This is an interesting (if dense) read, arguing from an ontological perspective that psychological phenomena are "unsteady" and that if we view the replication crisis in that way it isn't really a crisis. Provocative, but I don't know that I can fully sign on...
"In Popperian falsificationism, falsifiability is the hallmark of science: Taking reproducibility as a defining of science was anathema to this methodology."
Burgos, J. E. (2024). Getting ontologically serious about the replication crisis in psychology. doi.org/10.1037/teo0...#PhilSci#MetaSci
I get the authors decision to emphasize ontology over epistemology here, and the need to treat both as separate. But epistemology comes roaring back in when we ask how we know the observed probability (e.g. statstical whatever) maps to the propensity of the psychological phenomena.
Put differently, phenomena existing implies the potential for non-existence and if some phenomena exists others must not. There seems little reason to believe that the probabilities attached to a given experiment isolate a single phenomenon.
Whatever we observe would be a mixture of phenomena that exist, and there is no reason even to presuppose that our experiment is limited to psychological phenomena. Phenomenon a could not exist, but trying to find it dredges up propensities for b,c,d, and e to varying degrees which do exist.
I guess if wanted to keep strictly to the ontology: A white crow in green light can alter the propensity associated with green crows even if green crows do not exist.
I think, practically this winds up in a similar same place as the author where white crows exist but they don't reliably look white (failed replications for true phenomena) or they can often look green when they're white (successful rep, phenomenon doesn't exist).
Which isn't to say we cannot infer something about the existence of green and white crows but we don't get it simply from the experiment and whether it replicates or not. We may have to establish the existence of various colors of light to make sense of our green and white crows.
It was entirely predictable from sample sizes alone. Exacerbated by QRPs. Made worse by bloody awful identification problems (if I'm using that word right, my field doesn't have that issue).
It's not a replication crisis, it's a we don't employ enough professionally qualified statisticians crisis.😱
