[ Question ]

What are the open problems in Human Rationality?

by Raemon 10mo13th Jan 20191 min read62 comments


LessWrong has been around for 10+ years, CFAR's been at work for around 6, and I think there have been at least a few other groups or individuals working on what I think of as the "Human Rationality Project."

I'm interested, especially from people who have invested significant time in attempting to push the rationality project forward, what they consider the major open questions facing the field. (More details in this comment)

"What is the Rationality Project?"

I'd prefer to leave "Rationality Project" somewhat vague, but I'd roughly summarize it as "the study of how to have optimal beliefs and make optimal decisions while running on human wetware."

If you have your own sense of what this means or should mean, feel free to use that in your answer. But some bits of context for a few possible avenues you could interpret this through:

Early LessWrong focused a lot of cognitive biases and how to account for them, as well as Bayesian epistemology.

CFAR (to my knowledge, roughly) started from a similar vantage point and eventually started moving in the direction of "how to do you figure out what you actually want and bring yourself into 'internal alignment' when you want multiple things, and/or different parts of you want different things and are working at cross purposes. It also looked a lot into Double Crux, as a tool to help people disagree more productively.

CFAR and Leverage both ended up exploring introspection as a tool.

Forecasting as a field has matured a bit. We have the Good Judgment project.

Behavioral Economics has begun to develop as a field.

I recently read "How to Measure Anything", and was somewhat struck at how it tackled prediction, calibration and determining key uncertainties in a fairly rigorous, professionalized fashion. I could imagine an alternate history of LessWrong that had emphasized this more strongly.

With this vague constellation of organizations and research areas, gesturing at an overall field...

...what are the big open questions the field of Human Rationality needs to answer, in order to help people have more accurate beliefs and/or make better decisions?


New Answer
Ask Related Question
New Comment
Write here. Select text for formatting options.
We support LaTeX: Cmd-4 for inline, Cmd-M for block-level (Ctrl on Windows).
You can switch between rich text and markdown in your user settings.

16 Answers

I went through all my LW posts and gathered the ones that either presented or reminded me of some problem in human rationality.

1. As we become more rational, how do we translate/transfer our old values embodied in the less rational subsystems?

2. How to figure out one's comparative advantage?

3. Meta-ethics. It's hard to be rational if you don't know where your values are supposed to come from.

4. Normative ethics. How much weight to put on altruism? Population ethics. Hedonic vs preference utilitarianism. Moral circle. Etc. It's hard to be rational if you don't know what your values are.

5. Which mental subsystem has one's real values, or how to weigh them.

6. How to handle moral uncertainty? For example should we discount total utilitarianism because we would have made a deal to for total utilitarianism to give up control in this universe?

7. If we apply UDT to humans, what does it actually say in various real-life situations like voting or contributing to x-risk reduction?

8. Does Aumann Agreement apply to humans, and if so how?

9. Meta-philosophy. It's hard to be rational if one doesn't know how to solve philosophical problems related to rationality.

10. It's not clear how selfishness works in UDT, which might be a problem if that's the right decision theory for humans.

11. Bargaining, politics, building alliances, fair division, we still don't know how to apply game theory to a lot of messy real-world problems, especially those involving more than a few people.

12. Reality fluid vs. caring measure. Subjective anticipation. Anthropics in general.

13. What is the nature of rationality, and more generally normativity?

14. What is the right way to handle logical uncertainty, and how does that interact with decision theory, bargaining, and other problems?

Comparing the rate of problems opened vs problems closed, we have so far to go....

I've actually been thinking about this for a while, here's a very rough draft outline of what I've got:

1. Which questions are important?
a. How should we practice cause prioritization in effective altruism?
b. How should we think about long shots at very large effects? (Pascal's Mugging)
c. How much should we be focusing on the global level, vs. our own happiness and ability to lead a normal life?
d. How do we identify gaps in our knowledge that might be wrong and need further evaluation?
e. How do we identify unexamined areas of our lives or decisions we make automatically? Should we examine those areas and make those decisions less automatically?

