TalkToRanker: A Conversational Interface for Ranking-based Decision-Making
Conor Fitzpatrick - New Jersey Institute of Technology, Newark, United States
Jun Yuan - New Jersey Institute of Technology, Newark, United States
Aritra Dasgupta - New Jersey Institute of Technology, Newark, United States
Room: Bayshore I
Sunday, October 13th, 2024 @ 12:30GMT+00:00Change your timezone on the schedule page
6 months agoYour current time: Tuesday, Apr 8th @ 06:03
Full Video
Abstract
Algorithmic rankers have proven to be very useful in many real-world socio-technical systems, as they assist greatly in making decisions (e.g., who to hire, who to admit). Our conversational interface, TalkToRanker, aims to empower non-expert information consumers to engage with algorithmic rankers via multi-modal conversations involving text and visualizations. We leverage explainable AI methods and the generative power of large language models (LLMs) for facilitating such conversations. We demonstrate the capabilities of TalkToRanker via interactive scenarios from the perspective of an admissions officer.