1/4. I see a lot of consequential decisions about behaviors being made based on data generated from (constructed on-the-fly, non-validated) attitude measures. I don't have the energy to do a full Tweetorial about the attitude-behavior relation right now, but there are at least

2/4. two things to remember if you're going to build entire behavioral models (especially those with life/death outcomes) around attitudinal data. First, the relationship between attitudes and behaviors is modest, and moderated by host of factors (Glasman & Albarracin, 2006).

3/4. Second, even if you are successful in changing attitudes or intentions, that does not guarantee behavior change: medium-to-large changes in intentions yield small to medium changes in behavior. That relationship is also highly moderated (Webb & Sheeran, 2006).

4/4. Just...keep that in mind before you start betting people's lives on non-validated single-item survey questions measured at one time point and context that is very different than another time point and context (purely hypothetical example of course).

That's all I ask.

Addendum: If you want to read more about attitudes, their relation to behavior and/or how to measure them and interpret those measurements, the syllabus for my PhD seminar on those topics is here; these readings are a place to start: https://osf.io/hzr6v/