This is interesting because it seems to defy logic until you really understand what is happening with the springs. The application to traffic patterns later in the video is interesting too.

The part with the springs starts at 40 seconds, 2:46 explains it.

Braess’s Paradox video

That was pretty cool - thanks for sharing it!

okay, it was more of an optical illusion then. cutting the middle string changes the set up from series to parallel.

quote:

Originally posted by Rey HRH:
okay, it was more of an optical illusion then. cutting the middle string changes the set up from series to parallel.

It baffles me how it applies to other things like traffic.

Serial is multiplying and parallel is division.
(physically speaking. sort of...)

quote:

Originally posted by mrvmax:

quote:

Originally posted by Rey HRH:
okay, it was more of an optical illusion then. cutting the middle string changes the set up from series to parallel.

It baffles me how it applies to other things like traffic.

Thanks for prodding me to find out more. I came across this before in case studies of bank teller window lines or check out lines. Moving from separate lines where people choose which line to go to, resulted in more complaints when some people get hung up behind a person whose transaction is slow for some reason like their card isn't accepted, etc. So they force all lines into a single line and the one at the end goes to the next available teller or cashier. This theoretically would lower the average waiting time.

But it didn't have a name at the time.

Paradoxically, the average time in line increased but the complaints went down. When they analyzed it, they found that the teller/cashier spent more time with each customer because they did not have the pressure of the line behind their immediate customer which made them work faster. The complaints did go down as the customers perceived the set up as being fair as they aren't randomly stuck behind a slow customer ahead of them.

ISTR an article on "line theory" that appeared in a 50's or 60's edition of Scientific American. It was clearly shown that the "single queue/multiple servers" was the most efficient setup, both through mathematical analysis and actual experimentation. By "efficient" they meant moving the most people through the pattern of transactions in the shortest time. If my memory is holding up, the second best was with multiple lines/multiple servers where line selection was moderated, self-selecting lines was worse.

I don't remember that they addressed the issue of customer satisfaction, i.e. self-important concerns that some other guy was being treated better than they were.