Another brick in the wall: Threats to our autonomy as sense-makers when dealing with machine learning systems

Abstract

Autonomy, as proposed by the enactive approach to cognition, is the capacity that living organisms have to follow norms constituted by their own activity. This concept is linked to the concepts of sense-making and participatory sense-making, the former encapsulating the cognizer's ability to bring forth a world of meaning through its coupling with the environment – and being affected by its surroundings on an ongoing basis – and the latter being an extension of this idea, which applies to interactive processes in which at least two agents find themselves involved in. In this essay I advocate that, when dealing with machine learning systems, which cannot be considered autonomous, the agent or cognizer cannot sustain his or her autonomy in the same way as would be possible in an encounter with another agent. The reasoning is developed in three threads: the unbalanced encounter in which the cognizer's autonomy is threatened; the reduction of the range of experiences an autonomous agent could have; the lack of awareness of the cognizer concerning rules and potential risks of the systems he is dealing with. Even though these risks are focused on in the essay, the opportunities offered by machine learning systems are also recognized. To take advantage of them, it is necessary to seek a balance that encompasses the inherent human capacity for intersubjectivity permeated by affectivity.