Software applications, e.g., mobile apps, at times require multiple user interactions to fulfill certain actions. This disclosure describes an API that returns a predicted user action, e.g., selection of an option in a software application, in response to a query by the application. The prediction of user action is generated using one or more trained machine learning models. The models are trained, with user permission and expressed consent, on prior user interactions with various apps. When users permit, other contextual factors, e.g., data from device sensors, other apps that are running, operating system data, etc. can also be used as inputs for the trained models. The requesting application can present the predicted action as a default setting, or can automatically use the prediction as the user selection. In this manner, the techniques can enable apps to reduce user interaction, e.g., the number of clicks/taps to complete an action.
Feuz, Sandro and Carbune, Victor, "API for learning and predicting user interactions", Technical Disclosure Commons, (November 27, 2017)