My research interests are in the area of decision-making in human-centric complex systems, with emphasis on model-based approaches to system-of-systems design and assessment. My research found application in a variety of socio-technical domains like language modeling and search, systemic risk in financial markets, and smart scheduling and transportation systems.
Language Modeling, Search and Knowledge Acquisition
I work on the theoretical and computational aspects of a model-based integration of computational language modeling methods and cognitive decision making, reasoning, and search, through (Deep) machine learning methods like DSN, RNN, and (inverse) Reinforcement Learning.
Learning Knowledge Acquisition:
Knowledge acquisition is a non-Markov decision process of search over available sources of information for a given query. In this project, we first develop a framework to build (learn) a knowledge space using a set of textual documents. Then we exploit the rich structure of the resulting knowledge space to formulate the navigation behavior of an agent over the corpora in search of information as a sequential decision-making process. We develop an (inverse) reinforcement learning algorithm to learn the parameters of the navigation behavior as an online learning approach.
The process of searching for relevant legal materials is fundamental to legal reasoning. However, despite its enormous practical and theoretical importance, law search has been given inadequate attention by scholars. In this project, we define the problem of law search, examine its normative and empirical dimensions, and investigate one particularly promising computationally based approach. We implement a model of law search based on a notion of search space and search strategies and apply that model to the corpus of US Supreme Court opinions. We test the success of the model against both citation information and hand-coded legal relevance determinations.
Financial Markets Under Uncertainty
I work on developing a theoretical framework for modeling financial markets as a system-of-systems. This approach captures the dynamics of the interacting financial intermediaries from a competitive equilibrium model of a financial market that includes other participating agents like the regulators and non-financial institutions.
Modeling and Measuring Systemic Risk :
We propose a novel mathematical framework to model financial markets as complex systems of economic interactions and contractual and legal obligations aiming to understand, measure and mitigate systemic risk exposure of economy from the perspective of the policymakers and regulators. The proposed mechanism aims to exploit the systemic structure of contractual dependencies of the financial institutions in a competitive market where banks maximize their individual share of the financial market/profit. To do so, we need to develop an economy of such financial systems capturing the market dynamics and consistent with legal rules and regulatory frameworks.
Optimal Control Theory for Systemic Risk:
The 2007–8 financial crisis and its aftermath illustrated the fundamental shortcomings of the existing regulatory and policy-making approaches to recognize and mitigate systemic risk in financial markets and thereby maintain the stability of the economy. Ten years later, we still have no comprehensive framework to monitor/mitigate systemic risk and lack reliable and efficient methods to predict and avoid financial crises. This project aims for developing alternative regulatory methods to mitigate aggregate systemic risk exposure of the economy, subject to minimum regulatory and economic intervention and an economy of fixed size. We aim to show that, our methodology is a plausible and efficient framework to design better regulatory alternatives than the existing ones.