Humans suck at rational thinking -- but there's hope

published Apr 03, 2009, last modified Jun 26, 2013

We actually suck at a lot of things, especially at probabilistic math... and we need to get better, fast.

Quick, what's more probable?

  1. that a god exists
  2. that an omniscient god exists

The correct answer is 2.

Wanna take a guess at what the majority of people think?

Their answer is 1.

Go out there, poll your believer friends and family.  Verify this -- don't just take my word for it.  The more detail you include in an option, the more people will judge it probable.

Why is the people's answer wrong?  Surely the majority knows better!

Reality isn't democratic.  The answer lies in probability theory instead.

The probability of an event A (say, that you will crash in your car) will always, inevitably be less than (or at most, equal) than the probability of said event A (crash) happening in conjunction with an event B (that you have a passenger when you crash).  It doesn't matter that "you never crash your car", or that "you always have passengers in it", the likelihood that you will crash with a passenger will always be lower than the likelihood of you crashing under all possible sets of circumstances.

Yet people routinely fail at this elementary logic test.  They intuitively miscalculate probabilities and incur in this "conjunction fallacy" frequently, whether they are judging the probability of getting attacked by a terrorist, or the probability of getting a disease.

Yeah, but why do people make this mistake?

It's because we evolved that way for survival reasons.

Yeah, that sounds like a cop-out.  But it's the truth.  In the face of the "local maxima" limitations of evolution, we humans have evolved a number of "approximative" approaches to understanding the world.  Our innate physics tells us that rolling or flying physical objects have diminishing "oomph" (they don't, that's an effect of friction).  Our innate judgement of other people works in seconds, certainly not enough time to ascertain whether someone you met is "good" or "bad".  Our innate judgement of objects "animizes" them, confers them magical properties of living things.  These aren't preconceptions you learn; on the contrary, you "came" with these to the world, and you must unlearn these intuitive preconceptions if you want to be civilized.

And squarely among those preconceptions is our innate judgement of probability.  It errs.  A lot.  Of course, it errs on the side of caution, not on the side of curiosity.  This is why you get shit like gods and Daddy Government -- because we are unable to correctly judge on the basis of intuition alone, and so it becomes easier to accept all that shit than actually deduce the truth values of the propositions presented to us.  And, of course, sooner or later, we become emotionally attached to them and we cannot shed them, despite the overwhelming evidence against them.


  • Say you were a jungle man and you saw a tiger.  Now say you were curious.  OK, you're dead now.  Now imagine that instead of being curious, you imagined that a Tiger God did not want you to keep close to tigers.  It did not matter one iota whether said Tiger God existed or not -- what mattered was that you got to propagate your genome; it's pretty clear that "superstition" has evolutionary value.  (This is, incidentally, how many primitive cultures "animized" the world).
  • Say you were a random hunter-gatherer whose village was invaded by a strong enemy that let you live in exchange for 20% of your "production" (a form of robbery we now know with the name of "taxation").  Well, guess what, you lived!  Thus, onto feudalism and now "democracy"; even though the "strong enemies" are now vastly outnumbered by the productive sector of our society and really wield no power without their tax income, people haven't updated their probability estimates and continue to fund the "strong man".
  • Why do men overread romantic interest in women?  Because it pays.
  • Why do people fear guns and want to ban them?  Because they mistakenly think that banning guns will decrease the probability of them being shot (hint: it's actually the opposite way).
  • Look at the world.  What are people more afraid of? Terrorist attacks and crime, or car crashes?  OK, now look at how many dead people can be accrued to these two groups.  Does it make sense to continue being afraid of the bogeyman?

My point is that those approximative approaches kept us alive until today, but they are drastically, flat-out wrong -- and this is why "intuition" is never, never enough.

OK, but what does it matter?

It matters the world!  Probability errors misinform decisions around the planet.  Real personal and policy decisions, with real effects in the livelihoods of millions of individuals.  If you think that a god exists (a type of conjunction error where you update your probability calculations in the wrong direction on account of the existence of your planet), then you will act accordingly and plow yourself into buildings with planes.  If you think that terrorist attacks are more probable in certain scenarios, you will fail to foresee the other, more likely scenarios.  If you think that you will get more pussy by hitting on a limited number of chicks a certain way (instead of hitting on many chicks with varied approaches), you'll be sexless.  If you think that banning guns is better than owning and learning how to operate one, you will be serving yourself as a victim for the next robber.  If you think it's likely that we'll eradicate poverty with welfare and social services, you'll be doomed to slavery while the "strong man" gets richer and richer.  If you think that you are better off paying taxes to the "strong man" than paying a private security company to protect you from domestic psychopaths, you will continue to fund the statist machinery that empowered the "strong men" to murder 170 million people in wars and famines just in the 20th century -- overwhelmingly more than domestic casualties for the entire human history.

So, we are better off eradicating these primitive statistical errors from our minds completely (not that I think we will ever succeed as long as we have statism around and in charge of schools, but it is still something we need to fix).

Fortunately, as with all other human endeavors, there is hope.  Robin Hanson just wrote up that it's possible to dramatically lower the conjunction error (what we discussed in the first paragraph).  All it takes is coolly consulting with several people about one's judgement.  Great news!

Rationality has a long way to go, but as long as we have people like Eliezer, we're on the right path.