# Karl Popper on falsification of probabilistic statements

If in ten throws [of a die] I obtain « 1 » six times, then I will conjecture that something is wrong with the probability prediction [P(1) = 1/6]; if, say, in the very next five throws I get « 1 » two of three times again, I will be convinced that the prediction (for this case), and with its specific assumptions, are falsified: I will try to change the assumptions. Initially, I will not change the natural laws but assume that certain other assumptions were false. Above all, I will examine whether the die is not « loaded » (whether the centre of gravity of this die is at its geometric centre, etc.). (p. 152)

Popper, K. (2012). The two fundamental problems of the theory of knowledge. London: Routledge Classics. (Originally published 1978)

## Une réflexion au sujet de « Karl Popper on falsification of probabilistic statements »

1. Stéphane Vautier Auteur de l’article

This passage shows that, given that a probabilistic statement of form « P(x) belongs to ]0, 1[ » is not falsifiable since any sequence of events is possible (the probability of any sequence is strictly positive), Popper endorses a « minimal conventionalism », which allows one to take surprising outcomes as falsifiers.