I think this is moreso a misunderstanding - surveys on their own, in raw form, are not science

There’s all kinds of bs that can come up like:

• selection bias
• response bias
• general recollection errors/noise (especially for scary or traumatic experiences - there’s a bunch of papers on this behavior)

But data scientist can account for these by looking at things like sample selection (randomly selected so as to represent the nation/region/etc), pilot runs, transparency (fucking huge dude, tell everyone and anyone exactly what you did so we can help point out bullshit), and stuff like adjusting for non-responses.

Non responses are basically the idea that some people simply don’t give a fuck enough to do the survey. Think about a survey your Human Resources team at work might send out - people who fuckin hate working there and don’t see it changing anytime soon might not vote, which means there would be less people expressing their distaste which leads to a false narrative: that people like working there.

Hope this makes sense! Stay curious!!

PS/EDIT: Check out the SAGE method for data science for some more info! (There’s probably a YouTube vid instead of the book if you’d prefer I’m sure!)

Survey says, you need more data.