Dhanya Narayanan
Pronouns: she/her
Research Mentor(s): Sindhu Kutty
Co-Presenter:
Research Mentor School/College/Department: Computer Science and Engineering (EECS) / Engineering
Presentation Date: April 20
Presentation Type: Oral5
Session: Session 5 – 3:40pm – 4:30 pm
Room: Breakout Room 5
Authors: Dhanya Narayanan, Sindhu Kutty, Mithun Chakraborty
Presenter: 6
Abstract
Prediction markets are mechanisms that are used to provide insightful and reliable information about the future. Researchers are continually looking for tools that can improve the efficiency and accuracy of these markets. The goal of these markets is to aggregate information from individual traders to form a reliable prediction. Traders are financially incentivized to reveal their true beliefs via scoring mechanisms. One such mechanism is Hanson’s LMSR (Logarithmic Market Scoring Rule) Market Maker. Our research involves setting up an agent-based simulation of a prediction market ecosystem to study the efficacy of these markets. The simulation models a real-world market and has various parameters that can be altered to model different realistic scenarios. We can then run experiments to analyze the properties of the markets within such scenarios using probability theory and statistics. This analysis can tell us how accurate the aggregation mechanism is, and which specific features, if any, facilitate this. For example, will our prediction accuracy improve, or worsen, if we have a larger diversity of opinions among traders? While the market cares about its own prediction accuracy, traders are concerned with their own financial gain. Do traders with personal beliefs that are closer to the true probability always end up being compensated better? What if they were in the majority as opposed to the minority? All these questions, and more, can be explored within our set-up that can simulate both realistic and counterfactual scenarios.
Engineering
