[thinkloop]

Those were shocking events. Scientific polls, liquid prediction markets and experts, all predicted different outcomes. It would have been irresponsible to take different positions.

Either the tools we have to measure these things are inadequate, and must be improved. People simply changed their minds as the events drew near. Or we happened to hit the underdog outcomes (80% chance of Hillary is still 1 in 5 for Trump).

None gets solved by polling trolls.

[wutbrodo]

> Either the tools we have to measure these things are inadequate, and must be improved.

I think that's the kind of point he's getting at. Dismissing people entirely as not relevant because their conversational norms differ from yours is the kind of things that leads to shocking events. This doesn't imply you should feel any more positively or negatively about these people, just that their comments may add explanatory power to your model of the world. When reality doesn't match your model, don't blame reality.

[zigzigzag]

And the polls, markets and experts were all wrong. Apparently they aren't so scientific, liquid or expert after all.

Actually, a few journalists who spent serious time hitting the streets outside London concluded that Brexit would win or at least was quite likely, just by talking to people. They were largely ignored because they weren't "experts", just people talking to other people. John Harris being an example of that.

[forgottenpass]

>Those were shocking events. Scientific polls, liquid prediction markets and experts, all predicted different outcomes. It would have been irresponsible to take different positions.

Those were not shocking events. People wanted a prediction with more confidence than those sources of prediction were able to provide. Faced with the hard limitations of existing predictive tools, people that should damn well know better started listening when they found someone to tell them "Nate Silver has Trump at 25%" really meant "Hillary will win."

It would not have been irresponsible to account for the possibility of the less likely but still possible outcome. In fact it would be irresponsible to NOT be prepared for it.

>Either the tools we have to measure these things are inadequate, and must be improved.

We're never going to be able to predict the future with total accuracy. The question is how we want to face them limitations of that accuracy. People who don't take "no" for an answer will read too much into unreliable predictions and attempt to rebuild systems for more predictable results regardless of externalities. The question should be how we adapt to that reality.

[Chris2048]

Scientific polls have participation and subjectivity flaws, 'expert' opinion begs the question.

No one addresses the obvious weaknesses, they just say "this is the best we have", but then act with confidence in the predictions, rather than admit we don't know.