These are the kinds of questions addressed in Feldman and Warfield's short anthology Disagreement.1 In this post, I will primarily be sticking to one chapter which hits the central points.
"Epistemic peer" is a term that pops up in most of the chapters. It's a shorthand way of talking about two or more people who have access to the relevant evidence and are mutually familiar with the different arguments they all hold. Thomas Kelly offers these three examples:2
- Two members of a jury, when it's time to decide 'guilty' or 'not-guilty.'
- Two weather forecasters on the chance of rain tomorrow.
- Two professional philosophers on whether free will is compatible with determinism.
Disagreement isn't such a problem when it happens between non peers on a particular topic. If I find out one of my beliefs about the moons of Jupiter is in conflict with what an astrophysicist believes, I should take this to indicate it's very likely that I'm ignorant of some relevant evidence and I should lower my confidence until I can investigate further. Meanwhile, the astrophysicist need not consider lowering her confidence just because I disagree unless I can demonstrate status as an epistemic peer.
I found Thomas Kelly's chapter "Peer Disagreement and Higher-Order Evidence" to have the most helpful terminology, at least as a starting point. Search link:
Peer Disagreement and Higher-Order Evidence — Thomas Kelly
Here, Kelly examines two very different responses to peer disagreement, then ends up combining them. The first extreme response:
"The Equal Weight View: In cases of peer disagreement, one should give equal weight to the opinion of a peer and to one's own opinion."A pro-argument. You and I are carrying thermometers which have agreed in the past. I check mine and conclude it's 60 degrees Fahrenheit, but then I find out you just checked yours and have concluded it is 65 degrees Fahrenheit. I know you're as fanatical about accurate measurement as I am, so it seems I must lower my confidence in "60 degrees" and raise my confidence in "65 degrees" until my confidence is evenly split between the two.3
A con-argument. Given the evidence available, the rationally correct level of confidence that hypothesis H is true is 10% (quite unlikely!). You and I come to two different, higher levels of confidence than that. Suppose I'm 20% confident that H is true and you are 80% confident. According to the Equal Weight View, we should both split the difference and be 50% confident that H is true. But it seems wrong that I should have to adjust my close-to-appropriate confidence so much in response to your way-off-base confidence.4
Perhaps there are better con-arguments against the Equal Weight View, but I found this one unpersuasive. In the scenario, you and I have already tried our best individually to form a correct level of confidence in hypothesis H. Expecting us both to recognize that my answer is more justified by the evidence and your answer is less justified seems to be going against the conditions of the hypothetical situation.
The second extreme response:
"The No Independent Weight View: In at least some cases of peer disagreement, it can be perfectly reasonable to give no weight at all to the opinion of the other party."and its sibling:
"The Symmetrical No Independent Weight View: In at least some cases of peer disagreement, both parties to the dispute might be perfectly reasonable even if neither gives any weight at all to the opinion of the other party."(Though, really, the other extreme would replace "at least some cases" with "all cases." In other words, a person could always respond to peer disagreement by not making any adjustment to their own confidence level. I could take the stance that my peer must show me evidence or arguments besides mere disagreement before I should be expected to budge.)
A pro-argument. Suppose I examine the evidence and become 80% confident hypothesis H is true, but I also take the position that confidence levels 70% through 90% are within reasonable range. You examine the evidence and become 75% confident H is true, and also hold that 65% through 85% are within reasonable range. We meet and find out we disagree, but we're both close enough that we consider the other's answer reasonable. It doesn't seem like we should be rationally required to split the difference.5
Another quick example: an Atheist peer and a Deist peer may disagree quite substantially but still consider the other view within the range of reasonableness, given the nature of the evidence. That last bit — still considering the quality of the original evidence when dealing with disagreement — is what Kelly thinks is missing from a strict adherence to the Equal Weight View.
I don't have a specific con-argument. However, those who find the Equal Weight View strongly persuasive would want to know when it makes sense to set the Equal Weight View aside. More on that shortly.
If one view sounds pretty good and so does the other, why not try combining them? That's what Kelly does with his Total Evidence View. He argues that we should take all of the following into account during peer disagreement:
- The original evidence (the first-order evidence).
- My own confidence level.
- Your confidence level (which, together with my own confidence level forms the higher-order evidence).
The obvious question now is: "How, exactly, do does the nature (or quality) of the original evidence weigh in?" Can we construct a neat formula which reduces to the Equal Weight View as a special case and the No Independent Weight View as another special case? Kelly doesn't try to do this. Instead, he gives the following rough outline:
"In some cases, the first-order evidence might be extremely substantial compared to the higher-order evidence; in such cases, the former tends to swamp the latter. In other cases, the first-order evidence might be quite insubstantial compared to the higher-order evidence; in such cases, the latter tends to swamp the former. [...] In still other cases, the two kinds of evidence might play a more or less equal role in fixing facts about what is reasonable to believe."6An example of the first-order 'swamping' the second-order. Suppose we go out to dinner, decide to split the bill, and we both calculate what half of the bill should be. I come up with an answer of $20 each. You come up with an answer of $130 each. It's very obvious that $130 is more than the entire bill. Even if I have known you to be equally reliable at math before today, I can completely disregard your answer because it's so absurd.7
Frankly, I'm not sold on the Total Evidence View. In this dinner math example, it's doubtful we would still disagree after talking it over briefly. Maybe that's unfair since it's such a trivial example, but I thought the main point of this "philosophy of disagreement" stuff was how to handle disagreements with seemingly reasonable people with whom we've hashed out an argument as far as it seems we can on the first-order evidence. I do tend to think the quality of the original evidence should be "baked into" our confidence levels before considering whether disagreement itself constitutes additional evidence.
I still don't have any very strong views about whether peer disagreement counts as evidence. It's a good question which I hope to keep exploring.
When I started reading this book, I wondered if one of the writers would turn the lens of disagreement theory back toward disagreements about disagreement theory. Yep! Adam Elga delivered with his chapter "How to Disagree about How to Disagree."
What I found most interesting was the suggestion that positions about how to handle disagreements might be self-undermining. If I hold a "split the difference" style position and you disagree, how can I consistently hold fast to my position about how to handle disagreements? Elga argues that, yes, this would be hypocritical...unless we can somehow make a principled exception for the principle itself.
A principle that makes an exception for itself? How scandalous! But consider the hypothetical case that Consumer Reports magazine not only rates consumer products but also rates magazines of its own kind. Wouldn't it have to give itself top marks all the time for its product ratings to make any sense? How could Consumer Reports give Super Waffle Toaster a 3/5 rating (or however it's done) without implying that it stands behind its 3/5 rating as the most accurate rating any such magazine could give of the product?
"In order to be consistent, a fundamental policy, rule or method must be dogmatic with respect to its own correctness. This general constraint provides independent motivation for a view on disagreement to treat disagreement about disagreement in a special way."8Interesting. Now I'm on the lookout for when this principle of principled exceptions might apply in other contexts.
1. Feldman, R., & Warfield, T. A. (Eds.). 2010. Disagreement. Oxford: Oxford University Press.
2. p 111.
3. Adapted from p 114.
4. Adapted from p 122.
5. Adapted from p 118.
6. p 142.
7. Adapted from p 149.
8. p 185.