HM AND YES? HOW BACK-CHANNELING TRANSFORMS SMART-HOME DEVICES FROM PASSIVE AGENTS TO BE ACTIVE LISTENERS.
Open Access
- Author:
- Motalebi, Nasim
- Graduate Program:
- Information Sciences and Technology
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- August 15, 2019
- Committee Members:
- Saeed M Abdullah, Thesis Advisor/Co-Advisor
Carleen Maitland, Committee Member
Jeffrey Bardzell, Program Head/Chair
Benjamin Vincent Hanrahan, Committee Member - Keywords:
- Conversational Agents
Smart SpeakersHuman-Computer Interaction
Active Listening
Emotional Well-being.
Human-Computer Interaction
Smart Speakers
Emotional Wellbeing - Abstract:
- Smart Speakers assistants such as Amazon Alexa have become pervasive tools that support a variety of tasks including information search and controlling smart home devices. However, current use of smart-speakers is limited to episodic interactions (e.g., answering single query) that do not span more than a couple of user turn-takings. Moreover, user’s turn in talk is limited to short commands (mostly less than a minute); forcing the user to be positioned as a listener in most interactions. Such limitations narrow the use of smart-speakers to command driven interactions which reduces users’ turn in talk and as a result, reduces conversational engagement with smart-speakers. In this project, we aim to extend the capabilities of smart-speakers to support more engaging interactions with users. Towards this goal, we propose transforming smart-speakers to “Active-listeners”. More specifically, we have integrated the use of back-channeling (e.g., ‘hm’, ’yes’, ‘umm’) in smart-speakers to extend user’s turn in talk and to sustain interactions that contain multiple-turns with arbitrary durations. We believe the use of back-channeling will result in users perceiving these devices as active listeners and hence improving overall engagement and disclosure intimacy. Such improvement in engagement has potential impact across a number of application domains including effective and scalable delivery of mental health support such as self-centered therapy.