I’m Itai
I'm a CS graduate student at Harvard University, supported by the Nicole A. Chen and Karina A. Chen Graduate Student Research Fellowship. I am privileged to be advised by Prof. Ariel D. Procaccia.
Before joining Harvard, I worked for 4 years as a Quant for WorldQuant (under Millennium), where my main focus was alpha research in equity markets. Prior to that, I completed my undergraduate degree in Math and Economics at the Hebrew University of Jerusalem (HUJI), graduating as valedictorian. I was fortunate to work with Prof. Jeff Rosenschein and Prof. Assaf Romm during my time at HUJI.
Research
Generative Social Choice
Itai Shapira, Sara Fish, Paul Gölz, David C. Parkes, Ariel D. Procaccia, Gili Rusak, Manuel Wüthrich, 2023 [Working paper]
The paper introduces "generative social choice", blending traditional social choice theory with the generative power of large language models (LLMs) to facilitate collective decisions on complex topics. This framework divides design into two parts: first, ensuring rigorous representation using an "oracle" LLM for queries; and second, empirically validating these oracle queries using real-world LLMs. The approach is demonstrated by using LLMs to generate representative statements from free-form text in online deliberative settings.
[June 2023]: : Our project has been awarded a grant from OpenAI for developing experiments in establishing democratic processes that determine the rules AI systems abide by.
Expressivity of Shallow and Deep Neural Networks for Polynomial Approximation
Itai Shapira, 2023 [Arxiv]
This research analyzes the number of neurons required by ReLU neural networks to approximate multivariate monomials, exploring the expressive power of depth versus overall complexity.
Optimal Engagement-Diversity Tradeoffs in Social Media
Itai Shapira, Fabian Baumann, Daniel Halpern, Ariel D. Procaccia, Iyad Rahwan, Manuel Wuthrich, 2023 [Arxiv]
This paper investigates whether echo chambers are an inevitable result of high engagement in social media platforms, providing theoretical and empirical results on the engagement-diversity tradeoff.
Market Power and Dropping Strategies: A New Approach to Stable Matching Mechanisms
Itai Shapira, 2023 [PDF]
Introducing the dropping lemma and the concept of market power, this paper simplifies the analysis of one-to-one matching markets, providing insights on strategy, market power, and incentives in matching mechanisms.