Abstract
As AI systems enter into a growing number of societal domains, these systemsincreasingly shape and are shaped by user preferences, opinions, and behaviors.However, the design of AI systems rarely accounts for how AI and users shapeone another. In this position paper, we argue for the development of formalinteraction models which mathematically specify how AI and users shape oneanother. Formal interaction models can be leveraged to (1) specify interactionsfor implementation, (2) monitor interactions through empirical analysis, (3)anticipate societal impacts via counterfactual analysis, and (4) controlsocietal impacts via interventions. The design space of formal interactionmodels is vast, and model design requires careful consideration of factors suchas style, granularity, mathematical complexity, and measurability. Usingcontent recommender systems as a case study, we critically examine the nascentliterature of formal interaction models with respect to these use-cases anddesign axes. More broadly, we call for the community to leverage formalinteraction models when designing, evaluating, or auditing any AI system whichinteracts with users.