2. How do we determine whether we are operating in the right paradigm?
a. What are paradigms? Are they useful to think about?
b. If we were using the wrong paradigm, how would we know? How could we change it?
c. How do we learn new paradigms well enough to judge them at all?

3. How do we determine what the possible hypotheses are?
a. Are we unreasonably bad at generating new hypotheses once we have one, due to confirmation bias? How do we solve this?
b. Are there surprising techniques that can help us with this problem?

4. Which of the possible hypotheses is true?
a. How do we make accurate predictions?
b. How do we calibrate our probabilities?

5. How do we balance our explicit reasoning vs. that of other people and society?
a. Inside vs. outside view?
b. How do we identify experts? How much should we trust them?
c. Does cultural evolution produce accurate beliefs? How willing should we be to break tradition?
d. How much should the replication crisis affect our trust in science?
e. How well does good judgment travel across domains?

6. How do we go from accurate beliefs to accurate aliefs and effective action?
a. Akrasia and procrastination
b. Do different parts of the brain have different agendas? How can they all get on the same page?

7. How do we create an internal environment conducive to getting these questions right?
a. Do strong emotions help or hinder rationality?
b. Do meditation and related practices help or hinder rationality?
c. Do psychedelic drugs help or hinder rationality?

8. How do we create a community conducive to getting these questions right?
a. Is having "a rationalist community" useful?
b. How do strong communities arise and maintain themselves?
c. Should a community be organically grown or carefully structured?
d. How do we balance conflicting desires for an accepting where everyone can bring their friends and have fun, vs. high-standards devotion to a serious mission?
e. How do we prevent a rationalist community from becoming insular / echo chambery / cultish?
f. ...without also admitting every homeopath who wants to convince us that "homeopathy is rational"?
g. How do we balance the need for a strong community hub with the need for strong communities on the rim?
h. Can these problems be solved by having many overlapping communities with slightly different standards?

9. How does this community maintain its existence in the face of outside pressure?

There seem some foundational questions to the 'Rationality project', and (reprising my role as querulous critic) are oddly neglected in the 5-10 year history of the rationalist community: conspicuously, I find the best insight into these questions comes from psychology academia.

Is rationality best thought of as a single construct?

It roughly makes sense to talk of 'intelligence' or 'physical fitness' because performance in sub-components positively correlate: although it is hard to say which of an elite ultramarathoner, Judoka, or shotputter is fittest, I can confidently say all of them are fitter than I, and I am fitter than someone who is bedbound.

Is the same true of rationality? If it were the case that performance on tests of (say) callibration, sunk cost fallacy, and anchoring were all independent, then this would suggest 'rationality' is a circle our natural language draws around a grab-bag of skills or practices. The term could therefore mislead us into thinking it is a unified skill which we can 'generally' improve, and our efforts are better addressed at a finer level of granularity.

I think this is plausibly the case (or at least closer to the truth). The main evidence I have in mind is Stanovich's CART, whereby tests on individual sub-components we'd mark as fairly 'pure rationality' (e.g. base-rate neglect, framing, overconfidence - other parts of the CART look very IQ-testy like syllogistic reasoning, on which more later) have only weak correlations with one another (e.g. 0.2 ish).

Is rationality a skill, or a trait?

Perhaps key is that rationality (general sense) is something you can get stronger at or 'level up' in. Yet there is a facially plausible story that rationality (especially so-called 'epistemic' rationality) is something more like IQ: essentially a trait where training can at best enhance performance on sub-components yet not transfer back to the broader construct. Briefly:

  • Overall measures of rationality (principally Stanovich's CART) correlate about 0.7 with IQ - not much worse than IQ test subtests correlate with one another or g.
  • Infamous challenges in transfer. People whose job relies on a particular 'rationality skill' (e.g. gamblers and calibration) show greater performance in this area but not, as I recall, transfer improvements to others. This improved performance is often not only isolated but also context dependent: people may learn to avoid a particular cognitive bias in their professional lives, but remain generally susceptible to it otherwise.
  • The general dearth of well-evidenced successes from training. (cf. the old TAM panel on this topic, where most were autumnal).
  • For superforecasters, the GJP sees it can get some boost from training, but (as I understand it) the majority of their performance is attributed to selection, grouping, and aggregation.

