**Links to**: [[Active inference]], [[Ignorance]], [[De Docta Ignorantia]], [[Argument from silence]], [[12 Negintelligibility]], etc. # Active ignorance On the AIF account intelligent systems are negentropic systems that achieve prediction error minimization through the generation and revision of compressed representations. But this leaves us with some questions as it might reduce thought to a simplistic image of deviation from expectation. Which, in turn, promotes an _optimizing_ perspective on intelligence that, without special attention to what model-“revision” entails at the material-historical level, misses aspects of the permanent constructibility of thought’s own conditions. Our contention is that just as the complexity of physical systems is constrained by the inexorable tendency towards entropy, intelligent systems must build themselves out of, or in spite of, a compulsive tendency towards ignorance. This much is tacitly present in the AIF account, but there are several senses to understanding this tendency to ignorance. Firstly, every intelligent system operates within spatial, temporal, material, and information-computational constraints, so every intelligent system is ignorant of at least some processes or contingent outcomes that lie outside of the bounds of what is intelligible to it. But this, in some sense _passive_ ignorance, or constitutional finitude, is at the social-system level far exceeded by the _active_ processes of ignorance. Intelligent systems are actively ignorant firstly because the representation, or ‘understanding’ of an object demands its abstraction, so that the constituting details of an object’s features become ignored. This active ignoring is functionally necessary, revisionary-constructive, and largely (but not always) unconscious. However, there is another more important way in which intelligence tends towards ignorance – and that is the way in which a certain understanding, or compressed representation, which is ‘good enough’ as a model of the object (and reinforced by confirmation bias or model performativity), which forcefully obscures and excludes _other_ possible understandings. If AIF distinguishes between mere pattern recognition (correlation) in machines versus the causally-interested generative models in humans (Parr, Pezzulo, Friston 2022, p. 219), we need to dive deeper into what constitutes ‘interesting causation’—or an interest in causality—for situated human beings with historically-problematic models of reality, which tend to suppress that which does not coincide with the accepted tolerances of the time. <small>Keywords: active inference, active ignorance, salience, attention, patterns, intelligence, ideology critique.</small> This chapter is currently _out of order_, on embargo, as we are currently writing the paper on Active Ignorance, together with Jimena Clavel and Inigo Wilkins. Please refer back to the website entry on Active Ignorance (at www.n-o.ooo) to find the forthcoming publication. %% [[Active ignorance notes]]