Storytelling Science

The Science of Bureaucracy: Building a Win-Win Society

Amitabha Mukerjee


The other day, my taxi got stuck in a election-rally traffic jam. Instead of sticking his nose into whatever empty space was available like it happens most of the time, my driver waited patiently in the lane, and soon enough the traffic cleared up. In conversation later, he said that many professional drivers these days have realized that following rules is better for everyone.

You   She FollowRules Speedup
FollowRules 3 / 3
0 / 5
Speedup 5 / 0
1 / 1
This set me thinking about the science of this situation. Imagine that you and another car are approaching an intersection at the same time. Let us say that the maximum time you spend at the intersection, 5 minutes, happens when the other car shoves you to a corner and zips off. This is the Lose-Win scenario, where she saves 5 minutes and you save nothing. If both of you follow rules and negotiate the intersection in an orderly fashion, each of you will spend 2 minutes. This is the Win-Win scenario, where each of you save 3 minutes. If you speed up and she follows rules, you gain 5 minutes (Win-Lose). If both speed up, then you both end up stuck, and you save only 1 minute on the worst case (Lose-Lose).

This classic scenario is actually a famous problem in Game Theory, which studies how behaviour evolves within a group. If you are playing this game over and over again, you must decide, every time you encounter this situation - whether to try for Win-Win by following rules, or to try for Win-Lose? A good strategy in this situation is Tit-for-Tat - where you do to the other as they do unto you. If the other cooperates, you cooperate. If the other tries to act smart, next time, you do the same. In the beginning phases of the game, everyone ends up in the Lose-Lose square.

This is actually what has happened to us in India, where we are always trying to be one-up on the other and both of us end up losing.

What is surprising about all this is how easy it is for anyone to try it it out. You can take a computer, define a strategy for playing, and simulate how a number of players would play each other in this game. Try different strategies, and see which one is doing better. In the next step, you can also have a mechanism for mixing strategies, and between runs, you keep the winning strategies, and gradually prune out the poorer strategies. In this process (which is called Genetic Algorithms) you find that short runs (say 200 encounters), already show a tendency towards Tit-for-Tat (TFT), but with results more in the Lose-Lose square. As the number of encounters go to 1000, or 10,000, you find that co-operation strategies are dominating - this is because over time, if you get 3 rewards every run and a very occasional zero, it is better than getting 5 once in a while and 1 on the rest. Once everyone in the population starts to understand this, everyone starts to cooperate, and gradually this becomes the standard.

If you think about western nations, this is where they are ahead of us. Most people follow rules, and the entire society is better off for it.

However, if the rules of the game change suddenly, then soon enough everyone can be back again Lose-Lose - and the system needs a long time to recover its confidence and return to Win-Win again.

This is what happened to colonized nations like India - suddenly the rules changed on us and we were lifted into the industrial age - and our behaviours which might have stabilized earlier into some sort of a Win-Win, went back to Lose-Lose. At independence, the rules changed once more. The good news is now that we have had two generations of experience with being stuck at Lose-Lose, even our taxi drivers are beginning to see the value of following rules!

This analysis is of course very simplistic, but on the whole, Game Theory is a very powerful tool and can be used in many areas. Recently a group of students at IIT Kanpur used the games theory model to see how a bureaucracy works. In a bureaucratic setup, the organization meets its basic criteria even if only 50% targets are met. The employees are organized in a hierarchy with class 1 to class 4 employees. The class 1 employees are paid the most, and are evaluated the most stringently. In the study, they had a population of employees, and plugged it in the model, and lo and behold - most of the class 1 employees were working at 8 hours or more per day, whereas the class 4 employees worked for barely half day. (Adventurous souls can look up their report at

Game theory is not rocket science - any child can program a computer to run these computations, and share in some of these dramatic and interesting results. Sit with a child, share this story and see his imagination light up!