Someone tosses a coin ten times; it comes up heads every time. What's the probability it comes up heads on the next toss? (Pretty darn high—part of @nntaleb's work is unprogramming you from your high-school rules of thumb.) Now consider the (related) Gambler's fallacy...
In this case, it's a theory about compensation: the worse one's luck is, the more likely it is to see a reversal. On the surface, it's irrational. The more bad luck you have, the more you accumulate evidence that the system is rigged.
But there's also an anthropic component. If the luck is bad enough, it starts to become inconsistent with your survival. You've accumulated evidence for correlations in the environment, but these correlations (may be) inconsistent with (people like you) being in this environment.
An example. You're in a city where everyone takes public transport. You encounter a string of bad delays. It's reasonable to conclude they'll end—otherwise people wouldn't take public transport. It's unlikely that you happened to show up right when the network collapses.
Of course, that's a bad heuristic in a casino, which relies on a constant influx of losers. But in other environments, particularly with persistent populations and no evidence for sudden changes in the underlying laws, it makes sense.
Another example: the three-card monte scam in a big city. You meet some guys on the corner, who convince you to go in on the game. You win, a few times. The more you win, the more you ought to be convinced that your luck will turn. Otherwise, how are these guys there?
(In this case, the anthropic reasoning concerns the scamsters—it's unlikely that you showed up right when their operation starts to fall apart.)
The general principle, which again is due to @nntaleb, is a new kind of failure. The "IYI", intellectual-yet-idiot. IMO, this is driven by standardized testing—we started to promote people on the basis of their ability to internalize fake-but-difficult-to-master rules...
In a previous cycle, the British ran their empire by fast-tracking twelve year olds who could master Latin grammar. The ambiguities of interpretation make this a much better idea than the SAT version, where one learns more abstract "grammars"—e.g., integration by parts.
Everything can be degraded, of course. When I give CMU students the coin-toss problem, it's usually unfamiliar enough that they're thrown back on a more complex, and life-integrated, form of reasoning. But ask that question enough, and tutors will arise to teach the solution...
...which negates the original value of the puzzle. (BTW, nearly all the students "get it", which bodes well for our future engineers.)
A final thought on this, before I get back to my real work. If you pose this kind of problem as a teacher, there's a second level to what's going on: your students are also modeling you! (and your class.)
i.e., they're asking themselves—is this a class where we're all living in la-la land, or is there some substance here, some connection to our own lived experience?
So, yes, I'm proud that most of my students get it. :)
This is lovely—yes! "I can't be the only idiot" summarizes why we keep waiting for the train with rising hope.
You might also connect it to the bulk-vs-long-tail. "I'm not the only idiot" is true when there are repeated tests of the system of similar magnitude.
OK, truly one last thought. A scam is more likely to succeed if you're doing something new, but can convince people you've always been around, that this is "normal". Certainly psychologically obvious, but it's fun to look at it from the point of view of rational analysis.
(In the case where the scam operates by inducing the Concorde/sunk-cost fallacy—slowly extracting money from the person, who continues to believe in a final payoff.)

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Simon DeDeo

Simon DeDeo Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @SimonDeDeo

18 Jan
Say we gain complete control over genetics—we can code development like Python. After 10,000 years, which is the most likely outcome of our decision-making?
Can’t believe the elimination of men and permanent sexual immaturity/liminal stage are getting so few votes.
I don’t think speciation events are likely. Silicon software has been converging on near-total comparability for decades.
Read 5 tweets
30 Dec 20
2021 predictions (🧵)
Student loans will (finally) be understood as a bubble on a par with the housing market in 2008.
Some of the fallout will be temporarily explained away by COVID, but many universities below the billion-dollar endowment mark (and some above) will find themselves short.
Read 22 tweets
29 Dec 20
You guys told me julia was cool, but it turns out that it indexes arrays beginning with one? I'm sort of not joking—what possible justification is there to violate the near-universal standard?
This is very Bluebeard's Castle. What else is in the cupboard, "Julia"?
One-indexing considered harmful. cs.utexas.edu/users/EWD/ewd0…
Read 10 tweets
19 Dec 20
Let’s check in, as @PaulSkallas says, with the DMs. Five books after @nntaleb 🧵
I’ll take the Taleb-namecheck seriously. Black Swan is an amazing book, which merges life history, fiction, philosophy, and (by the by) plenty of complex mathematics. So this list will be a little weird.
First: the Dialogues. They should be read as dramatic pieces. “What kind of mind do I see?” Every kind of foolishness and grace is on display. Socrates himself grows and develops over time. Translation mostly irrelevant. amzn.to/3mBXXAx
Read 23 tweets
15 Dec 20
The truth appears to be the exact opposite. What Paul calls “low-level” employees seem to understand the lines very well, while middle (and sometimes top) management blurs them to protect revenue sources. newyorker.com/magazine/2020/…
Quite striking to see everyone jumping on to (1) agree with Paul, and (2) talk about how dumb the content moderators are, when the New Yorker article basically shows all of this to be total fantasy. For example...
It’s not that content moderators can’t understand context—rather, they are explicitly instructed by management to ignore it. Image
Read 13 tweets
2 Dec 20
One way to find failing institutions is to look for the ones that use, and endlessly debate, metrics.

Healthy systems have notions of success that evolve on the same timescales as the tasks themselves. (🧵)
The graduate admissions that I’ve seen is pretty healthy. The endless debates on whether or not to use GRE seem beside the point: discussions on who to admit, in my experience, end up as really interesting conversations about science itself.
Faculty hiring is more complicated, and it’s easier for people to look at metrics. Usually, how many papers in “top” journals; sometimes divided by the number of years the person’s been out.
Read 17 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!

Follow Us on Twitter!