You may have heard of the rationalists. What about irrationalists? Well, I am one! Here are some recent examples of how I may have been irrational — being less than “rational” when forming views. If you don’t see a bit of yourself in them, congrats, you’re probably much more rational than I am!
If you are (a bit) like me, I hope this post can at least get you thinking.
If you are a rationalist or super rational already, this post may not be very informative, but if you decide to keep on reading I would appreciate your advice!
When knowledge can hurt: “inside view” and base rate neglect
I was talking to someone about the prospect for growth in Sub-Saharan Africa.
“Sure they’re poor and will grow fast according to the Solow growth model, but accounting for bad governance — which I don’t see how it will improve (as it could be a viscous cycle) — I’m not positive about their growth in the next few decades.”
They asked me, “What do you think the growth rate of Kenya for the past 20 years was?”
Before you read on, think for 30 seconds how you’d answer this.
Here is what I did (don’t judge…).
I thought: Kibaki came to power in 2002 (after Moi), then Uhuru in 2013, none of them were particularly great presidents, but they’re probably not so bad that I’d guess 1%, so I’d say the growth rate was at most 3%.
It turns out their GDP per capita roughly tripled between 2000 and 2019, so the average annual growth rate was between 5 and 6 percent.
In retrospect, instead of immediately jumping to how I think the presidents were not great (Tetlock’s “inside view”), I should have started with the (“outside view” / “base rate”). I wasn’t supposed to look up data during this conversation (and shame on me for not knowing facts about Kenya’s growth), but perhaps doing the following with what I knew could have been better:
- US annual growth rate is ~2% (don’t know exact number but that seems to be the norm for the past few decades according to the growth class I took in 2011).
- If Kenya had great governance it would be much higher as it’s so much poorer.
- Bad governance brings it down. Will it be below 2% though? 2% feels low (Kenya’s governance isn’t a real disaster) so probably above.
- China was 8% a few years back, and double digit even earlier when it was poorer.
- Unsure if Kenya today is poorer or richer than China when it had ~10% growth. Let’s say similar.
- So if Kenya had the same quality of governance as China when it had 10%, Kenya should be ~10%. In reality it should be lower.
- (Turns out China had 10% growth as recently as 2010 when its GDP per capita was 4500 USD, higher than Kenya’s 1500 today, so I should project Kenya to be at >10% with similar quality of governance.)
Conclusion (based on slightly wrong assumptions): may be above or below 2 — probably above; quite likely below 10.
It’s unclear what I would have concluded then, e.g. whether I would have taken the midpoint of 2 and 10 (or if that’s even a wise strategy), but I would have considered many more factors (and “data-driven” ones) than merely “presidents haven’t been great since 2002”. As someone who works on “evidence-based development policy” (and recently read “Superforecasting”), I’m embarrassed that when confronted with this question, I did not start by utilizing the few numbers I had in my head (on GDP, growth rates etc.) but jumped right to my negative impressions of the presidents.
5-6% is higher than I expected (but perhaps not if I took the alternative approach). Perhaps this is the way to think about it: it should have been 10% if governance quality were better, but despite the government they’re growing at 5-6% (on average over the past two decades).
I probably could have done this for any Sub-Saharan African country, using 2% and 10% as benchmarks and adjusting in between based on 1) baseline poverty level, 2) governance quality (both of which I only have an “anecdotal” and very rough sense of, e.g. super poor vs. poor vs. no idea, horrible vs. mediocre vs. okay governance vs. no idea). In the end, the data (on 2018 growth rates so different from above) would have still blown my mind, because there is much heterogeneity across countries, sometimes wild year-to-year fluctuations within countries (e.g. Zimbabwe), and some surprising rankings (e.g. I would have never expected Zambia and Malawi to be lower than DRC! and so much lower than Kenya).
It’s a hard exercise and I probably would have been pretty wrong anyway (esp. imagine if they happen to pick another country), I do not have a good way of getting the right answer, but now I see starting by thinking about how (good / bad) of an impression I have of the president is not a good approach. Even if I do okay in this case by luck, this strategy would probably hurt me if used over and over again (maybe like some of Tetlock’s experts). When you think you know something, you need to be careful to not put too much weight on it because it’s likely that what you (think you) know is a smaller part of reality than you think. (This doesn’t mean I should stop reading about African presidents, but I shouldn’t overweigh this piece when making the overall prediction.)
