a pedagogical enterprise that is itself subject to empirical testing. But then, comes the retort from critics of X-phil, you’ve got to be very careful in designing your experiments. For all the talk of “intuition pumps” in thought experiments, you’ve got to be careful of experimental artifacts in X-phil results, the critics say. We know, we know, the X-phils answer, that’s why we work with social science experts in designing our experiments.
The debate goes on, as we know, and doesn’t seem to show signs of settling down any time soon.
I want suggest a third method for philosophers, beyond thought experiments and experimental philosophy: the case study, or as Deleuze calls it, “the method of dramatization.”
The problem with even the best designed X-phil use of the trolley problem, James Williams claims in his chapter on Deleuze and Gilbert Harman in The Transversal Thought of Gilles Deleuze: Encounters and Influences (Clinamen , 2005; available online here), is that the trolley problem yields pre-set, individuated, responses (pull the lever or not) and the experiments looks at the distribution of choices between individuated results (and also often at brain states and so on correlated with those choices).
But the trolley problem is not a “problem in the Deleuzean sense,” which for Williams looks like
vast inter-linked networks of conflicting ideal pressures and actual responses to these pressures…. A problem in Deleuze’s metaphysics is closer to the sense of a problem in macro-economics such as ‘Should we raise interest rates now?’ with its wide range of ethical, technical, social and political pressures and variables, than to a … problem … that might allow for a definitive answer. The economic problem is practical and experimental. It rests upon a variable context such that no final formula is likely to be resistant to new developments or even adequate in light of all currently available information.
For Deleuze, problems cannot be solved once and for all; they can only be dealt with practically and temporarily. They cannot be solved, but they can be resolved for the time being, for our actions now in resolving it change the conditions for future cases of the problem. And they can only be resolved in concrete situations by singular actions. This isn’t the place for a full-blown inquiry into Deleuze’s metaphysics, but the take-away point I think is that problems are differential fields, so that there are no pre-set, individuated, responses, but open-ended situations, and that no experiment can duplicate the complex pressures and real consequences of true problems. But we can write case studies of the real-life resolutions of problems, as I’ve tried to do with the case studies in Political Affect, case studies which respect and describe the multiple dimensions and interconnected pressures of “problems in the Deleuzean sense.”
That’s not to say we should only do case studies. But it is to say that the multiple and interconnected pressures and consequences of open-ended and evolving situations or “problems in the Deleuzean sense” can be studied better in detailed case studies than in experiments done in the armchair or even in the lab. There can be synergy here: an apprenticeship in case studies can help us identify key dimensions of situations that can be isolated from their real-world context and tested experimentally, and that very experimental knowledge helps us critique old case studies and produce new ones. For a case study itself involves the choice of what to include: a map that produces a 1-to-1 duplication of the territory, as in the Borges story, is no map at all.
In other words, both case studies and experimental design reduce the dimensions of a situation to gain control and clarity. The difference is that case studies strive to describe the interaction of as many dimensions as possible while still producing a coherent case, while experiments strive to isolate a few key dimensions and measure that reduced interaction (often the classic relation of dependent and independent variables, but not even multi-factorial analysis is going to deal with as many dimensions as a good case study). But we have to remember that the very clarity of experimental results comes from our choice of critical dimensions and our disregarding of the others. So I’d say that good experimentation needs a self-consciousness and openness to revision of our choice of critical dimensions, and that these virtues can be cultivated in a periodically renewed immersion in case studies.
Let me close with a challenging question I need to think more about: is it that X-phil methodology looks for “linear” relations between dependent and independent variables, whereas case studies try to describe the “nonlinear” feedback loops of our (Deleuzean, problematic) reality, which are resistant to the construction of dependent / independent variable experiments? So that it’s not really the quantity of dimensions that distinguishes experiments and case studies as their treatment of them?
Recent Comments