PolicyEngine uses artificial intelligence and machine learning to make policy analysis more accurate and accessible.
PolicyEngine combines tax-benefit microsimulation with cutting-edge AI to democratize policy understanding
Generate comprehensive policy analysis with natural language, tailored to different knowledge levels
Understand complex tax and benefit calculations through clear, accessible explanations of your household's finances
Combine computational precision with narrative insights to tell the complete story of policy impacts
PolicyEngine has leveraged artificial intelligence since our inception. In 2021, we pioneered the use of machine learning to enhance our microsimulation models, applying gradient descent to optimized survey weights and match administrative totals with unprecedented accuracy.
By 2022, our UK model had achieved up to 80% lower aggregate errors compared to other microsimulation models. This foundation of AI-enhanced accuracy has been central to our mission of providing reliable policy analysis.
We expanded our AI capabilities in 2023 with our Enhanced Current Population Survey (ECPS) for the US model, using quantile regression forests to integrate tax record information with survey data, creating the first open alternative to restricted tax microdata for policy microsimulation.
When OpenAI released GPT-4 in March 2023, we immediately recognized its potential to democratize policy understanding. Within just one month, we launched our AI-powered Analysis tool that translates complex computational results into accessible narratives.
In 2024, we extended this capability to household-level calculations with Anthropic's Claude API, enabling users to understand exactly how their taxes and benefits are calculated through plain-language explanations.
PolicyEngine integrates large language models with our computational tax-benefit engine to transform complex calculations into clear explanations.
For household calculations, we process thousands of intermediate values across tax and benefit programs, then use Anthropic's Claude API to generate plain-language explanations of eligibility, amounts, and potential changes.
For policy analysis, we leverage GPT-4 to weave narratives from our computational results, explaining reforms in terms anyone can understand - from simplified "ELI5" explanations to detailed technical analyses for policy experts.
Get the latests posts delivered right to your inbox.
© 2025 PolicyEngine. All rights reserved.