In my view, the most important ethical consideration in doing social sciences research on children is that researchers should do nothing that might harm the children (physically or psychologically). Of course, this is difficult to “operationalize” in that it is hard to know for sure when a research project might lead to psychological harm. However, this should be our guiding principle.
For example, while this was not exactly a research project per se, Jane Elliott’s decision to run an exercise in which she exposed children with the wrong eye color to simulated “racism” was a horrible decision. It is certainly interesting to see how children react when told that one group is worthy of being discriminated against. It is interesting to see how children are affected by “racism.” But hurting a group of children deeply (if temporarily) is not an ethical way to learn interesting things.
In more prosaic experiments, we have to be sure that we do not knowingly expose them to harm. For example, if we know for certain that a method of teaching is ineffective, we should not run randomized studies that assign some children to be taught in that way as a control group. At all times, our main guiding ethical principle should be the idea that we should do no harm to those whom we are studying.