[Previous] Discussion: Eating Candy | Home | [Next] Anti-Bias Procedures With Reach

Paths Forward or Prediction Markets?

In Can Foundational Physics Be Saved?, Robert Hanson proposes prediction markets to evaluate the future impact and value of scientists and research. This is intended to help address cognitive biases and incentive problems (like overhyping the value of one’s own research, and seeking short term popularity with peers, to get funding and jobs).

Markets strike me as too much of a popularity contest where outlier ideas will have low prices. I don’t think letting people bet on things will do a good job of figuring out which are the few positive outliers out of the many mostly-bad outliers. Designing good, objective ways to resolve the bets and pay out the winners will also be very difficult. And historians are often mistaken (in many ways, even more so than the news, which often gets the facts wrong about what happened yesterday), so judging by what future historians will think of today’s scientists is not ideal and can differ from what’s actually true.

So, in line with Paths Forward (including the additional information linked at the bottom like Using Intellectual Processes to Combat Bias), I have a different idea about how to improve science. It is online discussion forums and a culture of answering criticism.

Scientists and research projects should explain what they are doing and why it makes sense, in writing, and anyone in the public should be able to post criticism. Basic standards for discussion tools are listed in footnote [1].

Most readers are reacting by thinking discussions will be low quality and ineffective. There are many cultural norms, discussed in the linked essays and in books, which can improve discussion quality and rationality. But that’s not enough. People can read about how to have a truth-seeking discussion and then still fail badly. There already exist many low quality online discussions. Why, then, will the ubiquitous use of discussion venues help science?

Because of the expectation of answering every criticism received.

This often won’t be done. What’s the enforcement? First we’ll consider the vast majority of cases where a researcher or research project doesn’t get much attention. A few forums will get too many posts to answer, and we’ll address that later. But suppose some obscure scientist receives one criticism on his forum and ignores it. Now what?

Today, if I find a mistake by a scientist, I can write a blog post explaining the issue and arguing my point. Then I can tweet it out, share it on popular discussion forums, and hopefully draw some attention to it. What will people say? Many will try to debate. They will agree or disagree with my criticism. The Paths Forward approach will transform this situation into a different situation:

I write my criticism and post or link it at the discussion forum for the scientist or research project. They ignore me. Then I say to people: “I wrote X criticism and the relevant scientists did not respond.” And no one then debates with me whether my criticism is correct or not. That doesn’t matter. Everyone can clearly see the scientist has violated truth-seeking norms whether my criticism is correct or not. “He did not answer X argument…” is much more objective and clear-cut than “He is wrong because of X argument…”

Norms of having open discussions, where criticism is expected to be answered, would improve the current situation where hardly any criticism is written or answered, and little discussion takes place. And methodological criticisms – that someone did not respond to a criticism – are much easier to evaluate than scientific criticisms.

What if a scientist gives a low quality answer instead of a non-answer? This gives a critic more to work with. He can write a followup criticism. If he does a good job, then it will get progressively harder for the scientist giving a succession of bad answers to avoid saying anything that is wrong and easy for many people to evaluate. It’s hard to keep responding to criticism, including followups, and do it badly, but avoid anything that would noticeably look bad.

And this leads into the other main issue: What if scientists get too many criticisms to address and trying to keep up with them consumes lots of time? This would be an issue for popular scientists, and it could also be an issue when an obscure project gets even just one highly persistent critic. Someone could write dozens of followup criticisms that don’t make much sense. Methods for dealing with these issues are explained in the Paths Forward articles linked earlier, and I’ll go over some main points:

There’s no need to repeat yourself. The more your response to a criticism addresses general principles, the more you can re-use it in response to future inquiries. If people bring up points repetitively, link existing answers (including answers written by other people, which you are willing to take responsibility for just as if you wrote it yourself).

If there is a pattern of error in the criticisms, respond to that pattern itself instead of to each point individually.

If you get a bunch of unique criticisms you’ve never addressed before, you should be happy, even if you suspect the quality isn’t great. You can’t know if they are true without considering what the answers to those criticisms are. It’s a good use of your time to think through new and different criticisms which don’t fall into any pattern you’re already familiar with. That is a thing you can’t have too much of, and which is hard for critics to provide. The world is not full of too many novel criticisms. The vast majority of criticisms are boring because they fit into known patterns, like fallacies, and pointing that out and linking to a text addressing the issue is cheap and easy (and if people did that regularly, it would help spread knowledge of those common fallacies and other patterns of intellectual error, to the point that eventually people would stop making those errors so much).

