Probabilistic modeling and techniques of reducing ambiguity about the future are by large forward propagating. Machine learning and AI is back propagating.
Whenever you train a computer program (model) from example (data) you’re propagating stats from the field back to the machine (and by extension its designer).
Forward propagation is a hypothesis (intuition) realized (ethnography), spatially mapped (video posted below too) and simulated (using techniques like ABM for example). The information (one level up from data) is then measured against the field again, and once validated could be allowed to yield knowledge (such as levels of change in a complex system), and maybe even wisdom (more on that taxonomy at the end of this post).
Now the punchline: forward propagating design is the best avenue (I can think of, at the moment) to design for the world inside your head (and hence incredibly valuable for (1) leaders, (2) who manage systems).