Hacker News new | past | comments | ask | show | jobs | submit login
Braess's paradox: adding roads can increase congestion (crowddynamics.co.uk)
28 points by RiderOfGiraffes on Nov 8, 2009 | hide | past | favorite | 9 comments



I always prefer the string and spring version because people can physically see it happening.

For an illustration suppose we have two springs A and B connected in the middle. A is tied on the other end to the ceiling, and B is tied to a weight. The springs each expand 1 cm/newton of force, and the weight exerts 100 N of force pulling them apart. So each spring is 1 m long, and the whole arrangement is 2 m. Let's attach two "safety strings" of 1 m in length. One is attached from the top of B to the ceiling and the other is attached from the bottom of A to the weight. Cut the tie in the middle and the weight will rise up 50 cm!

People are shocked to see that cutting a string makes the whole arrangement stronger, but it is really easy to replicate.

According to http://www.maa.org/mathland/mathtrek_11_10.html this visually startling version of the paradox was developed by Joel E. Cohen.


Neat visualization. Am I getting this right?

Arrangement 1:

Ceiling -- A -- B -- Weight

Arrangement 2 (Picture is a bit weird, but both are attached to the ceiling and the weight and not to each other right?):

(Edit: Making picture clearer).

Ceiling -- A -- Weight

Ceiling -- B -- Weight

If thats right... it shouldn't be that surprising. In (1) B has to carry all of the weight, and A has to carry the weight and B (even if B weighs 0 it still has to carry it) so each spring is carrying the weight. In (2) they are in effect sharing the weight.

Its probably easier to think of the action movie scene where William falls off the cliff, Brad catches William but starts falling and Adam catches Brad and holds on. Brad has to hold up William and Adam has to hold them both up. But, if instead, Adam and Brad each took one of William's arms, it would be easier for both of them.


Change the picture for the second to

Ceiling -- A -- String -- Weight Ceiling -- String -- B -- Weight

and you've got it exactly. Also the reason why it works.

Once you understand it the result isn't that surprising. So the physical model provides both more intense initial surprise and also a direct way to visualize the explanation. Which is why I prefer that version. :-)


Some links about the paradox itself:

Wikipedia page: http://en.wikipedia.org/wiki/Braess%27s_paradox

English translation of the original paper: http://homepage.rub.de/Dietrich.Braess/Paradox-BNW.pdf

Preface to the translation: http://supernet.som.umass.edu/articles/preface_to_braess.pdf


This is reminiscent of (in effect, if not in mechanism) Belady's Anomaly. (http://en.wikipedia.org/wiki/Belady%27s_anomaly)


Side question: Is "paradox" commonly used to mean "what some people find unexpected", rather than "a logical or self- contradiction"?


It is used in both cases.

For example in probability the St. Petersburg paradox. http://en.wikipedia.org/wiki/St._Petersburg_paradox


A paradox need not be self contradictory.

There are generally 3 resolutions of a paradox 1) The premise(s) are wrong. 2) The logic is in correct. 3) The "answer" is not self contradictory.

If the resolution is 3 then there is no contradiction. Even if its 2, there isn't technically one (just a mistake). If its 1, there might be a contradiction.

I remember discussing this with someone and coming to the conclusion that the word "paradox" really should be "paradox for me" because its only a paradox to someone until they see how to resolve it.


I've always used it to mean "something that you find initially contradictory, but can be rationally explained".




Consider applying for YC's W25 batch! Applications are open till Nov 12.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: