CAREER: Towards Conversational Recommendation Systems: Explainability, Fairness, and Human-in-the-Loop Learning

Project Details

Description

Recent advances in Artificial Intelligence (AI) have accumulated a rich toolbox of models for information retrieval, natural language processing and personalized recommendation. By optimizing over benchmark datasets, many of the models were developed with an algorithmic consideration instead of putting human as the central consideration. However, the ultimate goal of AI is to serve humans, collaborate with humans, and, ultimately, benefit humans. As a result, algorithmic approaches to AI must put humans in the loop for model design, implementation and validation. This project focuses on conversational AI, a promising approach towards putting humans in the loop, which enables direct conversation between human and AI for model learning. In particular, the project explores conversational recommender systems to help users in information seeking and decision making. It will develop explainable and fairness-aware algorithms for conversational recommendation. Presentation of the work and demos will help to engage with wider audiences that are interested in computational research. By integrating transparency and fairness principles into computer science courses on areas such as Information Retrieval, Data Mining and Artificial Intelligence, results from the project will educate students to understand how AI can be not only useful but also socially responsible.

This project will develop a general framework for conversational recommendation that bridges natural language understanding and dialog state management. With the framework, the project will explore three directions. The first direction aims at developing explainable conversation strategies based on human-machine collaborative reasoning, which brings cognitive ease to users and helps to build trust between human and AI. The second direction explores fairness-aware conversation strategies based on short-term and long-term fairness learning, which helps to achieve fair recommendation experiences between advantages and disadvantaged users. The third direction aims at developing a learning to evaluate protocol for conversational recommendation, which unifies the advantages of online crowd-sourcing and offline model learning for evaluation. The project will also develop a prototype conversational recommendation platform as a class project to support the education of responsible AI. The project will result in the dissemination of shared data and evaluation platforms to the Information Retrieval, Data Mining, Recommender System, and broader AI communities.

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.

StatusActive
Effective start/end date10/1/219/30/26

Funding

  • National Science Foundation: $335,771.00

Fingerprint

Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.