8/ 🌐 Want to know more about the @AMLEDS1 initiative? Check out the full program, past recordings & upcoming webinars here: 🔗 https://t.co/Wp7YEQQFFi #EconTwitter #webinar #LLMs #AI #AMLEDS #Datascience #MachineLearning

7/ Don’t miss this conversation at the frontier of monetary policy + #AI + multi-agent systems. 🗓️ Friday, Nov 21 🕚 11 AM ET / 5 PM CET 🎙️ @TaraSinc & @SophiaKazinnik 🎤 Moderator: @dahi_chuda #MonetaryPolicy #Macroeconomics #AI #LLMs #CentralBanking #DataScience #EconTwitter

6/ 🚀 The result? A credible in silico environment to: — test counterfactual monetary policy scenarios — explore institutional design — study persuasion, norms & career incentives — compare behavioral vs. rational outcomes #EconTwitter #LLMs #AI #NLP

5/ 🔥 Findings • In baseline conditions → both tracks converge within the 4.25–4.50% policy range • Political-pressure → more dissent, greater dispersion • Labor-market data revisions → dovish drift but still inside the range #LLMs reveal subtle frictions #EconTwitter #AI

4/ Both systems start from identical priors constructed using: • macro indicators • Beige Book intelligence • #FOMC speeches • member-specific profiles This allows a clean comparison of behavioral vs. rational dynamics. #EconTwitter #LLMs #AI #NLP

3/ 🔍 What’s the paper about? It builds a dual-track framework to simulate FOMC decision-making using: 1️⃣ #LLM-driven agents Debate, persuasion, personality heterogeneity, & real-time macro data → all embedded in deliberation. 2️⃣ A Bayesian Monte Carlo voting model #EconTwitter

2/ A small plot twist: due to a last-minute change, the session will be moderated by @dahi_chuda (@BrandeisU) — proving once again that macroeconomists are exceptionally good at handling shocks, even outside DSGE models. #EconTwitter #LLMs #AI #NLP

1/ 📢 This Friday! Join us for the next #AMLEDS #webinar featuring @TaraSinc (GWU) presenting her #FOMC In Silico" paper with @SophiaKazinnik (@Stanford). 🕚 11AM ET-5PM CET 💻 Zoom 📄 Paper → https://t.co/TXV4N68M3k #EconTwitter #AI #LLMs #NLP

🚨 Happening tomorrow! 🚨 🕔 11am EDT(5pm CEST) @ellliottt (ETH) joins @AMLEDS1 to explore “#AI #Preferences & #Economics.” Modern AI isn’t just about language — it’s about human preferences. Implications for economics & society? 💻 https://t.co/pMI5vJsVrx #LLMs #EconTwitter

7/ #AI #Economics #MachineLearning #SocialScience #LLMs #GenerativeAI #AMLEDS #Research #ETH #LSE #EconTwitter #EconLinkedin #ArtificialIntelligence #BigData #Econometrics #TextMining #TextAsData

6/ Don’t miss this chance to hear from Elliott Ash (@ellliottt), one of the leading voices bridging #AI and #economics. 📅 Friday, Oct 24 | 🕚 11am EDT / 5pm CEST 🔗 Register now: https://t.co/pMI5vJtth5 #EconTwitter #Webinar #AMLEDS @AMLEDS1

5/ This session will interest anyone working at the crossroads of: 🧠 #AI & #economics 📈 Machine learning & empirical social science 🗣️ Textual data, #LLMs, and preference modeling #EconTwitter #AI, #Machinelearning #Datascience #Textualdata

4/ 🎙️ Moderated by Roberto-Rafael Maura-Rivero (@rrmaura) (@LSEEcon) 📚 Part of the @AMLEDS1 series on Applied Machine Learning, Economics, and Data Science. #EconTwitter #AI #Webinar

3/ In this #AMLEDS #webinar, Elliott Ash (@ellliottt) (@ETH_en) will explore how #AI’s underlying modeling of preferences opens new frontiers for: * Economic modeling & behavioral research * Aligned AI system design * The synergy between AI & human decision-making #EconTwitter

2/ 💡 Modern #AI systems — from #NLP to #LLMs and #GenerativeAI — aren’t just models of language. They are models of human intent and preference. What does this mean for #economics? For #research? For the design of aligned AI systems? #EconTwitter #AMLEDS

🧵 1/ 🔹#AMLEDS #Webinar | #AI, Preferences&Economics🔹 📅 Friday, Oct 24 🕚 11am EDT | 5pm CEST 📍On Zoom Join us for a deep dive into how AI can help us understand human preferences — and what this means for economics and social sciences. 🔗 https://t.co/pMI5vJtth5 #EconTwitter

🎯 Why it matters? If you’re evaluating #LLMs on pre-cutoff data your #forecasting may be a polished replay of past truths 🔍 This paper offers the tools to audit and fix that 🔗 https://t.co/pMI5vJsVrx #EconTwitter #AI #DataScience #Econometrics #NLP #GPT4 #AMLEDS #Webinar 7/

✅ Solution: Convert firm-level info into abstract economic logic Anonymize identifiers = Forecasts with lower recall risk + solid directional accuracy (≈ 51–58%) 💡 Simple long–short strategy based on this yields a Sharpe ratio ≈ 2.09 #EconTwitter #AI #LLM #AMLEDS 6/

🧠 Embeddings leak signal 🎭 Motivated reasoning adds polish to memory 💥 Recent & prominent firms show strongest recall 📉 Small caps = higher refusal/error rates It’s not just memorization—it’s selective memorization. #EconTwitter #AI #LLM #AMLEDS 5/

📊 Pre-cutoff performance = eerily accurate 📉 Post-cutoff = collapse to noise ⚠️ Even “fake” cutoffs in prompts don’t work - #LLMs still #recall with high precision. And masking? Not enough. The model can deanonymize firms with 100% accuracy in some cases. #EconTwitter #AI 4/