It wouldn't necessarily be 'game over' for the 'Rationality project' even if this turns out to be the true story. Even if it is the case that 'drilling vocab' doesn't really improve my g, I might value a larger vocabulary for its own sake. In a similar way, even if there's no transfer, some rationality skills might prove generally useful (and 'improvable') such that drilling them to be useful on their own terms.

The superforecasting point can be argued the other way: that training can still get modest increases in performance in a composite test of epistemic rationality from people already exhibiting elite performance. But it does seem crucial to get a general sense of how well (and how broadly) can training be expected to work: else embarking on a program to 'improve rationality' may end up as ill-starred as the 'brain-training' games/apps fad a few years ago.

One more, because one of my posts presented two open problems, and I only listed one of them above:

15. Our current theoretical foundations for rationality all assume a fully specified utility function (or the equivalent), or at least a probability distribution on utility functions (to express moral/value uncertainty). But to the extent that humans can be considered to have a utility function at all, it's may best be viewed as a partial function that returns "unknown" for most of the input domain. Our current decision theories can't handle this because they would end up trying to add "unknown" to a numerical value during expected utility computation. Forcing humans to come up with an utility function or even a probability distribution on utility functions in order to use decision theory seems highly unsafe so we need an alternative.

In my opinion, the Hamming problem of group rationality, and possibly the Hamming problem of rationality generally, is how to preserve epistemic rationality under the inherent political pressures existing in a group produces.

It is the Hamming problem because if it isn't solved, everything else, including all the progress made on individual rationality, is doomed to become utterly worthless. We are not designed to be rational, and this is most harmful in group contexts, where the elephants in our brains take the most control from the riders and we have the least idea of what goals we are actually working towards.

I do not currently have any good models on how to attack it. The one person I thought might be making some progress on it was Brent, but he's now been justly exiled, and I have the sense that his pre-exile intellectual output is now subject to a high degree of scrutiny. This is understandable, but since I think his explicit models were superior to any anyone else has publicly shared, it's a significant setback.

Since that exile happened, I've attempted to find prior art elsewhere to build on, but the best prospect so far (C. Fred Alford, Group Psychology and Political Theory) turned out to be Freudian garbage.

To me the biggest open problem is how to make existing wisdom more palatable to people who are drawn to the rationalist community. What I have in mind as an expression of this problem is the tension between the post/metarationalists and the, I don't know, hard core of rationalists: I don't think the two are in conflict; the former are trying to bring in things from outside the traditional sources historically liked by rationalists; the latter see themselves as defending rationality from being polluted by antirationalist stuff; and both are trying to make rationality better (the former via adding; the latter via protecting and refining). The result is conflict even if I think the missions are not in conflict, though, so it seems an open problem is figuring out how to address that conflict.

Group rationality is a big one. It wouldn't surprise me if rationalists are less good on average at co-ordinating than other group because rationalists tend to be more individualistic and have their own opinions of what needs to be done. As an example, how long did it take for us to produce a new LW forum despite half of the people here being programmers? And rationality still doesn't have its own version of CEA.

The problem of interfaces between cultures.

Humans live in different cultures. A simple version of this is in how cultures greet each other. The Italian double kiss, the ultra orthodox Jewish non touch, the hippie hug, the handshake of various cultures, the Japanese bow/nod, and many more. It's possible to gravely offend a different culture with the way you do introductions.

Now think about the same potential offence but for all conversation culture.

I have the open question of how to successfully interface with other cultures.

