Thought experiments can be useful in suggesting new hypotheses and in identifying flaws in established theories. Thought experiments become harmful when they are used as intuition pumps to provide evidence for the acceptance of hypotheses. Intuitions are neural processes that are poorly suited to provide evidence for beliefs, where evidence should be reliable, intersubjective, repeatable, robust, and causally correlated with the world. I describe seven ways in which philosophical thought experiments that purport to establish truths are epistemically harmful. In the philosophy of mind, thought experiments support views that run contrary to empirically supported alternatives.
Thanks to Mike Stuart, Yiftach Fehige, and two anonymous referees for helpful comments. My research is supported by the Natural Sciences and Engineering Research Council of Canada.