I'd say that (based on comments here) the real different is intent. People seem to be suggesting that the exact same experiment, but measuring ad-clicks instead of emotional state, would be a perfectly fine A/B test. It doesn't matter that the experiment manipulated people's emotions, it matters that Facebook wanted to manipulate people's emotions.