Chirag Shah
University of Washington, WA·Computer Science
Smart assistants on phones and speakers mostly wait to be asked. They answer only what a user explicitly asks, so they work best for people who already know the right question.
- Fields
Computer Science
- Topics
conversational AI
conversational agents
proactive dialogue systems
- Grants
- 1 known grant, through May 2027
What they work on
- Build a proactive conversational agent on an offline hierarchical model-based deep reinforcement learning framework, in which a high level picks conversation strategies and a low level generates responses.
- Develop conversation state tracking and proactive action planning so the agent can decide when to join in, expand topics and knowledge, and manage multi-level task goals.
- Add a critique-and-validation step that checks proposed responses for opportunities and pitfalls, and evaluate the system through offline dataset experiments, crowdsourced judgments and user studies.
More about the research
This puts users with lower information literacy at a disadvantage. Tasks like exploratory searching, brainstorming and complex decisions also need a partner that actively contributes. This project, a collaboration with another institution, aims to build a conversational agent that can both answer questions and take part in proactive, two-way conversations. The agent would brainstorm, generate new insights and add ideas of its own.
Why it matters
Conversational assistants that contribute ideas, rather than only answering direct questions, could better support people who are unsure what to ask. Such systems could help with exploratory search, brainstorming and complex decisions, making digital help more useful to users with lower information literacy.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Anthony Sicilia
West Virginia University, WV·Computer Science
Develop a Probably Approximately Correct (PAC) learning theory that measures how user disagreement, suggestions and topic shifts change the calibration of an AI system's confidence during a conversation.
Prerna Juneja
Seattle University, WA·Computer Science
Build a taxonomy of user interactions with AI companion apps, covering motivations, challenges and values, by analyzing large-scale social media data and user interviews.
Anita Sarma
Oregon State University, OR·Computer Science
Identify the needs and expectations of students and instructors for a chatbot that supports open source software contribution, using participatory design.
The school
University of Washington: 81 labs on PI Pages, $1.7B a year on research, 694 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.