<i>"...in an environment punctuated by slow, progressive changes followed by cataclistic changes in the opposite direction, individuals that track the enviromment better will overfit (and die). More inaccurate individuals will be the ones surviving long term."</i><p>Reading this made me think of the numerous financial firms in history that become very efficient at making money in a particular type of market environment, which inevitably changes suddenly in unexpected ways, causing those financial firms to blow up and maybe even start a financial crisis:<p>* <a href="https://en.wikipedia.org/wiki/List_of_stock_market_crashes_and_bear_markets" rel="nofollow">https://en.wikipedia.org/wiki/List_of_stock_market_crashes_a...</a><p>* <a href="https://en.wikipedia.org/wiki/List_of_banking_crises" rel="nofollow">https://en.wikipedia.org/wiki/List_of_banking_crises</a><p>Society might be better off with financial firms that are "dumber!"
I don't think humans evolve to be inaccurate - I think a better description is that they evolve to not all be the same. If you look at two species, the one that is very uniform is more likely to die out over the long term than the one that has more variation. That variation can be physical, or in the mind and how it makes choices. But this variation makes the species more adaptable when conditions change drastically.<p>If you look at the long term, in the world that the article describes (slow, gradual with rare cataclysmic changes), most wrong decision makers will still die at a higher chance than correct decision makers. It's just in those rare situations that they survive.
Related to this topic: someone (Derbasti) once recommended me [1] "Thinking: Fast and Slow" by economy Nobel-winning Kahneman. This is a book about dozens of ways humans fail to reason properly and a huge part of this book addresses the problem of heuristics we use to estimate risk, probability, costs... absolutely a must-read.<p>[1]: <a href="https://news.ycombinator.com/item?id=11555148" rel="nofollow">https://news.ycombinator.com/item?id=11555148</a>
Well it would make sense if that were the case.<p>You have some system that tried to balance bias, variance, and utility. Sometimes that best way to do that is to have some bias. The classic example of this is you may jump when you see something that looks like a predator, but then turns out not to be. The cost of being wrong (see nothing when predator, see predator when nothing) means it's optimal to sometimes see danger that isn't there.
The argument seems essentially similar to theories on the evolution of aging that suggest we age because senescence improves evolutionary fitness when the environment changes on a comparatively short timescale [1]. The world changes, therefore a race to the bottom arises for ways to improve fitness that also happen to make life miserable and short for individuals. Miserable and short outcompetes hydra-like immortality or naked-mole-rat-like negligible senescence in the vast majority of niches.<p>[1]: <a href="https://arxiv.org/abs/1103.4649" rel="nofollow">https://arxiv.org/abs/1103.4649</a>
I do not think decision making is a good model for this concept. Culture is. In particular:
Traditionalism vs. Progressiveism
as cultural influences: the former acts to not overfit current conditions, while the latter tries to fit current conditions. Both influences have been critical for human survival.
Interesting argument presented in the paper, though I wouldn't frame it as "we've evolved to be inaccurate": it's really that the world can be so suddenly unpredictable that setting up strong, working paradigms of decision making in the short term can be worse in the long run than just winging it.<p>It's worth considering, especially in light of the authors' suggestion that we use computer/human decision-making systems to improve performance, as the world is still unpredictable, and can still break our paradigms. The biggest danger of setting up a good system to improve knowledge is that you'll think you've got a perfect one--we could improve our rationality and decision-making with computers for a long time, before an unexpected case cracks the system, and we're left floundering.
<i>maybe a hybrid computer / human solution will fare better</i><p>This is the purpose of technology. To enhance our skills. From fire to machine learning, tools are built to make our lives easier and help us make decisions better.<p>In the end we're better off with more empirical computing in our decision loops. Eventually hopefully we totally replace ourselves with better, more consisitently optimized decision making systems.
A similar conclusion was made using AI. Here is the video: <a href="https://youtu.be/dXQPL9GooyI" rel="nofollow">https://youtu.be/dXQPL9GooyI</a><p>The takeaways:
- The path to success is through NOT trying to succeed
- To achieve our highest goals we must be willing to abandon them
- It is in you interest that others DO NOT follow the path you think is right
Yes, of course we are inaccurate decision makers.<p>Although there's so many factors and chance that influence the results of every decision and you can only have so much information and perspective, so you can only do the best you can.<p>The map is not the territory. We work with models of how the world works when deciding things, not the actual world, so it's bound to not be 100% accurate.<p>Computers do the same, although they can crunch a lot more data than we can, they still work with models of the world, not the world itself.
There's a tradeoff between accuracy and speed.<p>For survival reasons the brain needed to evolve to make both quick decisions and well thought decisions.<p>If a predator appears in front of you you might not be able to give a lot of thinking to the decision of what you need to do.<p>If you are a nomad during the ice age and you need to collect food and prepare a shelter, or track a prey for long distances, you probably need to give it some thought.
I don't know if we're really that inaccurate, or that the complexity of the problem is vastly underestimated. If we think that making accurate decisions is so simply, we haven't we made AI yet?