If what you want to know is “what characteristics do LW participants and ex-participants have” then you want to survey LW participants and ex-participants, and responses from other people will not help to answer that question. And if their survey answers don’t clearly distinguish the other people from the participants and ex-participants, they will make the answers less useful. I forget how clearly the questions ought to make it possible to distinguish, but given that people make mistakes and don’t fill everything in I suspect that in practice they distinguish much less than perfectly.
If what you want to know is “what characteristics do LW participants and ex-participants have” then you want to survey LW participants and ex-participants, and responses from other people will not help to answer that question.
I don’t think that’s true. Having a control group quite often does help you to know more.
Having more data is good regardles of the semantics.
Not if the “data” is noise.
Most of those who haven’t ever been on Less Wrong will provide data for that distinction. It isn’t noise.
If what you want to know is “what characteristics do LW participants and ex-participants have” then you want to survey LW participants and ex-participants, and responses from other people will not help to answer that question. And if their survey answers don’t clearly distinguish the other people from the participants and ex-participants, they will make the answers less useful. I forget how clearly the questions ought to make it possible to distinguish, but given that people make mistakes and don’t fill everything in I suspect that in practice they distinguish much less than perfectly.
I don’t think that’s true. Having a control group quite often does help you to know more.