Game Theory
This blog post is about game theory and how it can be used to help you make decisions. I am going to start with a funny example on my take for why vehicles are getting larger. Then, I am going to provide an example for a situation at work. In the realm of software engineering, game theory can be applied to model and analyze various scenarios, such as distributed consensus algorithms, load balancing strategies, task scheduling algorithms, and even software development processes involving multiple stakeholders. By representing these situations as games and analyzing the costs/benefits, we can gain insights into the optimal strategies for achieving desired outcomes. We may not always achieve the outcomes that we want, but we at least have a tool which helps.
It seems that every year cars just get bigger and bigger. My same make and model of vehicle I bought 8 years ago is still made, but it is almost unrecognizable. The car is at least 20% larger than when I bought it. Why is the exact same model getting bigger? Anecdotally, looking at the vehicles driving around I notice bigger vehicles every day. Parking spots that seemed big enough 10 years when they were new now cannot even hold those same vehicles. What gives? It could be car manufactures making larger vehicles for more revenue/profits. It could be EPA emissions, commonly referred to as the truck/suv loophole. But, I blame game theory for safety. Generally speaking a larger vehicle is safer. If an F-250 3 to 4 ton truck hits your sub-compact 1 to 2 ton car, which one wins? I don’t care what any crash rating says, one clearly wins. This is commonly referred to as a symmetric game, because everyone gets the same benefits. When you buy a larger vehicle you feel safer. The catch-22 is that this game plays on repeat - hence small increases to average vehicle size compound over time making larger and larger vehicles. When will it stop?
| Everyone else stays in a small vehicle | Everyone else gets a larger vehicle | |
|---|---|---|
| You stay in a small vehicle | 0, 0 | 0, 1 |
| You buy a larger vehicle | 1, 0 | 0, 0 |
I had a fun situation at work and I won’t bore you with the details, but it ended up being a simple prisoners dilemna. It is funny, because I did not realize that was the situation I was in until someone called it out. Once they called it out, I knew exactly how to solve the problem! If I had identified the problem before hand, I could have prevented a lot of wasted effort. The problem really boiled down to doing work now or doing work later. The nuance of the situation is that this is a shared launch, so either we both make changes or no one makes changes. Making the changes today is perfectly quantifiable, we all know the changes that we need to make. Making the changes later is unquantifiable.
| They make change | They don’t make change | |
|---|---|---|
| We make change | -1, -2 | -INF, -INF |
| We don’t make change | -INF, -INF | -2Y, -Y |
If we think that Y is going to be really large, then we would all agree on making the change. But, there could be a chance (however small) that Y is 0 or 1. So, now what? Well, we could prove that Y is large which would convince people - but without getting into the details we do not have time to do that. Welcome to a non-zero sum prisoners dilemma. We can save effort by making a change now or we can delay and maybe someone else will have to do the effort later.
Many of our decisions in both life and work boil down into a simple cost-benefit matrix. If you represent this cost-benefit matrix in the normal form of game theory, you will understand the situation. This will help remove any bias you may have and look at the situation objectively. There are certainly scenarios where not all parties have perfect information and they can be really tricky to model, but they help you analyze the situation too. Am I making the correct decision? Is my “opponent” making the right decision? At the very least, you should have enough information to solve the problem.
The post is done, but you are still reading? Well, that situation is a bit more nuanced. Being the narcissist I am, what are my options and what are my cost/benefits? I can chart out all branches. In this case, I am evaluating two variables X and Y. I personally value X much much more than Y, so I really want to minimize actions that cause me to lose X. While I do not want to lose Y either, how much is Y really worth? Well, when I look at it from this perspective my choice is clear. Lose-lose-lose-lose situations are not a fun place to be in! But, I get to choose and live with my decisions. Option 2 sucks, but the potential outcomes are more palatable.