Kahneman is half-right, though. You have to accept the conclusion of a study as true. You can't accept the study as true and then keep believing in the thing the study just disproved.
What you can do, though, is question whether the study itself is true. Sometimes, it's not.
One can consistently accept a study as well-implemented/operationalized while acknowledging a reasonable chance that the results are garbage, and that repeating the study might not yield significant results.The proper conclusion of a statistically oriented study is a probability distribution across several potentially contradictory propositions. One is never forced to "exclude the middle" as a result of a study, and conclude either A or not A.
Re-reading what you said I think we're in agreement already. I was nit-picking at the idea that a study has a single conclusion which is either true or not true. The conclusion of most studies is properly a vector and a hefty pinch of salt imo.
Actually I would say that you are more likely to be correct if you wait until there are multiple replicated results before you "have to accept" it as true. One study is interesting evidence, depending on the protocol, but publication bias/filedrawer effect is a real thing and studies are massively biased towards positive results.
>You have to accept the conclusion of a study as true. You can't accept the study as true and then keep believing in the thing the study just disproved.
People doing that weren't who Kahneman addressed there though. It was people thinking the study was BS/fishy/wrong.
What you can do, though, is question whether the study itself is true. Sometimes, it's not.