It’s important, with suspected bad ideas, to either address them individually or address them by connecting them to some kind of general pattern which is addressed (sometimes we criticize types or categories of ideas, e.g. there are criticisms of all ad hominem arguments as a group). Ignoring a suspected bad idea with no answer - no ability to actually say what’s bad about it – is irrational and allows for bias and ignoring important, good criticisms. There is no way to know which criticisms are correct or high quality other than answering them. Circumstantial evidence, like whether the first words of sentences are capitalized, whether it uses slang, or whether the author has over 10,000 fans, are bad ways to judge ideas. Ideas should be judged by their content, not their source.

If you get tons of attention, you ought to be able to get some of your many admiring fans to help you out by acting as your proxies and answering common criticisms for you (primarily by handing out standard links). You can give an issue personal attention when your proxies don’t know the answer. You can also hire proxies if you’re popular/important enough to have money. Getting lots of interest in your work, and having resources to deal with it, generally happen proportionally. Using proxies to speak for you is fine as long as you take responsibility for what they do – if someone does a bad job, either address the issue yourself or fire him, but don’t just let it continue and then claim to be answering criticism through your proxies.

I’m not attempting to present an exact set of rules for people to follow, nor an exact set of instructions for what people should do. Science is a creative process. It requires flexibility and individuality. These are rough guidelines which would improve the situation, not exact steps to make it perfect. These proposals would increase the quantity of discussion and offer some improved ways for interested parties to engage. It offers mechanisms for identifying and correcting errors which offer better clarity and transparency when they are violated. I think the same proposal would improve every intellectual field – philosophy, psychology, history, economics – not just science.


[1] Forums should:

  • allow public access
  • have permalinks for every comment which are expected to still work in 20 years
  • don’t moderate or delete content for being disagreeable (only delete things like doxing, shock porn, and spam bots advertising viagra, not mere flaming, ad hominems, rudeness, or profanity)
  • no restrictive length limits. should be like 100k words, not 10k or 280 characters
  • no time limits after which additional comments are disabled
  • allow links to external sites
  • support nested blockquotes.

These simple standards are egregiously violated by currently popular forums like Twitter, Facebook and Reddit. The violations are intentional, not a technical issue.

Note that people don’t need to run their own forums. Each project can add a new sub-forum on a site that hosts many forums. Technologically, creating thousands of mostly-silent forums can be very cheap and easy. And there can easily be tools to monitor many forums at once and be notified of new posts. This technology pretty much already exists.


Elliot Temple on December 17, 2018

Comments (4)

Sadly, I just posted the following on Robin Hanson's blog post (the one linked at the start, which I wrote this in reply to, and where I had posted the full text of my comments):

> Did you delete my comment proposing a different way to improve science, and also ban me from commenting via my Twitter account that I'd been signed in with? If this is a glitch, automated filter, or misunderstanding, let's fix it. If it's intentional, how is that rational? (BTW I don't appreciate the lack of communication about what the issue is – my comment just disappeared without explanation and clicking the log in with Twitter button results in the message "There was an error submitting the form.") Here is a copy of what I wrote, for reference: http://curi.us/2161-paths-forward-or-prediction-markets. Agree or disagree, it's a on-topic good-faith effort to contribute to the discussion.


curi at 2:19 PM on December 17, 2018 | #11471 | reply | quote

Prof.

Consider a hiring committee with several candidates. Concretely, how does your proposal influence which candidate they favor? What exactly are they supposed to do differently?


Robin Hanson at 7:01 AM on December 18, 2018 | #11472 | reply | quote

> Consider a hiring committee with several candidates. Concretely, how does your proposal influence which candidate they favor? What exactly are they supposed to do differently?

It helps indirectly by making the reputations of people and projects more related to their quality and truth. This reputational information can reach hiring committees by the usual means.

And the hiring committee can look at the comments on someone's work to get information. They could see obscurity, unanswered criticism, well-answered criticism, or a difficult to evaluate discussion.

