Inquietudini digitali. Il problema del Black box effect

The essay analyses the black box effect, which refers to AI systems, devices, or programs that provide useful information without revealing their internal workings. The greatest risk is that it is unrealistic to expect that machine-learning decision support systems will be able to generate, under all circumstances, complete explanations for the predictions they make (immanent opacity). The incommensurability of algorithmic systems derives from their mathematical structure, because there is no algorithm to decide whether and when a program will stop (mathematical opacity). Many areas of our society are opaque. The opacity of algorithms increases as the opacity of society