TItle: Modeling and Improving User Engagement
Description: There are many mobile applications that wish to collect information from users such as ground-truth labeling, experience sampling, and crowdsourcing. A typical and convenient way to grab user attention is by popping out mobile phone notifications, which instruct users to provide inputs. However, ill-timed notifications can cause annoyance and distraction, hence decreasing users’ willingness to provide inputs. Widely common approaches to address interruptibility are postponing notifications to future transitions of activities, and filtering out undesired notifications based on content, user preference, or user context. However, skipping or deferring the notifications can cause biased inputs. For example, if a user enables ‘don’t disturb’ mode during meetings, we can never receive user responses during meetings. This project aims to model user engagement, i.e., learn human perception of interruption. To this end, we are going to answer the following questions: As a mobile device, when is a reasonable time to interact with users while the disturbance? How do we infer such a timing from mobile sensors or IoT devices? How do we keep track of user's cognitive status?
Status: Active Project
Main Research Area: Pervasive ComputingBack