Essentially, you identify numerous hypothetical scenarios, analyse the impact they would have, and use that information to decide whether a particular course of action is a good idea or not.
This creates a faster feedback loop inside of the larger, noisier one. Your feedback is now guided around the question of "what makes the rubric itself better?" This can be done on principle, with limited access to external information. Philosophical thinking is eminently suited to this style of problem, but it can be supplemented with short-term empirical studies that add some falsifying points and narrow your cone of uncertainty.
At the end you've generated a list of yes/no questions forming the rubric of whether the course of action is likely to succeed. It can be turned into a ranking score, or a pass-fail threshold.
If you're frustrated by the idea of just making it up on principle, that's a frustration with philosophy itself; it rarely "works" until you accept some pragmatic premises around what is "good" or "true". The point of having a large number of questions, using a wide variety of perspectives, is that they test the overall coherency of the premise. Something can work fine from one perspective, and then completely fail in another. When that happens, it's a good sign that you have more to improve.
It's quite an important life skill to practice. It's easy to go along with the crowd, but this is a way of breaking away from it.
"Next, we eliminate some incorrect worlds by testing them against noisy data. This data should come from intermediate steps and can’t be quantified. If the data conflicts with an imagined future, that future is unlikely to occur. Noisy information may be bad for feedback loops, but it’s great for elimination, because the wider range of outcomes is more effective for pruning. It may even be a bit faster than accurate data, since the variance of accurate data is lower!
As we near decision time, we stop the world construction and elimination, and use judgment and experience to pick the best idea out of the remaining ones. Process improvement happens by creating many parallel processes and discarding the ones that are unfit, not by iteratively improving a single process."
I can't quite understand what he is saying.