Keynote talks
Dr Qian Zhao is an Engineering Manager in Bloomberg's AI Engineering group, which consists of 350+ researchers and engineers responsible for applying technologies such as machine learning, natural language processing, dialog understanding, graph analytics, time-series analysis, information retrieval, recommendation systems, speech recognition, computer vision, optimization, and generative models to applications in the finance sector. Some of the product areas that the AI Engineering group has built AI solutions for includes News, Research, Communications, and Finance. Bloomberg's AI Engineering group is also an active contributor to the academic community; its researchers have published more than 100 peer-reviewed papers in the last three years.
Before joining Bloomberg, Qian earned his Ph.D. in computer science from the GroupLens lab at the University of Minnesota, where he was advised by Professor Joe Konstan. Qian's research interests are focused broadly on the intersection of machine learning and human-computer interaction (e.g., user-centered design and evaluation of online interactive recommender systems). Before that, he worked in industry in China for a few years, where he focused on large-scale data mining and machine learning systems for video recommendation and computational advertising products.
Keynote Talk: User-Centric Research on Financial Recommender Systems
The Bloomberg Terminal delivers a diverse array of information, data, news, and analytics to facilitate decision-making by the world's leading capital markets professionals (e.g., investment bankers, institutional investors, etc.). In this talk, I’ll briefly introduce how the AI Engineering Group at Bloomberg utilizes AI technologies to enhance the Bloomberg Terminal product. Then, in the spirit of this workshop, I’ll focus my talk on the user-centric research related to financial recommender systems. First, I’ll try to explain what I mean by user-centric research and share some interesting findings from my past work in this field. I’ll then dive into the financial recommender systems domain and walk through some inspiring social theories in the literature about financial wellness, and consumers’ AI/RecSys/technology acceptance and adoption. I’ll highlight some very interesting prior work in this still understudied field. Finally, I’ll explore the question of why it is understudied in terms of challenges, and propose the exploration of some exciting ongoing opportunities in accessing real systems and optimizing for performance vs. diversification.
Dr Ekaterina Svetlova is an Associate Professor at the BMS faculty of the University of Twente. Her interdisciplinary research sits at the intersection of finance, ethics, Science and Technology Studies (STS) and economic sociology. Ekaterina’s projects focus on financial models, valuation studies, and AI ethics, as well as on risk reporting by firms, central governments and local authorities. Ekaterina is the author of the book Financial Models and Society: Villains or Scapegoats? (2018) and the co-author of the book Chains of Finance: How Investment Management is Shaped (2017).
Keynote Talk: Ethics of ignorance and transparency requirements for recommender systems in finance
The talk argues that the increasing use of recommender systems in finance should bring the concept of ignorance to the forefront of ethical and regulatory debates about transparency. Using examples of AI-based financial products, the paper reasons that a key challenge for designers, marketers, and investors is to communicate issues that are not, or cannot be, fully understood due to the inherent complexity of the products. Severe epistemic insufficiencies, such as data quality, algorithm behavior in unexpected situations, and the reasons behind algorithmic decisions, should place ignorance as a central issue in ethical and regulatory considerations. However, ethicists and regulators often overlook this challenge, while calling for comprehensive understanding and knowledge of what is going on within and around an algorithm. It is why they require humans-in-the-loop assuming that AI-related knowledge limitations are difficult yet surmountable obstacles for humans in the pursuit of accountability, traceability, transparency, and explainability ideals.
The talk engages with literature that questions the attainability of such ideals, suggesting that expectations of full transparency and explainability may be overly demanding for human (market) participants. While acknowledging the moral duty to know and inform, the essay explores ways to elevate ignorance from a perceived negative quality to an accepted element in ethical and regulatory discussions about recommender systems in finance, and digital finance more generally. The paper examines how ignorance might be embraced in feasible regulatory norms and ethical obligations, particularly for digital finance professionals tasked with conveying both the capabilities and limitations of new technologies to clients.
Mr Stephen Choi is the Co-Founder and CEO of IRAI, a startup focused on leveraging AI for crypto market curation. Previously, he co-founded and served as CTO of Asklora, an SFC-licensed stock trading platform that utilizes Generative AI to emulate a private banking experience. IRAI's LLM-ADE ("lemonade") framework aims to disrupt AI in finance by modifying and optimizing the architecture of open-source foundational LLMs. Before his entrepreneurial ventures, Stephen held senior positions as a quantitative trader and fund manager at leading financial institutions including Merrill Lynch, HSBC, Citi, Kyobo AXA, and Mirae Asset. His research work has been published in peer-reviewed financial journals and presented at AI conferences. Stephen holds an MS in Financial Mathematics from the University of Chicago and a bachelor's degree from Northwestern University. His unique blend of expertise in quantitative finance, AI, and entrepreneurship drives his passion for innovating at the intersection of technology and finance.
Keynote Talk: Personality-Driven AI: Enhancing LLM Performance in Financial Services
This presentation explores the intersection of behavioral finance, personality psychology, and artificial intelligence in the context of financial services. Drawing on insights from the PolyU-Asklora Fintech Adoption Index, we demonstrate how the Five-Factor Model (OCEAN) can be applied to understand and potentially influence retail investment behavior. We also introduce ConFIRM, a novel methodology that utilizes the OCEAN personality model to efficiently generate representative synthetic datasets for fine-tuning Large Language Models. This approach offers promising applications for developing more personalized AI-driven services for financial wellness.
