Papers of the day   All papers

The AI Economist: Improving Equality and Productivity with AI-Driven Tax Policies


Richard Socher: Excited to introduce the AI Economist: Extends ideas from Reinforcement Learning for tackling inequality through learned tax policy design. The framework optimizes productivity and equality. Blog: Paper: Q&A:

63 replies, 2202 likes

hardmaru: Using pure simulation and data-driven approach to the design of optimal tax policies. Both workers (agent) and the policy maker (environment) are RL agents optimizing for different things. Reminds me of Jay Forrester's work on System Dynamics (his ideas are used to make SimCity)

4 replies, 293 likes

Natasha Jaques: Can reinforcement learning in multi-agent gridworlds lead to discovering better tax policies? Some fun results, including zero-shot generalization to taxing humans playing the same game.

3 replies, 160 likes

Stephan Zheng: Very happy that the AI Economist is out: reinforcement learning finds tax policies that improve social welfare, even when RL agents game taxes. This is a hard two-level RL problem: taxes affect agent rewards, so agents and social planner need to constantly adapt. Summary to come!

4 replies, 75 likes

TechCrunch: Salesforce researchers are working on an AI economist for more equitable tax policy | by @ron_miller

3 replies, 59 likes

Wojtek Kopczuk: Back to the 1970s: what the world needs is an optimal tax framework with no intuition behind it, but one that instead performs well in simulations.

5 replies, 48 likes

Yu Bai: The AI Economist: Using multi-agent RL to simulate complex economic systems, guide policy designs, and improve social equality. Impressive work by colleagues @StephanZheng @alexrtrott and all at @SFResearch!

0 replies, 28 likes

Salesforce: Our team at @SFResearch, led by @RichardSocher launched the 𝑨𝑰 π‘¬π’„π’π’π’π’Žπ’Šπ’”π’•, an AI framework for economic policy design. The AI Economist is a new way that reinforcement learning can drive positive, social change.

0 replies, 22 likes

Jonathan "PhD in mask wearing" Hersh: Robots taking our jobs has gone too far! Okay, not really. Really fascinating stuff out of @salesforce research on using reinforcement learning to design optimal tax policy.

3 replies, 22 likes

Dr Simon Osindero: Lots of neat scope in this direction -- social-policy design using communities of adaptive agents. It'll be interesting to see what details are needed to make useful real-world predictions. Also important to understand where/how/for whom/etc such models fail. Very cool work!

0 replies, 21 likes

Nazneen Rajani: Amazing work done by my colleagues on the AI Economist that uses RL for optimizing equality and productivity. The simulations evolve through interactions between the AI government and several AI agents.

1 replies, 16 likes

Xander Steenbrugge: I'm a big fan of these approaches to complement existing regulation policies. Rather than trying to sell political ideas or bickering about personal opinions, these frameworks allow for objective metrics to steer decision-making. We got to get out of the marketing politics model!

0 replies, 15 likes

David Masad: RL + agent-based modeling continues to be one of the things I'm most excited about right now.

0 replies, 14 likes

Nikhil Naik: Excited about this work and it's potential for impact on RL and economics. We show that an RL-based economic planner can learn optimal tax policy for a stylized economy with participating RL agents and improve social welfare. (1/4)

1 replies, 14 likes

Salesforce Research: Introducing the AI Economist, an AI framework for economic policy design. The AI Economist is a new way that reinforcement learning can drive positive, social change. Learn more about this exciting innovation: Q&A: Blog:

1 replies, 14 likes

Pierre Richemond: Exciting that the RL framework allows for dynamic solutions - as opposed to most work in optimal tax theory !

1 replies, 13 likes

Alex Trott: Super excited to share our work and proud to be a part of this amazing team! With @StephanZheng Sunil Srinivasa @nikhil_ai @melvingruesbeck David Parkes @RichardSocher and tons of support from everyone at @SFResearch

0 replies, 13 likes

Thang Luong: Nice ideas and studies on AI and economy by Salesforce! The illustration video is cool, letting me wondering if in the future, doing AI research can be as easy and fun as playing video games (after which AI will generate a paper for you) :)

0 replies, 13 likes

Ronny Chieng: Can you plug A.I Economist into twitter to argue with #MAGA bots?

0 replies, 12 likes

ML and Data Projects To Know: πŸ“™ The AI Economist: Improving Equality and Productivity with AI-Driven Tax Policies Authors: @StephanZheng, @alexrtrott, Sunil Srinivasa, @nikhil_ai, @melvingruesbeck, David C. Parkes, @RichardSocher, @SFResearch Paper: Blog:

0 replies, 12 likes

ray: Awesome research from @SFResearch and @hseas on designing adaptive tax policies with deep reinforcement learning (using RLlib) from @RichardSocher and others! These policies can effectively trade off productivity and equality. Article by @ron_miller.

0 replies, 11 likes

Rob McCargow: Interesting innovation from @RichardSocher of @salesforce FYI @jhawksworth5 @BarryMurphyPwC @moulla @maria_axente @J__RA @nevinomics @ben_combes @StellaAmissPwC @GolbinIlana @andy_woodfield @amal_larhlid @EmmaHLCox @blairsheppard

1 replies, 10 likes

Anugraha Sinha: Reinforcement Learning simulating economic policies. May be this is relevant in today's COVID context as well, where nations are running towards opening the economies again. We could probably use such simulations to see what could happen if we do not open countries so soon?

0 replies, 6 likes

Cambridge Land Economy: Unortodox thoughtℒ️ : This is just Animal Crossing for economists (It’s also really interesting so check it out, I’m just finding the comparison funny) #ACNH #AnimalCrossing4economists #CambTweet

0 replies, 5 likes

MONTREAL.AI: The AI Economist: Improving Equality and Productivity with AI-Driven Tax Policies Zheng et al.: #GeneralEconomics #AIGovernance #MachineLearning

0 replies, 5 likes

Lorenn Ruster: Fascinating! #AI #equality #economist #RL #tax #policy

0 replies, 4 likes

Salesforce News: Introducing the AI Economist: a new AI-powered framework for economic policy design from @RichardSocher and @SFResearch. The tool aims to help economists, governments and others design tax policies that optimize social outcomes in the real world:

0 replies, 4 likes

Nathan Lane: The policies of an AI planner - whelp there goes my morning. This is an interesting application of reinforcement learning and optimal policy design context.

1 replies, 3 likes

paco nathan #Cascadia: Brilliant

0 replies, 2 likes

Richard Liaw: rllib for tax policy research? who wouldve thought...

0 replies, 2 likes

Daniel Mulet: Cc @professor_ajay

0 replies, 1 likes

Frank Dellaert: Reminiscent of Westworld III plot, but still very cool :-)

0 replies, 1 likes

Vignesh: I had read the paper couple of days ago and was blown away and didn't think it could get any better, but this animation is just damn!

0 replies, 1 likes

Lav Varshney: Ever since I read about economic simulation @sfiscience in Waldrop's book, Complexity as a kid, [B. Arthur, D. Farmer, et al.], I've wanted to see this kind of thing full on. Really glad my colleagues @SFResearch have pushed it to this level using RL!

0 replies, 1 likes

Matthew Pirkowski: One thing's for certain: As we begin to lean upon such systems, the tune sung by many economists with respect to the disruptive impact of automation will begin to change at an accelerating rate. It's one thing when it's someone else being displaced. It's another when it's you.

1 replies, 1 likes

ioana c.: This is so cool πŸ’œ.

0 replies, 1 likes


Found on Apr 29 2020 at

PDF content of a computer science paper: The AI Economist: Improving Equality and Productivity with AI-Driven Tax Policies