[vintermann]

In the deontological view, both pulling the lever and inaction may be perfectly permissible.

Consider: Who knows what happens? It's perfectly possible the problem description lied, and the opposite thing happens. Either way real life doesn't come with problem descriptions. Since "what happens" is profoundly inaccessible, the more important question is what you wanted and why.

So, it's permissible to redirect the train away from one and towards many, if directing it away from the one was what you wanted. It's equally permissible to redirect the train away from the many and towards the one, if directing it away from the many was what you wanted. It's easy to imagine a world where you got what you wanted, but the bad thing that looked like would happen as a side effect didn't happen. Maybe that's the one we live in.

However, if you really wanted a specific person on the track to die, then you should pull it away from them. Not for their sake, but for yours. What happens is still uncertain, but the important thing is that you did not act on this bad desire.

(By the way, virtue ethics is just a stupid "third way" branding exercise. To say goodness isn't derived from outcomes is fine. To say it is, is at least a coherent position. But thinking you can dodge that problem by talking about "virtues" instead is just nonsense.)