This paper outlines how probability theory and the concept of randomness was elaborated in modern science, and became a central tool or framework in which disparate fields of scientific inquiry could be unified, particularly with the development of information theory, computability theory, and algorithmic information theory. It argues that there is a tendency to ideological reification of the concepts of randomness and probability when principles from physics and mathematics are applied to the social domain. It distinguishes between two senses of randomness and probability and then describes how this opposition has been overcome in the game-theoretical restatement of probability in finance, and in the current turn in cognitive science called predictive processing (PP) or the active inference framework (AIF), which understands cognitive systems as an approximately Bayesian hierarchically nested generative model of the environment. Finally, it argues that a similar ideological reification of randomness is at play in the AIF literature, in its failure to address the social conditions of uncertainty reduction. It concludes by suggesting that a politicized reconceptualization of AIF is possible, which would take into account the active power of practices and representations to construct different conditions of uncertainty reduction.
Inigo Wilkins (Tue,) studied this question.