A seminar by Associate Professor Xi Dong from Baruch College
Title: Does Democratized Information Access Demand Democratized AI? A Tale of the Rich, the Poor, and the AI (ChatGPT)
with Xiumin Martin and Changyun Zhou
Abstract: Regulation Fair Disclosure aims to democratize access to information such as earnings calls, though it remains contentious whether such openness achieves its purpose. AI democratization adds fresh tension to such pursuit. We conduct the first analysis on the implications of democratized AI on human traders in utilizing democratized information for investment insights. Leveraging on the longest sample of 19-year earnings calls, we find that ChatGPT-generated AI-sentiment strongly predicts returns for 12 months. An AI-sentiment-based strategy generates a monthly alpha of 1%. In the years before ChatGPT exists, short-selling over the two weeks after calls already aligns with AI-sentiment, while retail-trading doesn’t. Post ChatGPT’s wide-deployment, however, retail alignment with AI-sentiment surges 65-fold, while short-sellers-AI alignment can even weaken. ChatGPT outages reduce the retail-AI alignment, further supporting the causal role of AI. Bid-ask spreads decrease for stocks with increased retail-AI alignment. Our findings suggest that democratization of long and complex textual information might not have effectively democratized information access, until the democratization of AI, which deciphers long-term investment insights valuable for the average Joe. This advocates for “open” AI (usage) to bridge the information gap between the privileged and the general public.
For further information, please contact RSFAS Seminars.
All information collected by the University is governed by the ANU Privacy Policy.