We have exciting new work on optimally annotating outcomes for doubly-robust causal inference, motivated by an ongoing collaboration in homelessness services. It works really well! paper