Often experiments (controlled randomized trials) can't be done for ethical or practical reasons (E.g. finding out which TV ad is more successful). In these scenarios we need quasi-experimental methods like "difference in differences", "propensity score matching", and tools like directed acyclic graphs to display causal relationships and deduce the right calculations.

Of course there is material out there (https://causalinference.gitlab.io/book/) but it doesn't follow the khan academy didactics, so it's way harder to learn than necessary.

A course could not only help understanding scenarios where experiments are impossible but they'd also help to complement and help with understanding of statistical inference