Sometimes "looking at the thing itself" is too costly or too difficult. How can the proverbial "bright sixteen-year-old" sitting in a high school classroom figure out the truth about, say, the number of protons in an atom of gold, without having to accept the authority of his textbooks and instructors? If there were a bunch of well-funded nutcases dedicated to arguing that gold atoms have seventy-eight protons instead of seventy-nine, the only way you can really judge who's correct is to judge the relative credibility of the people presenting the evidence. After all, one side's evidence could be completely fraudulent and you'd have no way of knowing that.
Far too often, reference classes and meta-level discussions are all we have.
Then let us try to figure out whose authority is to be trusted about experimental results and work from there. Cases where you can reduce it to a direct conflict about easily observable facts, and then work from there, are much more likely to have one dramatically more trustworthy party.
I really liked Robin's point that mainstream scientists are usually right, while contrarians are usually wrong. We don't need to get into details of the dispute - and usually we cannot really make an informed judgment without spending too much time anyway - just figuring out who's "mainstream" lets us know who's right with high probability. It's type of thinking related to reference class forecasting - find a reference class of similar situations with known outcomes, and we get a pretty decent probability distribution over possible outcomes.
Unfortunately deciding what's the proper reference class is not straightforward, and can be a point of contention. If you put climate change scientists in the reference class of "mainstream science", it gives great credence to their findings. People who doubt them can be freely disbelieved, and any arguments can be dismissed by low success rate of contrarianism against mainstream science.
But, if you put climate change scientists in reference class of "highly politicized science", then the chance of them being completely wrong becomes orders of magnitude higher. We have plenty of examples where such science was completely wrong and persisted in being wrong in spite of overwhelming evidence, as with race and IQ, nuclear winter, and pretty much everything in macroeconomics. Chances of mainstream being right, and contrarians being right are not too dissimilar in such cases.
Or, if the reference class is "science-y Doomsday predictors", then they're almost certainly completely wrong. See Paul Ehrlich (overpopulation), and Matt Simmons (peak oil) for some examples, both treated extremely seriously by mainstream media at time. So far in spite of countless cases of science predicting doom and gloom, not a single one of them turned out to be true, usually not just barely enough to be discounted by anthropic principle, but spectacularly so. Cornucopians were virtually always right.
It's also possible to use multiple reference classes - to view impact on climate according to "highly politicized science" reference class, and impact on human well-being according to "science-y Doomsday predictors" reference class, what's more or less how I think about it.
I'm sure if you thought hard enough, you could come up with other plausible reference classes, each leading to any conclusion you desire. I don't see how one of these reference class reasonings is obviously more valid than others, nor do I see any clear criteria for choosing the right reference class. It seems as subjective as Bayesian priors, except we know in advance we won't have evidence necessary for our views to converge.
The problem doesn't arise only if you agree to reference classes in advance, as you can reasonably do with the original application of forecasting costs of public projects. Does it kill reference class forecasting as a general technique, or is there a way to save it?