Re:We don't know how to fix science (2021)
by openfuture on Wednesday July 27, 2022 @11:06AM
(#32251368)
Science is built on trust, no one reproduces every single finding and in the first place we measure the likelihood of some conclusion holding up in reality so we have probabilistic measures of trust in our conclusions, turns out that we can definitely build a society on this kind of trust. See my other responses to people responding to me (or datalisp.is which is a work in progress) to piece together what I am describing.
[ Parent
| on HN ]
Re:We don't know how to fix science (2021)
by naasking on Wednesday July 27, 2022 @05:52PM
(#32257514)
It's not based on trust, but verification by replication. The closest description is perhaps the security adage, "trust, but verify".
Re:We don't know how to fix science (2021)
by openfuture on Wednesday July 27, 2022 @07:00PM
(#32258191)
So you trust that a conspiracy is less likely than people just replicating and getting the same results (within tolerance).
If you are unsure you can convince yourself by replicating and then others can trust easier because you're probably not part of a conspiracy.
But it's all based on trust in the end. Again, no _one_ replicates everything.
Re:We don't know how to fix science (2021)
by naasking on Thursday July 28, 2022 @07:45AM
(#32263047)
I don't trust that a conspiracy is less likely, I infer that due to logical arguments of parsimony (see Solomonoff induction), and then I verify that such a conspiracy didn't take place to the best of my ability.