Alireza Fatollahi

Newton famously rejected the use of hypotheses in natural philosophy, in stark contrast to many of his contemporaries, such as Descartes, Huygens, and Leibniz, who employed hypothetical methods. This disagreement is often framed as one concerning the Hypothetico-Deductive (HD) method, but I argue that this is mistaken. The relevant hypothesis-based methods at issue were what I call inference to the best hypothesis and its stronger version, inference to the only plausible hypothesis. These methods were far more nuanced and plausible than HD, and they enjoyed widespread popularity among early modern thinkers, even among prominent experimental philosophers. Newton rejected them, nonetheless.

This essay critically examines well-known arguments against Bayesian Information Criterion (BIC). These arguments claim that I) BIC violates axioms of probability calculus in the case of nested families of statistical hypotheses (‘models’). II) BIC cannot explain why scientists sometimes prefer models with fewer adjustable parameters. And III) BIC’s verdict on a model is sensitive to arbitrary choices of priors over members of models or their parameterization. We argue that although these criticisms contain valuable cautionary advice against the improper use of BIC, they have no force against its general and proper use.

This essay critically examines a widely held assumption in interpreting Leibniz's modal metaphysics: that whatever is necessarily actual is necessary. I argue that Leibniz rejected this axiom for principled reasons having to do with his views on the grounding of metaphysical modalities in divine power and intellect (but not divine will). I also argue that if we read him in light of this rejection, his per se possibility theory becomes (contrary to its reception in the literature) quite successful in addressing the problem of necessitarianism.

The “No Miracle Argument” for scientific realism contends that the only plausible explanation for the predictive success of scientific theories is their truthlikeness, but doesn’t specify what ‘truthlikeness’ means. I argue that if we understand ‘truthlikeness’ in terms of Kullback-Leibler (KL) divergence, the resulting realist thesis (RKL) is a plausible explanation for science’s success. Still, RKL probably falls short of the realist’s ideal. I argue, however, that the strongest version of realism that the argument can plausibly establish is RKL. The realist needs another argument for establishing a stronger realist thesis.

Forster and Sober (1994) introduced the “sub-family problem” for model selection criteria that recommend balancing goodness-of-fit against simplicity. This problem arises when a maximally simple model (family of hypotheses) is artificially constructed to have excellent fit with the data. We argue that the problem arises because of a violation of the general maxim that balancing goodness-of-fit against simplicity leads to desirable inferences only if one is comparing models for the consideration of which one has a positive reason independently of the current data.

There has been a lively debate in the philosophy of science over _predictivism_: the thesis that successfully predicting a given body of data provides stronger evidence for a theory than merely accommodating the same body of data. I argue for a very strong version of the thesis using statistical results on the so-called “model selection” problem. This is the problem of finding the optimal model (family of hypotheses) given a body of data. The key idea that I will borrow from the statistical literature is that the level of support a hypothesis, _H,_ receives from a body of data, _D,_ is inversely related to the number of adjustable parameters of the model from which _H_ was constructed. I will argue that when _D_ is not essential to the design of _H_ (i.e., when it is predicted), the model to which _H_ belongs has fewer adjustable parameters than when _D_ is essential to the design of _H_ (when it is accommodated)_._ This, I argue, provides us with an argument for a very strong version of predictivism.

This paper discusses two conservative ways of treating evidence. (I) Closing inquiry involves discounting evidence bearing on one's belief unless it is particularly strong evidence; (II) biased assimilation involves dedicating more investigative resources to scrutinizing disconfirming evidence (than confirming evidence), thereby increasing the chances of finding reasons to dismiss it. It is natural to worry that these practices lead to irrational biases in favor of one's existing beliefs, and that they make one's epistemic condition significantly path-sensitive by giving a bigger role to batches of evidence obtained earlier in the course of inquiry compared with those subsequently acquired. However, I argue that both practices are demanded by considerations of practical rationality. I also argue that, contrary to initial appearances, there is little reason to worry about the effects of these practices on the dynamics of one's beliefs.