I'm a scientist, and I don't think things so easily become consensus. In order to get to consensus, a model needs to have been tested from many different angles. It's never enough to get to statistical significance, things either get replicated or, more often, triangulated.
Consensus is an extremely time-consuming thing to build, and it's extremely important to be aware of when there exists a consensus, where there isn't one, and what the consensus is.
It isn't an appeal to authority fallacy, it's a form of deferral to expertise, and it's one of the most important heuristics we have.
Absolutely not true. I’ve seen things become consensus because of one single questionable image in one paper from a “trusted” lab.
In my alma matter field the question was whether her2 receptor recycles (has implication of all breast cancer antibody therapies that target her2) and a paper from a Genentech lab had ONE FIGURE (where different data points from different experiments) as proof. An entire sub field including projects in my own lab were spawned assuming this.
Whenever I point out this flaw in lab meetings I’d be shut down by my professors as “they know the authors they trust them”.
I think I'm going to continue disagreeing with you on this point, but I will point out that our difference of opinion is a semantic one.
To say that everyone "believes in" a model could mean that everyone accepts it as plausible (and thus worthy of further exploration), or it could mean that everyone is justifiably certain that it maps properly to a real phenomenon.
I never say that something is "consensus" in the first case: IMO, the term should be reserved for the latter case, or appropriately qualified.
In any case, the situation you describe appears to lack meaningful triangulation. [1]
Consensus is an extremely time-consuming thing to build, and it's extremely important to be aware of when there exists a consensus, where there isn't one, and what the consensus is.
It isn't an appeal to authority fallacy, it's a form of deferral to expertise, and it's one of the most important heuristics we have.