Machine Learning for Financial Wellness: Human-Centric Recommender Systems

Financial wellness was introduced in 1989 and has gained considerable attention within academia and industry in finance and actuarial science. The scope of financial wellness is centered around five major themes: education, resources, decision-making and behavior, resilience, and values and goals. If AI is to serve investors by supporting them in their financial wellness, algorithm implementations must use investor modelling to truly capture their needs and behaviors.

One major avenue for supporting financial wellness is recommender systems; these include digital and virtual customer assistants for conversational recommendations regarding financial information, products, transactions, or advice. Current state-of-the-art recommender systems use a vast amount of user-item interaction data to infer users’ preferences. While it is essential to consider individual investors' preferences, this should not be the only objective of the model. Financial recommender systems must consider the risk and return of a recommended asset or portfolio relative to the existing investor portfolio and preferences, as well as future predictions of the asset. 

Despite the significant advancements in algorithmic AI and recommender systems, approaches that enable financial wellness must incorporate human-centric considerations to ensure methodologies are ethical, transparent, and aligned with human values. Human-centric AI emphasizes the importance of designing AI systems that prioritize user needs, enhance human decision-making, maintain a balance between automation and human oversight, and deliver explainable and transparent services to clients. In the context of finance, this means developing AI solutions that are explainable, fair, and capable of mitigating biases that could adversely affect individuals or groups. By focusing on the human aspect, we can ensure that AI-driven financial services are accessible, trustworthy, and beneficial to a broader range of stakeholders.

Investor demographics and trading behaviours can directly influence the guidance recommender systems provide, and thus the accurate understanding of investors through modelling is paramount. Recommender systems utilizing investor modelling can target different end-users, such as clients, financial advisors, dealerships, regulators, etc. Incorporating investor modelling into recommender systems designed for clients, households, advisors, and dealerships is pivotal to the future of financial decision-making. 

Currently, there is a gap between industry and academia in designing ML in investor modelling and recommender systems that integrate human-centric principles. This workshop will bring together academic and industry participants working on state-of-the-art research in designing recommender systems that incorporate investor preferences, demographics, historical and predicted trading behaviors, and financial details. This workshop builds on previous workshops on machine learning for investor modelling where we identified recommender systems as an emerging area of quantitative behavioral finance. This workshop reaches more broadly to bring together researchers and industry participants working on new methodologies in AI and recommender systems and share their recent results in this fast-moving field.

Key dates

• Paper submission deadline: 20th October, 2024

• Author notification: 29th October, 2024

• Workshop: 15th November, 2024

Organizing Committee

Prof. Andrea Kő (Corvinus University, Hungary)

Prof. Richard McCreadie (University of Glasgow, Scotland)

Mr. Thomas J. De Luca (The Vanguard Group, Pennsylvania)

Mr. Chuck Grace (Canada's Financial Wellness Lab, Canada)

Prof. Yong Zheng (Illinois Institute of Technology, Illinois) (Lead Organizer)

Prof. John R.J. Thompson (University of British Columbia, Canada) (Lead Organizer)