(This may not be the best example on base rates despite being a real life one from me. If you’re not familiar with “Superforecasting” this podcast may be interesting. Also, all views are my own. Some may not like my views about Kenya’s economy or its presidents, but I have to be honest about what I actually thought if I were to discuss how my thought process was irrational. My pessimism about growth in Kenya was unwarranted, not only because I underestimated its recent growth trend, but also even if growth in GDP per capita were only 2% per year, after 50 years it will more than double — the power of exponential growth. A separate question is how much of that gain falls to the extreme poor, but since China and India lifted so many out of poverty via growth, you’d have to assume growth in SSA is much, much less inclusive for it to not help much of the extreme poor — though I haven’t looked much into this.)
Anecdotes vs. data
I was discussing with someone the value of on-the-ground experience in (development / developing countries) vs. data (by which I mean rigorous evidence: representative sampling, good quality etc.) in forming views on things in developing countries. They said, “On-the-ground experience is largely anecdotal evidence. Unless it’s some basic common sense, I ultimately put little weight on anecdotes compared to (good quality) data.” (Another example they gave: “You can’t say you understand Trump supporters because you read ‘Hillbilly Elegy’ — that’s just one guy. I would want survey data to understand why people vote for Trump.” — Not saying the author supports Trumps but he is from the same demographic group as many Trump supporters.)
I feel they have a point. I do think I learned a lot about how things work in the developing world by spending time there, but now I wonder if I overweighed anecdotal evidence. And I definitely made sweeping generalizations from what I saw in one place in Kenya to all of Kenya and probably all of Sub-Saharan Africa, if not all of the developing world. Such generalizations may be safe at times — I’ve taken matatus plenty times in Kenya, never took a tro tro in Ghana or danfo in Nigeria but they probably work in similar ways. But it could be dangerous other times, probably when I don’t even know it.
I started wondering:
- When I update on a piece of evidence (“anecdotal” or “rigorous”), can I notice it, and ask myself if I’m overweighing anecdotal evidence, or generally overweighing the evidence relative to (my best assessment of) its quality?
- When I make a generalization, do I have evidence to back it up? Even if not, (do I have reasons to expect it to hold / what kind of conditions would make it hold or fail, and do I have any evidence on such conditions)? (Like a “crux”)
- Across Sub-Saharan African countries, what things have the highest levels of heterogeneity? (Which of those would surprise me the most / what’s my prior on them now)?
- (Digression: In the case of SSA, I’m wondering if there is more cultural similarities across countries than say Sweden and Italy, or we are more willing to generalize simply because most of us from outside of the continent are ignorant of the differences. Despite being aware of and often anecdotally speaking of the differences between the “stereotypical” Kenyan vs. Zambian vs. Nigerian etc., we probably still generalize too much.)
- When I hear a statement about (Kenya / Africa) (or even China / US / Canada) from someone who has experience there, what kind of things (surprise me / I disagree with)? Why would I and that person have different views on this thing? Whose view do I think is closer to the truth?
I do not have good answers now, but I hope to remember to ask myself these.
(Africa in the example with which I started this section, but it could apply to anything, as I probably overweigh anecdotal evidence in all realms!)
(For context: I’m a development economist and have cumulatively spent between 6 months and a year in Kenya for field work, and couple weeks in each of Uganda, Ghana, Zambia, and Myanmar for field work.)
Views and where they come from
When I was younger I once felt shocked by how most people had opinions on many more things than I did (this may have been after I moved to Canada or the US). I did not have a view on most things, I thought, because I didn’t have enough information. I felt stupid, that there must be something wrong with me, and that I should try to have more knowledge of the world (party in order to have opinions on most things and look smarter).
Now I’m older. I have more knowledge of the world, and also more views on things, but I’m afraid the latter is disproportional to the former — that many of my views aren’t informed by facts I know, but by something else. Some by anecdotes as mentioned above, which could be biased; scarier yet, some could be influenced by merely opinions of others — whether peers or “thought leaders” of my “tribe” (whatever that might be).
