Hacker News Reader

Comment by taneq | Hacker News Reader

I don't think it's necessarily that everyone in the whole psychology profession has worse than highschool level math skills. I think it's more just that, at some level, you have to trust that your colleagues have done their work correctly. You focus on the important parts of the paper (ie. the hypothesis, the method, and the results) because there's not enough time in your life to give full scrutiny to every detail of every paper you read. So sometimes, something slips through the cracks, until it's caught.

View on HN

At which level in the sciences are you ever supposed to trust that your colleagues have done their work correctly? That sounds more like the opposite of what you are supposed to be doing.

At the level where you cite the work of others.

I'm reminded of the Russian proverb; "Trust, but verify".

A citation does generally mean that you hold a given paper with a certain level of trust, but that isn't the same as having trusted the people who wrote it to have done their job properly.

The trust in the paper could be because you trust the folk who wrote it to have gone and done their job properly, however the trust could also be from having actively gone and checked the working.

Similarly, "Trust - but take care in whom."

At the level where it takes more than one lifetime's research to reinvent your field from ground zero, it's literally impossible to progress the field without taking something on faith.

Well before that, though, it becomes impractical to re-check every single thing. And the longer-established a field is, the more you have to take on faith.

So then you have to go to other people in that field and have them look at it. Sure, you have to take them on faith too, but a little bit less each time unless they all wildly disagree, and they can explain their reasoning to you in layman terms when you canvas their opinion. Now, none of this guarantees success, but it is a long way from trusting others to just do their work properly.

What do you recommend? Without a criterion for trusting some previous work this would make science impossible.

Reply on news.ycombinator.com