Algorithmic Bias, Economic Efficiency, and Social Welfare

Professor Wu, Yanhui   (Project Coordinator (PC))

Professor Li Jin   (Co-principal investigator)

Professor Xia Lu   (Co-principal investigator)

Professor Hao Yu   (Co-principal investigator)

Professor Liu Zhuang   (Co-principal investigator)

36

2025-06-15

3000000

Algorithmic Bias, Economic Efficiency, and Social Welfare

""Algorithmic bias"", ""AI for social good"", ""social media"", ""digital economy"", ""platform regulation""

EconomicsArtificial Intelligence and Machine learning

Business Studies (B)

C7003-24G

Collaborative Research Fund (CRF) - Group Research Project 2024/2025

2024

On-going

1. The intellectual objective of this project is to make substantial contributions to two emerging research areas: economics of AI and algorithmic social engineering. First, we will study algorithmic bias from the economics perspective, aiming to elucidate the economic and social sources of algorithmic bias and quantify its welfare effects. Second, we will develop novel theoretical models and empirical frameworks with implementable tools to solve challenging problems in welfare assessment and algorithm design. The key methodological innovation of this project is the two-way interactions between economics and data science. First, we will embed economic analysis into algorithm design to develop machine-learning methods and AI tools that aim to achieve the economic-social dual goal. Second, we will incorporate cuttingedge machine learning and algorithmic optimization methods into econometric analysis so as to gauge the extent of algorithmic bias and quantify its economic and social impacts. In multiple ways, the proposed project will produce high-impact research outputs that are aimed for publications at top economics, management, and statistics journals. 2. In the first work package, we aim to investigate how information addiction and self-isolation is driven by social media's algorithmic recommendation and what are the consequences and remedies. We will conduct surveys, observational studies, and field experiments with leading Chinese social media platforms: Xiaohongshu and Douyin. We aim to produce two to three high-quality papers published at top economics and management journals. 3. In the second work package, we aim to study the determinants of decision bias in AI-human interactions and the consequences of this bias with applications in medical diagnosis and legal decisions. The theoretical model involved is aimed at publications at top general-interest economics journals, while the two empirical settings (medical diagnosis and legal decisions) are aimed at publications at specialized journals in AI research, health economics, and legal studies. 4. In the third work package, we aim to develop an estimable framework to quantify the welfare effect of platform bias and apply it to the content-generation industries—online-novel writing and short-video platforms. We aim to produce two to three high-quality publications at leading economics and management journals. 5. In the fourth work package, we aim to develop a general framework that is flexible in incorporating various algorithmic designs and robust to the use of different machine-learning techniques. We then apply our method to two case studies related to the labor and Fintech markets. We aim to produce two or three high-quality publications at top statistics and economics journals. 6. The policy objective is to help regulators understand the pros and cons of algorithmic decision-making in terms of social welfare. Our project will generate regulatory applications with regard to social media, content-generative platforms, AI-based medical services, online labor markets, and Fintech. These applications are important for China and particularly relevant to Hong Kong. As a world leader in the digital economy, China is facing the urgent need of building a regulatory framework under which information and AI technology is used for social good rather than merely for self-interested firms and individuals. To thrive in the digital transformation, Hong Kong must establish its position as a high-tech hub with international regulatory standards. Our project aims to design a regulatory framework that promotes welfare-enhancing algorithmic design and mitigate the detrimental effect of algorithmic bias. The proposed research will not only provide intellectual support for advancement in this direction but also offer concrete solutions to some key regulatory problems encountered by policy makers and firms alike. Our project will generate regulatory applications with regard to social media, content-generative platforms, AI-based medical services, online labor markets, and Fintech. These applications are important for China and particularly relevant to Hong Kong. As a world leader in the digital economy, China is facing the urgent need of building a regulatory framework under which information and AI technology is used for social good rather than merely for self-interested firms and individuals. To thrive in the digital transformation, Hong Kong must establish its position as a high-tech hub with international regulatory standards. The proposed research will not only provide intellectual support for advancement in this direction but also offer concrete solutions to some key regulatory problems encountered by policy makers and firms alike. 7. The educational objective is to promote high-quality interdisciplinary study among research students and enhance education in the social sciences, business studies, data science, and civil engineering. This project cuts through many subject areas including economics, statistics, computer science, finance, business strategy, and information systems. It will set an example of interdisciplinary study for research students who are involved in the project. Some research outputs can be developed into teaching material for doctoral, master, and undergraduate levels of study in economics and business subjects as well as in data science and computer science.