9 Feb 2015

How To Make Real Life Decisions Using Game Theory

Introduction

In a previous blog I mention that our greatest power lies between stimulus and response...in the freedom to choose. This means that our life is primarily, though not exclusively; a sum of all the decisions we make.

We make hundreds of decisions every day. From what we should do upon waking up all the way through to when we go to bed. Wouldn't it be nice to know what the best decision to make is? Thankfully there is an area of study that looks into how to make good decisions. This study is called Game Theory, my area of interest.

I am a fourth year computer science student in my final term of university. I am specializing in artificial intelligence; specifically game theory, social choice and mechanism design. With this blog, I hope to explain how we can use concepts in game theory and apply them to our lives to make better, more informed decision.

Defining a game

A game is defined by:
  1. Finite set of agents (or players) N
  2. Each agent i has an action space Ai
    Ai is non-empty and finite
  3. Each agent has a utility function ui: ( A1 x ... x An ) -> R (real numbers)
If you didn't understand the math above, I will explain it in context of real life
  1. In any given scenario, there are some number of participants. Let's call that number n. Now if n=1, we are in a scenario where it's only us. We have to make a decision with things and ourselves.
  2. Every participant involved has a finite number of options to make a decision from. We will call this their action space, a set of possible options that an individual has to chose from. It is important to note that very rarely does a person know their entire action space in interdependent situations. That is to say, there is almost always 'another alternative' that was there, but we simply didn't have knowledge about at the time.
  3. Every participant has some degree of knowledge to how favorable different outcomes will be, given all users taking actions in their action space.
It is crucial to note that the 'size of our action space' and 'estimates of utility' are based on our maturity and preferences. This will be made more clear once we look an example.

Making decisions

In the individual setting, n=1. We are the only participant. One major difference in the individual setting is that our utility function is U(a)->R and we are concerned with finding the action to take that gives us the maximum utility.

The best way to demonstrate how to optimize decisions is through an example.
  • It's 11pm and we just finished an episode of our favorite show.
    Our potential action space could be A={ watch another episode, go to bed, study }
    We could estimate U(watch another episode) = 1
                                   U(go to bed) = 5
                                   U(study) = 4
    From this, we could make the decision to go to bed.
Action space
The above could be the action space of a high school student.
  • A person suffering from a drug addiction may include using again after the episode.
  • An older person may also include activities such as reading, stretching and other preferred activities with high utility.
  • A parent may also make sure to include checking well-being of children.
Utility
  • If we have a test the following day, the utility to study may be higher and to watch TV lower.
  • If we are experiencing a breakup, the utility for all our actions may be slightly biased.
  • The drug addict may have an extremely flawed estimate of the utility of using again.
There are 3 factors that influence our action space and utility, and how much they influence our utility can be caught in the 3 sequentially arranged Greek words ethos, pathos, logos.
  1. Ethos - Character The drug addict may have a flawed view on drugs heavily impacting the utility of that action leading to poor decisions, regardless of many other factors.
    On the contrary, a parent may have a very accurate view on child upbringing and highly prioritize the care of the children, regardless of many other factors.
  2. Pathos - Emotion After ethos comes pathos. A breakup will be influencing our actions, but not nearly to the degree that a drug addiction or a child may. 
  3. Logos - Logic Time management influences our actions as do other logical realities, but do so far less than our character or our emotion.

Increasing action space and utility estimates

  • Our action space is directly influenced by our wisdom. Wisdom is your perspective on life, your understanding of how the various parts and principles apply and relate to each other.
    In independent situations, we can seek to know who we are to increase our action space.
    In interdependent situations, we can seek to understand others to increase our perspective and hence increase our action space.
  • Our utility estimates are directly influenced by our guidance. Guidance is your source of direction in life. Encompassed by your internal frame of reference that interprets for you what is happening out there. The standards or principles or implicit criteria that govern utility estimates.
    Having more accurate utility estimates requires learning from others and our own successes and shortcomings. Utilities should increase or decrease after evaluating results of important decisions.

Think before you decide!

What is the point of this blog? To simply understand that we must consider all our options and pick the best one.