Tom Vladeck, official brother, wrote that “opinions suck” on his website. Excerpt:

they suck for two reasons:

(1) they are subjective, and impossible to verify or falsify. how do you make any decisions based on these inputs?…

(2) most of the time “differences of opinion” are actually different projections of how the facts actually are. this shit wrankles me to no end. consider: “i think it will rain tomorrow” “i don’t think it will rain tomorrow”. there is a big difference between imperfect information and subjectivity. most differences of opinion are not in fact disagreements over subjective matters, but different projections over (unknown) objective facts. in these cases “let’s agree to disagree” (or any of its variants) is a shitty response. the answer is to get more data or to track which imperfect model of the world performs better in explaining it. 

The whole thing is worth reading. But I disagree. Here’s what I wrote in response:

I think you’ve misinterpreted “differences of opinion,” and your “I think it will rain tomorrow” example is indicative: it’s a strawman “difference of opinion” that doesn’t really tell us anything about opinions.

When two people disagree about whether it’s going to rain tomorrow, they’re using different heuristics to assess tomorrow’s weather based on easily available information. (E.g., the season, the cloud cover, whether it rained today, what the weatherman said, etc.) In this instance, where the cost of information gathering is extremely high and the payoff is low, saying “let’s agree to disagree” is just a polite way of saying “you’re an idiot but it’s not worth the time to prove I’m right.” Heuristics like these – even when people’s use of them differ – are important, because they get us past what can be crippling paralysis in the face of huge data.

However, the “weather” difference is a much less interesting and instructive difference of opinion than, say, “I think we should raise taxes on the wealthy.” Like the weather disagreement, there is certainly an empirical element: what the Laffer curve actually looks like w/r/t increasing marginal tax rates, the effects of capital flows to other jurisdictions, and so forth. But the key difference is that, even after accounting for the empirical questions, normative differences of opinion will undoubtedly remain.

This is where opinions get useful: they address the toughest and most interesting questions we deal with, precisely because they cannot be reduced to net present value. How much should society’s wealthy citizens pay to improve the lives of other less wealthy citizens, many of whom may have made poor decisions? In my opinion, a significant amount – but my argument is based primarily on principles, not statistics.

There’s more in the comment I left on his website