CMU Secure Blockchain Summit Speakers

Talk title

Incentivizing Decentralized Content Contribution in Information Aggregation Pipelines

Abstract

In many decision-making processes, a decision maker summarizes content from multiple content providers. Examples include DAOs (which often summarize a discussion among delegates) and LLM search engines (which provide a summary of input documents). However, it is often unclear how to fairly assign credit to content providers. Various DAOs (e.g., MakerDAO, Optimism, and Hop) have experimented with different ways of rewarding delegates. However, these methods either (a) rely on human-intensive evaluations of contributions from delegates and/or (b) rely on metrics that can easily be gamed, such as number of posts or votes. LLM search engines, on the other hand, sometimes provide crude, automated estimates of relevance for various sources. In this project, we aim to study new methods for assigning value to content providers' contributions based on their alignment with a final output explanation. Concretely, we explore natural language processing techniques for attributing value to a final summary of multiple input documents, taking inspiration from the classical Shapley value. We evaluate these algorithms on applications to LLM question answering.

Bio

Giulia Fanti is the Angel Jordan Associate Professor of Electrical and Computer Engineering at Carnegie Mellon University. Her research interests span the security, privacy, and efficiency of distributed systems, and she is a co-director of CyLab-Africa and the Initiative for Cryptocurrencies and Contracts. She has served on the Department of Commerce Information Security and Privacy Advisory Board and is a two-time fellow of the World Economic Forum’s Global Future Council on Cybersecurity. Her work has been recognized with several awards, including best paper awards, a Sloan Fellowship, an Intel Rising Star Faculty Award, and an ACM SIGMETRICS Rising Star Award.