Putting Users at the Centre: A Path to Trustworthy AI Recommendations
Online shopping and entertainment platforms often provide recommendations based on your purchases/the films you viewed. Do you find them annoying or useful?
The latter is what Professor Chen Li from the HKBU Department of Computer Science, and the research group she formed strive for, in their research work themed “Towards Trustworthy Recommender Systems: From Users’ Perspective”. “If you don’t trust the system is doing for your good, you won’t be likely to use it,” said Professor Chen, who has been dedicated to developing user-centric recommender systems.
The Project’s keywords “From Users’ Perspective” clearly manifests Professor Chen as a caring researcher who aims for human well-being. Besides being an active researcher with a track record of over 120 published papers, most in high-impact journals and top-tier conferences, she is also a benevolent supervisor who has trained more than 20 research postgraduates and post-doctorate students, earning her the President’s Award for Outstanding Performance in Research Supervision 2022/23. Her trained students secured research positions in industrial labs or faculty positions at renowned universities.
Collaborating with her research students and researchers from communication, psychology, and business, this transdisciplinary project is based on Professor Chen’s research findings from some RGC-funded projects. “Over the past 14 years, I’ve been very fortunate to have received vast support, guidance, and encouragement from the Department and the University, leading to more opportunities of receiving external grants,” said Professor Chen.
Trust is Difficult to Build in Human-Computer Interaction
Trust is like a mirror, once broken, it never looks the same. Professor Chen found trust is even more difficult to build and much easier to lose in human-computer interaction. Her research team went beyond defining, measuring, and promoting appropriate user trust in recommender systems, focusing on the users’ perspectives.
The three-layered trust model Professor Chen’s team focuses on
Founded on current literature reviews, Professor Chen’s team aims to build a trustworthy recommender system by establishing an integrated user-centred trust evaluation framework. It considers system-related factors (explanation – reasons for the recommendations), user-related factors (users’ personality), and context-related factors (task complexity). For the system-related factors, the research team extracted sentiment (opinion) features from product reviews and combined them with static specifications for the explanation, and further developed personalised explanation generation methods based on pre-trained language models.
Multi-group model comparison between high and low curiosity groups
Professor Chen’s team realised the relation between the users’ personalities and the diversity of the recommendations for building trust, and conducted a large-scale user survey on Mobile Taobao (one of the most popular mobile e-commerce apps in China). The survey reveals that curiosity behaves as a significant moderator in strengthening the causal relationship from novelty to serendipity (i.e., recommending relevant but unexpected items), and from serendipity to satisfaction. As for context-related factors, the team investigated and discovered the key qualities of conversational recommender systems in influencing users’ trust.
Interfaces of two conversational music recommenders employing different initiative strategies
Trust the Process: Impact and Implications
Professor Chen’s team has been approached by some enterprises showing interest in their published papers and keynote presentations, and thus has various collaborations with platforms including Mobile Taobao (affiliated with Alibaba Group), Douban (a popular mainland China social media platform/online database that allows users to write books/films/music reviews), and Wisers (a local provider of all-media big data smart business intelligence solutions). Professor Chen’s studies have been proven to be constructive, bringing them more long-term collaborations.
Get through the Challenges: Trust Issues and Individual Differences
The challenges faced by the research team are exactly the team’s research objective – To develop and promote appropriate user trust in recommender systems. While people are generally concerned with Artificial Intelligence (AI)’s privacy and security issues, another critical challenge is modelling and understanding users, as substantial varieties exist among individuals’ interests, behaviours, and backgrounds. To tackle with, Professor Chen conducted a series of qualitative and quantitative studies, comprising user studies, co-design workshops, field studies, in-person interviews, online evaluations, and eye-tracking experiments for more in-depth understanding and more accurate measurement of user perception and satisfaction of their system.
Trust the Future to the Hands of Progress
Professor Chen, a curiosity-driven and well-planned researcher, is optimistic and enthusiastic about continuing her current studies and collaborations with the HKBU Faculty of Social Sciences and other enterprises. Looking forward, she will work with social work experts, exploring how her expertise could help with the mental health and emotional problems resulting from the COVID-19 pandemic, such as incorporating music therapy and narrative therapy into their system.
She has long-term plans to extend her research work in education (e.g., intelligent systems for language training) and digital humanities, “I hope AI technology can be further developed and incorporated into more domains, and more users like school children and elderly can enjoy using AI systems. It’s a long road ahead, but it’ll be an exciting journey that I’m looking forward to.”
About the Researcher
Professor Chen Li is currently a Professor and Associate Head (Research) in the Department of Computer Science at Hong Kong Baptist University (HKBU). She obtained her PhD degree in Computer Science from the Swiss Federal Institute of Technology in Lausanne (EPFL), Switzerland, and her Bachelor's and Master's degrees from Peking University, mainland China. Her recent research focus has mainly been on conversational and explainable AI, with applications covering various domains including entertainment, digital media, education, e-commerce, and psychological well-being.
She has authored and co-authored over 120 publications, with 8,900 citations so far (H-index 46). Her co-authored papers have received several awards, such as the CHI’22 Honourable Mention Award, UMAP’20 Best Student Paper Award, UMUAI 2018 Best Paper Award, and UMAP’15 Best Student Paper Award. She received the President’s Award for Outstanding Performance in Research Supervision 2022/23, and has been included in the list of the world’s top 2% most-cited scientists by Stanford University since 2021. She is now an ACM senior member, co-editor-in-chief of ACM Transactions on Recommender Systems (TORS), executive committee member of ACM Conference on Recommender Systems (RecSys), editorial board member of User Modeling and User-Adapted Interaction Journal (UMUAI), and associate editor of ACM Transactions on Interactive Intelligent Systems (TiiS). She also served as the general co-chair of ACM RecSys’23, the program co-chair of ACM RecSys'20, and the program co-chair of ACM UMAP'18.
Click here to learn more about Professor Chen’s research group and their studies about intelligent technologies for well-being.