If the experiment was designed to be small enough to not affect the outcome of the election, then they're not experimenting with affecting the outcome, they're investigating and learning about tactics that, on a larger scale, could be used to affect the outcome of the election.

I did say it was a bit shady, but I'm not really sure how to do this sort of research in a way that doesn't potentially affect the real world, because the whole point is to see how this sort of thing affects real people. Doing it in simulation doesn't help because that only tests your simulation.

I'm taking it as a given that this kind of research is important to be able to identify and combat actual interference of this kind from malicious entities in the future.