Dr Asefeh Asemi is an Associate Professor at Corvinus University of Budapest's Institute of Data Analytics and Information Systems. Her expertise lies in financial data analytics, machine learning, and the development of advanced recommender systems. Dr. Asemi’s research focuses on applying AI-driven methods to enhance financial decision-making, particularly through neuro-fuzzy inference systems and personalized financial tools. She has led numerous research projects in collaboration with financial institutions, developing data-driven strategies to optimize investment outcomes. In addition to her research, she has played an integral role in academia, supervising postgraduate students and contributing to the development of future financial technologies. Dr. Asemi is also an active participant in the global research community, with frequent contributions to international conferences and journals in the field of FinTech and data science.
Keynote Talk: Behavioral Insights of Investors Using Generative AI: A Case Study Based on Survey Data
The rapid evolution of financial markets and the increasing complexity of investor behavior present unique challenges for developing personalized and human-centric financial recommender systems. In this keynote presentation, we explore a novel application of Generative AI (GPT-3/4) to analyze survey data on investor behavior, focusing on their risk preferences, financial decision-making patterns, and behavioral tendencies. Utilizing the comprehensive dataset derived from a detailed investment questionnaire, we apply Generative AI models to uncover hidden patterns in the responses, providing insights into how different demographic and psychological factors influence investment strategies. By analyzing the behavioral data, this study reveals how AI-driven insights can support financial wellness by generating personalized recommendations that are sensitive to each investor’s unique needs. Furthermore, the presentation will highlight how human-AI collaboration can enhance the effectiveness of investment decision-making by combining the power of AI-driven data analysis with expert human oversight. This approach offers a fresh perspective on the future of ethical, transparent, and human-centric AI in the realm of financial advisory systems.
Mr Yongjae Geun works in the AI Innovation Team at the Korea Financial Telecommunications and Clearings Institute (KFTC), which has various financial data such as online banking transactions and electronic bonds. Leveraging this data, he is responsible for business strategy and planning to advance the Korean financial industry through ML and generative AI. Recently, he has led projects on alternative data for credit scoring and fraud detection services and has collaborated closely with financial authorities. Through these experiences, he has developed a deep understanding of the Korean financial industry and is actively contributing to AI and data-driven financial services.
Keynote Talk: Alternative Data-driven Credit Scoring Approach in KFTC
In the credit scoring domain, alternative data not traditionally used is emerging as a valuable resource to support financially disadvantaged groups. This presentation introduces the process of identifying alternative data from KFTC sources, such as automatic transfers, bills, and bonds, and explores how financial institutions can leverage this data for ML-based credit scoring. We will discuss how alternative data can help financial institutions overcome the limitations of traditional approaches, contributing to risk management and personalized services. Our presentation will also cover data pre-processing, alternative data analysis, and case studies of financial applications.
Accepted papers and invited presentations to be announced
Mr Juchan Kim is a master’s student in the Department of Industrial Engineering at Ulsan National Institute of Science and Technology (UNIST). He holds a bachelor’s degree in mathematics from the same institution. He has a strong interest in optimization and ML/AI model development for financial services and has participated in several industry projects. Currently, he is analyzing trading data from millions of individual investors at one of South Korea's largest securities firms and developing ML/AI models tailored to retail investors.
Accepted Paper: The Taxonomy of Retail Investors with co-author Yongjae Lee (UNIST)
Recent events in finance domain highlighted the growing influence of retail investors. Although there have been many studies on retail investors, most have treated them as a uniform group or categorized them solely based on demographic data. This method fails to account for the considerable heterogeneity within the retail investor, which leads various characteristics compared to institutional investors. In this paper, we aim to establish a taxonomy of retail investors by classifying their diverse trading patterns using a combination of investment-theoretic and practical metrics. Specifically, we focus on how individual investors prioritize six key metrics when making trades: momentum, maximum drawdown (MDD), maximum drawup (MDU), maximum daily return, volatility, and skewness. We calculate these six metrics for each investor and apply clustering analysis to identify distinct investor profiles. Through this approach, we aim to (1) define distinct investment styles among retail investors, (2) evaluate the performance differences across these styles, and (3) track how these investment patterns change over time. Our analysis seeks to provide a deeper understanding of the heterogeneity among retail investors, ultimately laying the groundwork for developing truly personalized investment solutions. Our analysis aims to offer a deeper insight into the diversity among retail investors, ultimately paving the way for creating genuinely personalized investment solutions.
Mr William Gazeley is the CTO at IRAI Labs, focusing on domain-specific applications of Large Language Models. Previously Head of AI in the financial sector, his work bridges academic AI research with industry implementation.
Accepted Paper: Conversational Factor Information Retrieval Model (ConFIRM) with coauthors Stephen Choi (IRAI Labs), Siu Ho Wong and Tingting Li (LORA Research)
This paper introduces the Conversational Factor Information Retrieval Method (ConFIRM), a novel approach to fine-tuning large language models (LLMs) for domain-specific retrieval tasks. ConFIRM leverages the Five-Factor Model of personality to generate synthetic datasets that accurately reflect target population characteristics, addressing data scarcity in specialized domains. We demonstrate ConFIRM's effectiveness through a case study in the finance sector, fine-tuning a Llama-2-7b model using personality-aligned data from the PolyU-Asklora Fintech Adoption Index. The resulting model achieved 91% accuracy in classifying financial queries, with an average inference time of 0.61 seconds on an NVIDIA A100 GPU. ConFIRM shows promise for creating more accurate and personalized AI-driven information retrieval systems across various domains, potentially mitigating issues of hallucinations and outdated information in LLMs deployed.