Samenvatting

Aletta Smits, professor in Human Centred Technology, zoomed in on human-AI interaction: how to give people autonomy and control. She introduced us to terms like recommender systems: any tool that helps you select one or more from a host of options; and decision support systems: any tool that helps you select which of a limited set of options is best. She gave several examples, familiar to most of us, like Netflix, Google and Spotify, that adjust results based on your input, like Netflix recommendations and Spotify Discover Weekly. She highlighted the interaction domain of the three AI domains Victor introduced, which were: the technical domain, legal domain and user domain. Aletta zoomed in on the user domain; the interaction between the user and AI. To the example of Netflix and Spotify, in which the user’s usage of the platform steers the algorithm, she proposed the term of algorithmic affordances. Affordances provide strong clues to the operations of things. In essence, affordances are aimed at the user’s influence on what gets recommended to them on a platform. This sense of agency, the experience of having control over one's actions, and responsibility is vital in the success of AI. The perception of users about the interactions they experience. They determine aspects such as: trust, return to tool, sense of control, confidence to act, and ownership/accountability. In any decision support system, algorithmic affordances matter.