Many employees question the effectiveness of workplace meetings, despite their importance to collaborative decision-making. Sometimes attendees have trouble staying focused, or else one individual dominates the conversation; in both cases, productivity is down. The ability to conduct a meeting and foster effective participation by all is influenced by subjective and unquantifiable factors such as relationship dynamics, gender biases, and willingness to accept criticism. This research will explore whether it may be possible to avoid these complications by using an impersonal entity - the computer - to conduct group meetings and provide both live and post-meeting feedback that enables participants to stay on topic, take part equally, and listen closely to one another, all while generating analytics about the process. The framework that will be developed in this research may serve as a tool for improving meeting efficiency and effectiveness, while opening up the possibility of designing interventions for individuals with Asperger syndrome or social phobia who find it difficult to assert themselves in group settings. Providing live feedback during a meeting will require real-time recognition and analysis of the sensed data, but live feedback is only valuable when its interpretation does not impose a cognitive overload on the participants. This research will make contributions on designing live feedback that is effective yet not distractive. Project findings will also be applicable in the context of online learning, where a computer facilitator could engage every student equally. In addition to standardized test scores, computers can also create objective analytics on, for example, students' ability to ask questions, participate effectively and maintain a positive attitude. Recognizing students for their effort may positively impact their scores. Project findings may help virtual assistants (such as Alexa or Google Home) mediate conversations with appropriate and respectful intervention strategies, and in addition they may yield insights into the role played by gender and racial diversity in improving team performance.
The interplay of human behaviors (e.g., facial expression, tone of voice, language, interruptions, participation rates, and turn-taking) is considered subtle, sometimes contradictory, and often confusing. These behaviors can be highly individual-specific, and may require temporal modeling with real-time performance if they are to be harnessed automatically for effective meeting mediation. While the raw sensed numbers are useful when developing algorithm benchmarks, they add very little value for users. This research will use principles of human centric computing to visualize, interpret and reveal insights from raw behavioral data. In particular, the following research questions will be addressed using human-centric approaches in the context of online video conferencing: 1) What possible mediation strategies exist that avoid being disruptive and disrespectful to human participants; 2) Which is most helpful: live feedback, post-meeting feedback, a combination of both, or no feedback at all; 3) Can we implement an automated process to facilitate meeting productivity; and 4) Does feedback improve group performance in any measurable way?
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
