-------------------------------------------------------------------------------------------------------



                 Department of Systems Engineering and Engineering Management

                                 The Chinese University of Hong Kong



-------------------------------------------------------------------------------------------------------



Date: Jan 20, Monday, 4:30p.m. - 5:30p.m HKT



Venue: ERB 513, The Chinese University of Hong Kong



Tilte: Eliciting Risk Aversion with Inverse Reinforcement Learning via 

Interactive Questioning



Speaker: Prof.Ziteng Cheng, HKUST(GZ)



Abstract:



This talk presents a framework for identifying an agent's risk 

aversion using interactive questioning. Our study is conducted in two 

scenarios: a one-period case and an infinite horizon case. In the 

one-period case, we assume that the agent's risk aversion is 

characterized by a cost function of the state and a distortion risk 

measure. In the infinite horizon case, we model risk aversion with an 

additional component, a discount factor. Assuming the access to a 

finite set of candidates containing the agent's true risk aversion, we 

show that asking the agent to demonstrate her optimal policies in 

various environment, which may depend on their previous answers, is an 

effective means of identifying the agent's risk aversion. 

Specifically, we prove that the agent's risk aversion can be 

identified as the number of questions tends to infinity, and the 

questions are randomly designed. We also develop an algorithm for 

designing optimal questions and provide empirical evidence that our 

method learns risk aversion significantly faster than randomly 

designed questions in simulations. Our framework has important 

applications in robo-advising and provides a new approach for 

identifying an agent's risk preferences.







Short bio:



Dr. Ziteng Cheng joined the FinTech thrust in HKUST(GZ) as an 

assistant professor in August 2024. Before joining HKUST(GZ), he was a 

postdoctoral fellow in the Department of Statistical Sciences at 

University of Toronto, mentored by Dr. Sebastian Jaimungal. Dr. Cheng 

got his Ph.D. in Applied Mathematics from Illinois Institute of 

Technology under the supervision of Dr. Tomasz R. Bielecki and Dr. 

Ruoting Gong. His research focus resides within the realm of 

stochastic processes, machine learning, and their financial 

applications.



Everyone is welcome to attend the talk!