Recommender systems have been widely applied in different scenarios of our daily lives. Existing recommender systems mainly focus on predicting the user’s future behaviors (e.g., rating, clicking, watching, and purchasing) items but ignore the interactions between the user and the recommender agent. Therefore, they usually generate un-satisfactory recommendation results and cause bad user experience. This motivates the development of interactive recommender systems (IRS) that encourage the interactions between the user and the recommender agent. In recent years, interactive recommender systems have attracted increasingly research attentions from both academia and industry. This workshop provides a forum for researchers and practitioners to share the new techniques and applications of interactive recommender systems, and discuss the future research directions.
Abstract
Recommender System (RecSys) has been heavily researched for more than 20 years. There are not short of RecSys models in research publications, or practical systems that support our daily needs. Although various experiments have been reported in research papers, the evaluation of RecSys has not been a key focus. In this talk, I will review the challenges in evaluating RecSys and then discuss the possible challenges in evaluating interactive RecSys.
Bio
Dr. Aixin Sun is an Associate Professor at the School of Computer Science and Engineering (SCSE), NTU Singapore. He received Ph.D. in Computer Engineering from NTU in 2004. Dr. Sun is an associate editor of ACM TOIS, Neurocomputing, an editorial board member of JASIST and the Information Retrieval Journal. He was a member of the best short paper committee for both SIGIR2020 and SIGIR2022. He has also served as Area Chair, Senior PC member or PC member for many conferences including SIGIR, WWW, WSDM, EMNLP, AAAI, and IJCAI.
Abstract
Trust in AI/autonomy has been identified as one central factor in effective human-AI interaction. Despite active research in the past 30 years, most studies have used a “snapshot” view of trust and evaluated trust using questionnaires administered at the end of an experiment. This “snapshot” view does not fully acknowledge that trust is a dynamic variable that can strengthen and decay over time. With few exceptions, we have little understanding of the temporal dynamics of trust formation and evolution nor of how trust changes over time due to moment-to-moment interactions with autonomy. In this talk, I will present the results of two studies examining trust dynamics in human-autonomy interaction. In study 1, we identify and define three properties of trust dynamics, namely continuity, negativity bias, and stabilization. The three properties characterize a human agent’s trust formation and evolution process de facto. In study 2, we propose a computational model of trust dynamics that adheres to the three properties and evaluate the computational model against existing trust inference models. Results show that our model provides superior prediction performance and guarantees good model explainability and generalizability.
Bio
Dr. X. Jessie Yang is a Richard Wilson faculty scholar and associate professor in the Department of Industrial and Operations Engineering, University of Michigan, with a courtesy appointment at the School of Information. She received her PhD in Mechanical and Aerospace Engineering (Human Factors) from Nanyang Technological University, Singapore, and was a postdoctoral fellow in the Computer Science and Artificial Intelligence Lab at MIT prior to joining U-M. Her research interests include human-autonomy/robot interaction and human factors in high-risk industries. Her work is supported by grants from the National Science Foundation, National Institute of Health, Department of Defense, including an NSF CAREER award, as well as from various industrial collaborations, including Boeing and GM.
Abstract
Search and recommendation systems suffer from the basic information asymmetry problem between the user and system. Conversation offers a natural and convenient way to bridge the two. Although various findings or models have been reported in research papers and deployed in real systems, the proactiveness of such conversation systems has been less investigated. In this talk, I will first introduce the proactive conversation paradigm and then discuss methods to equip conversational search and recommenders with the ability to interact with end users in a more proactive way.
Bio
Dr. Lizi Liao is an Assistant Professor of Computer Science at Singapore Management University. She received Ph.D. from National University of Singapore in 2019. Dr Liao’s research interests center on task-oriented dialogues, proactive conversational agents, and multimodal conversational search and recommendation as the application target. One of her work was nominated in the ACM MM Best Paper Final List in 2018. She serves as Senior PC member or PC member of prestigious conferences. She also serves as organizing committee members of ACM MM 2019, NLPCC 2022 and WSDM 2023, while chairs sessions at KDD and SIGIR.
This workshop welcomes submissions from both academia and industry researchers relevant to interactive recommender systems. Topics of interests include but are not limited to:
Submissions must be in PDF, up to 8 pages long (plus unlimited pages for references), and shorter papers are also welcome. All manuscripts should be formatted according to the new ACM format published in ACM guidelines, selecting the generic “sigconf” sample. Papers can be submitted through easychair.
All submitted papers will be peer reviewed by the program committee. Accepted submissions will be presented at the workshop and published on the workshop’s website, but will not be archived.
Activity type | Time | Title | Speaker/Author(s) |
---|---|---|---|
Opening | 1.30pm-1.35pm | ||
Invited talk | 1.35pm-2.05pm | On the Proactiveness in Conversational Search and Recommendation | Prof. Lizi Liao |
Invited talk | 2.05pm-2.35pm | On Challenges in Evaluating Interactive Recommender Systems | Prof. Aixin Sun |
Invited talk | 2.35pm-3.05pm | Combining Empirical and Computational Approaches to Model and Predict Trust Dynamic in Human-AI Interaction | Prof. X. Jessie Yang |
Coffee break | |||
Paper oral | 3.30pm-3.45pm | Minimalist and High-performance Conversational Recommendation with Uncertainty Estimation for User Preference | Yinan Zhang, Boyang Li, You Yuan, Chunyan Miao |
Paper oral | 3.45pm-4.00pm | Balancing Accuracy and Fairness for Interactive Recommendation with Reinforcement Learning | Weiwen Liu, Ruiming Tang |
Paper oral | 4.00pm-4.15pm | Leveraging Context-dependent Click Model for Off-Policy Evaluation of Ranking Policies | Haruka Kiyohara, Nobuyuki Shimizu, Yasuo Yamamoto |
Paper oral | 4.15pm-4.30pm | Dynamic Adaptation of User Preferences and Results in a Destination Recommender System | Asal Nesar Noubari, Wolfgang Wörndl |
Paper oral | 4.30pm-4.45pm | Improving Recommendation Systems with User Personality Inferred from Product Reviews | Xinyuan Lu, Min-Yen Kan |
Paper oral | 4.45pm-5.00pm | Counterfactual Explanations for Visual Recommender Systems | Neham Jain, Vibhhu Sharma, Gaurav Sinha |
Closing | 5.00pm-5.05pm |