What is it about?
Experiments are informative but often costly. We build a “Theory Instead of Experiment” (TIE) system of digital twins that replace experiments by modeling how different participants respond to policy changes. In Tencent’s creator valuation task, TIE achieves a 97% correlation with real experimental results while avoiding these costs. TIE has since been deployed as Tencent’s default system for creator valuation.
Featured Image
Photo by Markus Spiske on Unsplash
Why is it important?
Experimentation is often costly—it may disrupt user experiences, demand substantial manual effort, and take time to yield results. TIE helps eliminate unnecessary experiments, saving resources for cases where experimentation is truly needed.
Perspectives
It’s not feasible to test every policy through experiments. Theory can provide sufficient insight in many cases, reserving experiments for when they are truly necessary.
Lei Huang
Massachusetts Institute of Technology
Read the Original
This page is a summary of: Theory Instead of Experiment (TIE): A Creator Valuation System at Tencent, August 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3711896.3737267.
You can read the full text:
Resources
Contributors
The following have contributed to this page
