Synthetic populations and digital twins for human decision-making
Build representative simulated populations and individual digital twins that reproduce human attitudes and behavior, then let organizations test products, messages, policies, interfaces, and other decisions against those simulated people before deploying them.
From Latent Space — Simulating Humanity: from Generative Agents to 8 Billion Digital Twins — Joon Sung Park, Simile AI at 18:22
Problem: Human studies and panels are costly, time-consuming, difficult to scale, and often cannot ask participants enough questions. Decision-makers also need to understand how interventions can shape future outcomes, not merely predict what will happen.
For: Organizations that make frequent decisions about people, including CPG companies, developers, designers, policymakers, marketers, public companies, enterprises, and organizations that historically use human research panels.
Examples
- Wealthfront used the system for product testing beyond asking people what they thought, including multimodal inputs and agents traversing Figma mockups or websites.
Behind this: 14 build steps · 4 tools and how each is used · how to validate demand · 3 more real examples · 12 things the video never answers.