It’s best to form belief based on (good quality) evidence, but researching everything is too hard. Why do people like me let others opinions influence their own? This has a few advantages:
- Quick and easy.
- Reliable on many things, e.g. very common sense things (we need to eat food), some issues of morality (if most people feel strongly against X, there may be a good reason).
- Decreases tension between me and (peers / “tribe”) which would be costly.
However, it has some disadvantages:
- Others’ opinions aren’t always close to truth, e.g. if they were all from the same biased source rather than independently formed, if they were motivated by decreasing dissonance with others (or other “political” reasons) rather than seeking truth.
- The pressure to conform makes me examine my views less, find it more costly to converse with those (holding differing views / from other “tribes”) and hence miss out on some opportunities to get closer to truth, etc.
- Apart from potentially causing me to skew my information source, it may also lead me to selectively interpret evidence (e.g. confirmation bias).
I can’t exactly tell when the good or bad things are happening, but examining my views more is a good start.
Unfortunately over the years I have not only given in to the pressure to “have view on things” (in order to not look stupid), and become more “tribal”. (Like everyone else I have multiple “tribal identities”, including e.g. the Effective Altruism community — one of the most evidence-based “tribes” which encourages critical thinking, but one should still beware following the group rather than thinking for oneself!)
Instead of saying my views I would have loved to ask the following: what are important factors that would sway my view one way or the other on this issue, and what do I know about them? I may realize I don’t know much and hence can’t possibly be confident in any view. (This is like finding “cruxes” with myself.)
A middle approach is to say: “At the moment (I’m inclined towards / my preliminary view or gut feeling is towards) position X but I am not very confident. Let me examine what would sway my view: factors A, B, C. What do we know about them?” And so on. And be willing to change my mind if I find evidence on (A/B/C) do not support my view, and downgrade my confidence if I realize I know less about (A/B/C) than I thought (though I could also upgrade my confidence under the opposite finding, that’s probably rarer).
Robin Hanson is a proponent of not having to have a view on everything. I don’t always agree with him, but I like this thing he said on the Sam Harris podcast (my paraphrasing):
Try to live your live so that you don’t have to rely too much on things that may not be true. One way is to not have so many opinion on things. We’re in a society where you’re expected to have an opinion on half the things you hear. Just don’t do that. Just be agnostic about things you haven’t looked into. Pick your specialty, learn a few things and know them well, and find people you roughly trust on the other things. I don’t have an opinion on immigration in Europe. I don’t want to have an opinion on everything. I only want to express opinions on things I’ve studied and can tell you something about.
I would have expected an academic like him who is much more public-facing than most to (have / need to have / feel the pressure to have) views on many things, but he doesn’t. I feel encouraged by that. At least for consequentialist reasons, it seems better to spend the effort examining evidence in order to form beliefs closer to truth on a few things that matter the most to you, compared to giving in to the pressure to look smart, having views on lots of things and risking being wrong on many (including the ones that really matter).
I would like to be better at forming rational beliefs, not only because it seems good in itself and useful for everyday life, but because I am trying to have a lot of impact in my career. If I get to a position with high potential impact, I don’t want to have (on average) pretty wrong beliefs and subsequently a bunch of negative impact!
What to do? I’m not really sure but I would start by reading “The Sequences”. Any other (advice / tips / help practicing (e.g. doing “double crux” with me and help me realize how wrong I am)) all very welcome! If anyone feels they’ve made good progress on this, tell me about your journey and tricks!
Let me end by summarizing things I want to do going forward (mentioned above):
- When making a guess, start with the base rate not “inside view”
- When updating on a piece of evidence, consider if I’m weighing it proportional to its quality
- When making a generalization, consider whether the necessary conditions hold
- Examine my views — how much informed by evidence vs. influenced by peers
- Don’t be so quick to state my view; finding cruxes is more important; be open to changing my mind in light of cruxes (evidence / arguments)
- When someone expresses a different view, try to understand why
Doing this all the time is too expensive so it’s definitely good to adopt Robin Hanson’s strategy of being okay with not having a view on most things unless they really matter (for our work, life, voting etc.).