Abstract

The no-miracles-argument (NMA) for scientific realism asserts that the only plausible explanation of the predictive and explanatory success demonstrated by mature scientific theories is that those theories are at least approximately true. Sprenger and Hartmann (2016, 2019) attempt a probabilistic reconstruction of the NMA to assess its status from a Bayesian perspective.

Their reconstruction shows that the stability of scientific theory over time is a crucial factor when assessing the NMA in a given context. Periods of theoretical stability substantially improve the persuasive force of the argument in favour of the scientific realist.

In this talk, I critically assess their reconstruction, which I call the 'two-variables NMA'. I argue that Sprenger and Hartmann assume that the predictive and explanatory success of a scientific theory is less correlated with its stability than what may be genuinely expected. I show that under more plausible assumptions about this correlation, the described improvement is not substantial.