Your first argument is implying that the ethics of an experiment is dependent on its sample size. That seems ridiculous.

Your second argument is just incorrect. Given that obesity is arguably the foremost health issue in the country, inducing appetite is not "at worst neutral."

The bakery example actually seems pretty comparable to FB. IMO they are both fine because the magnitude of the negative effect of being made hungry by the smell of a bakery and being made angry by seeing a FB post are both insignificantly small.