Exploring the oxymoronic concept of the "harmless wrong"
This is an interesting take. My issue with Haidt's example is that probabilistic concerns are legitimate. If someone made a baby play one round of Russian roulette and the baby lived and was amused by the clicking noise of an empty chamber, there is no harm, but there could've been. Similarly, there is likely to be harm with Julie and Mark.
I don't think that all ethical concerns reduce down to harm. I think that bodily violations are immoral even if there is no harm. But defending this position takes a long time and it's seems on its face weird, but I think it's defensible.