Hacker News Reader

Comment by monkeynotes | Hacker News Reader

1) People love drama 2) People say stuff they don't really mean just to perpetuate a drama in their otherwise predictable lives 3) Social media is a cancer on society 4) Statistics are sometimes nonsense

View on HN

Whether it's people trolling or not understanding the question you can end up with some nonsensical percentages when you poll the public at large. I don't know if this effect has a name but I remember my statistics professor talking about this.

The name of one of the factors is "selection bias". The type of people willing to take the survey may be those that already feel strongly on the topic, meaning you get results that favor one or both extremes, with fewer in the middle. Or the survey delivery mechanism makes response from some population segments less likely:

Results collected purely online may not get as maney older people less comfortable with technology (although that's a dwindling group).

Results collected through email solicitation may not get as many younger respondents who don't use email as much.

A phone survey sent out during the day will miss many people who aren't home, which will skew the results by excluding many people who are already comfortable going out, shopping, back to work, etc and move results towards "permanent lockdown".

Assuming there actually is a survey and people are actually being asked. 87% of the time it's people pulling numbers out of their ass.

If you want fake stats to support a specific viewpoint it's much less risk and very easy to simply phrase everything as a leading question directed towards certain answers. Like an uncentered Likert scale:

1) Perfect 2) Great 3) Very Good 4) Good 5) Okay 6) Neutral 7) Poor

5 out of 7 options are positive and only 1 is negative. As a bonus, the neutral & negative are listed last so you have to read through lesser positive options and the neutral before you get to the negative.

Lizardman's Constant: https://slatestarcodex.com/2013/04/12/noisy-poll-results-and...

There's another cute (if unofficial) name for this that is slightly more specific than just "selection bias" which is (likely) what your professor was talking about: it's called the "lizardman constant"

The name is originally based from the polling "results" of this article: https://www.publicpolicypolling.com/wp-content/uploads/2017/...

And (imo, aptly) named by Scott Alexander in https://slatestarcodex.com/2013/04/12/noisy-poll-results-and...

"3) Social media is a cancer on society"

Yessir.

Hacker News sometimes too.

Reply on news.ycombinator.com