In the last case, I would advise the researcher to write a discussion summary, with plenty of links and quotes, to make it easier to evaluate. (The summary should itself be subject to critical comments. This doesn't spiral into infinite, unclear discussion because the discussion of the summary is easier to evaluate. The good criticisms of the summary would not be attempts to re-litigate the original issues, but would instead address the simpler matter of whether the summary is accurate. They'd say things like "Joe said X, here is a quote and link. I thought that was an important point, but you didn't respond and then left that out of your summary." or "Your paraphrase about Y is inaccurate, here are the quotes and then I point out 3 clear differences...").

That's the end of my direct answer to your question.

## Markets and Critical Discussions

Critical discussion doesn't lend itself to summary numbers. I think that's OK and that my proposal would encourage critical discussion and make it easier to evaluate compared to today. *You could have both Paths Forward and prediction markets*, and regard the additional critical discussion as a source of information for the markets to use. *But* there is a way the proposals are in conflict:

Prediction markets can give something a high score/price even though there are published refutations of it which are being ignored instead of answered. And they can give a low score to something which has answered all outstanding criticism. You can say "I have sought out criticism and meticulously answered all of it, publicly. Here are summaries and full records. If anyone knows any mistake I've made, please tell me." And then no one tells you another mistake, but you still have a low score on the market. This kind of thing is common today, except with a low score in terms of fame, reputation, social status, etc.

Markets don't have direct mechanisms to make them relate to the current state of critical discussion/debate anymore than current social interaction does. Market prices could be proxies for lots of the same social status issues (including which credentialing authorities approve of something) which determine hiring today.

Perhaps after there is a shift in cultural attitudes and people become more educated about how reason and also social status work, then the markets would be more effective. I expect prediction markets to reflect society's ideas, rather than solve the problem of improving those ideas. I think markets would help people do what they're already doing, but more efficiently, but I don't see them as addressing the core problems related to irrationality.

## Example 1

Here are two current examples of a good state of critical debate but a low reputation. I don't think that, today, a prediction market would give these high prices:

https://www.hedweb.com/manworld.htm

This document answers 41 objections to the Many Worlds Interpretation (MWI) in quantum physics. There are other documents with convincing arguments, IMO, which are unanswered, like _The Fabric of Reality_ by David Deutsch (chapter 2).

Most people reject MWI without saying which objection they have that isn't on the list of 41, or which of the 41 answers is false and why. Of course, it would help if the FAQ had a comment box at the bottom where people could say such things or could see that no one before them has posted any (unanswered) criticism (including a link to an essay or a reference to a book passage, there's no need to rewrite things which have been said before).

## Example 2

https://www.amazon.com/Philosophy-Popper-Library-Living-Philosophers/dp/0875481418/?tag=curi04-20

In this 2-volume book set, Karl Popper solicited 33 critical essays about his philosophical work and wrote detailed responses to them all. I have never found a Popper critic who builds on this by saying where Popper's answers to his critics went wrong, or by saying he has a new criticism which wasn't in those 33 essays. Instead, they consistently reject Popper due to reasoning and arguments which have already been addressed in print. I think Popper's reputation doesn't match the state of critical debate, and I've put a lot of effort into finding out whether I'm right about that, but it's hard to tell due to the lack of public discussion and open forums where people care about things like addressing all criticisms of their positions instead of ignoring some (and that ignoring, even if it were sometimes defensible, lacks procedures to prevent it from being done in a biased way, and lacks transparency for the public to evaluate and criticize it when it is biased). Basically no one thinks it's their responsibility to answer Popper – even if they would be wasting their careers if his criticisms of induction and alternative evolutionary epistemology are correct – and they all seem to think Popper was already answered somewhere by someone but they aren't specific about that (in my experience, people who reject Popper won't pick a specific document and say "That is my answer to Popper, and if it's mistaken then I'm mistaken", they don't provide any position for criticism, any way for their error to be corrected if they are in error).

FYI my proposal, which sees criticism and error correction as central to rationality, was developed by building on Popper's philosophy. It has some compatibility with other perspectives, but it comes out of the Popperian way of looking at things.


curi at 10:40 AM on December 18, 2018 | #11473 | reply | quote

No

#11471 No, I didn't delete your comment or ban you from commenting.


Robin Hanson at 5:10 AM on January 7, 2019 | #11563 | reply | quote

(This is an unmoderated discussion forum. Discussion info.)