1 + 2. Generalizability is not always a feature of random surveying, and I’d have to look at how exactly they did the survey taking and any analysis if any, it’s often not unreasonable. The wording they use “consumers lose interest” could definitely be changed, but it doesn’t seem a big issue. The real big issue is the loss of a question. We have no idea of the responder loading before versus after announcement and how it may have changed, but it is clear that of individuals who believe they possess enough information to make a decision, there are a great number who are uninterested.

3 + 4. What you said under 3 is true, but I believe 4 is linked to it, and 4 is wrong. People are more likely to pick an uncertain option that leaves room for a change in the decision if they don’t possess a lot of information than they would be to express a certainty. It’s much more likely that the positive responses are over-represented for that reason than the negative responses.