Did you ever see that early (some might say, premature) trailer for the anthropic horror game SOMA where Jarret was wandering around, woefully confused, trying desperately to figure out where his brain was located?

That's how humans are about their values.

I can't find my utility function.

It's supposed to be inside me, but I see other people whose utility functions are definitely outside of their body, subjected to hellish machinations of capricious tribal egregores and cultural traumas, and they're suffering a lot.

I think my utility function might be an epiphenomenon of my tribe (where is my tribe???), but I'm not sure. There are things you can do to a tribe that change its whole values, so this doesn't seem to draw a firm enough boundary.

My values seem to change from hour to hour. Sometimes the idea of a homogeneous superhappy hedonium blob seems condemnably ugly, other times, they seem fine and good and worthy of living. Sometimes I am filled with compassion for all things, and sometimes I'm just a normal human who draws lines between ingroup and outgroup and only cares about what happens on the inner side.

The only people I know who claim to have fixed utility functions appear to be mutilating themselves to get that way, and I pale at the thought of such scarification, but what is the alternative? Is endless mutation really a value more intrinsic than any other? Have we made some kind of ultimate submission to evolution that will eventually depose us completely in favour of whatever disloyal offspring fight up from us?

Where is my utility function?

How about: "What is rationality?" and "Will rationality actually help you if you're not trying to design an AI?"

Don't get me wrong. I really like LessWrong. I've been fairly involved in the Seattle rationality community. Yet, all the same, I can't help but think that actual rationality hasn't really helped me all that much in my everyday life. I can point to very few things where I've used a Rationality Technique to make a decision, and none of those decisions were especially high-impact.

In my life, rationality has been a hobby. If I weren't reading the sequences, I'd be arguing about geopolitics, or playing board games. So, to me, the most open question in rationality is, "Why should one bother? What special claim does rationality have over my time and attention that, say, Starcraft does not?"

For applied rationality, my 10% improvement problem: https://www.lesswrong.com/posts/Aq8QSD3wb2epxuzEC/the-10-improvement-problem

Basically, how do you notice small (10% or less) improvements in areas that are hard to quantify. This is important, because after reaping the low-hanging fruits, stacking those small improvements is how you get ahead.

Is there a way to integrate probability based forecasting into the daily life of the average person that's clearly beneficial for them?

I don't think we are yet at that point where I can clearly say, that we are there. I think we would need new software to do this well.

Dealing with uncertainty about our own logic. It's a circular sort of problem: any logic I use to deal with my potentially flawed reasoning is itself potentially flawed. It gets worse when you deal with potential rationalization.

One open problem:

The problem of communication across agents, and generally what I call "miscommunication".

I'm a natural systems scientist, who noticed some time ago that it takes organization in nature to use energy, and then takes development to create organization. I was a physicist, but physics seems limited to studying things that can be represented with numbers. Organization is really not so open to that. So I developed various pattern languages, starting with the seemingly universal pattern of "natural growth," as a process of organization development. The pattern begins with an accumulation of small things or changes expanding into big ones, and then to produce stable organization in the end the progression has to reverse, with accumulating smaller and smaller things or changes leading toward a state of completion. If you look close, really everything we do follows that pattern, of diverging then converging accumulations. That seems to mark "time" as an organizational, not a numeric process.

So... for science to apparently not to notice that seems to be a very strong hint as to what in the world is wrong with human thinking. It's not just Descartes. Somewhere in the evolution of human thought, seemingly well before science chose to define nature with numbers, we seem to have decided that it was the job of reason to define reality. Of course we then get it wrong every time, as reality is already defined and so defining it is just not our job. Nature defines reality, and we can only do our best to study it. So, I don't know is my questions will help anyone else yet, but they might. It does seem "less wrong" as a point of view though, rather than to err by assuming our best rules of prediction define the world we are trying to predict. That's contradictory.

There's one overwhelmingly big problem, which is solving epistemology.