Tag Archives: Tom Griffiths

Algorithms to Live By: The Computer Science of Human Decisions by Brian Christian, Tom Griffiths

  1. Studying algorithms and how they might be implemented to help us better solve every day problems. Thinking about human cognition and behavior through this computer science lens helps shed light on how we think, why we make the mistakes we make, why and how we have such incredible computational powers, and what rationality really means. We can learn how to make the best decisions given the limited knowledge, time and other resources we have and how to do it with imperfect insights all while dealing with yourself and other messy people. Many problems are intractable but these algorithms will at least give you a jumping off point to begin.
Key Takeaways
  1. Master key algorithm for getting stuff done
    1. Earliest due date and shortest processing time is the master key to determining what to work on and in what order. Work on what has the highest value when importance is divided by completion time. Something must be twice as important if it takes twice as long.
    2. If all you want to do is get through tasks and reduce your to do list, do those things you can accomplish quickest first.  There are many algorithms to follow, it all depends on what your goal is and what you want to maximize.
  2. The Optimal Stopping Problem
    1. These cases you should have two phases: a looking phase where you commit for a certain period of time (usually 1/3 of the total amount of time you’re willing to look) and then a leap phase where you take anything that’s better than what you’ve seen during the look phase
    2. If there is some objective criteria you could set, you can then create a threshold and anyone or anything above the threshold should be accepted
    3. Our time horizon or the intervals of which were looking at strongly determine how much we explore and try new things and how much we exploit – going back to well known favorites. Since the interval determines the strategy we can also determine the strategy from the interval. An overload of sure things such as sequels is a good signal of short-termism.
    4. Optimism is the best solution for regret and we should give people, things, and experiences the benefit of the doubt because we don’t know their upper bound – how good they can be – because we don’t have enough information yet. You should be willing to explore when there’s not enough information to make a reasonable conclusion. However, in real life people tend to over-explore and not know when to lean towards the optimal solution. Win – stay, lose – shift
    5. Older people tend to have fewer social connections but that’s because they have refined over decades the type of people they want to spend time with and that naturally seems to decrease over time. This ties together our explore / exploit phenomenon because younger people who have a longer time frame are more on the explore phase and older people with a more finite time frame are in the exploit phase. As you get older and switch from seeking pleasure from exploitation versus exploring, your quality of life will necessarily improve as you are going back to well-known favorites more often
  3. A | B Testing
    1. Tinkering on an extreme scale is done today by some of the world’s largest companies to see what little tweaks between two options can cause. This iteration is done over millions of times per day so that the product/service/experience is ever improving, at least maximizing what is being measured and sought after. You can use this iteration mindset to make small changes and adjustments to your routine, habits, behaviors, thoughts, and see how it impacts you and others over time
  4. Sorting
    1. Fundamental lesson learned about sorting is that scale hurts.
    2. Simply by breaking tasks or projects down into more manageable units can sorting be reduced by multiples.
    3. However, the first question should be whether it needs to be sorted at all. Efficient sorting which is unnecessary is extremely inefficient and sometimes mess and disorder is the optimal solution
  5. Cache
    1. Keeping around pieces of information that you refer to often or anticipate needing shortly at hand so you can quickly retrieve it
    2. Keep things you use often in close physical proximity so that you can get them quickly
    3. It has been found in many different domains that events that have recently happened are more likely to happen in time and the longer it goes without happening the less likely it is to happen again (Lindy Effect)
  6. Over-Fitting
    1. Over fitting is when we try to use too much data too many factors into making our decisions and they not only make things more complex but actually lead to worse predictions and decisions. If there is high uncertainty and unlimited data, paint with a broad stroke and make it simple. Going into the nitty-gritty only hurts you
    2. It’s better to be approximately right then precisely wrong
  7. Other
    1. Procrastination is often associated with laziness but it can simply be that people lose sight of the important things and are racing through their tasks. They have the right strategy for getting things done but it is the wrong metric – favoring the easy over the meaningful
    2. Be aware of context switching costs. Flow and deep work sometimes takes an hour just to warm up and get into the flow and interrupting people or getting interrupted can ruin hours worth of work or more.
    3. There is a constant tension and trade off between throughput and responsiveness. If you’re too responsive you got nothing done and if you’re throughput is all you’re maximizing you’ll never respond to anyone.
    4. Thrashing is the point when your interrupted so often and have so much to do that you get no actual work done and at this point you can step back and reevaluate and often just do whatever you can get done and not worry about the optimal way to do it.
    5. Batching tasks and having set times to do things such as only looking at emails first thing in the morning and at night is a good way to keep from being interrupted too often
    6. You can become better at predicting by knowing if you’re dealing with power laws or normal distributions and the better information you have of course the better guess you can make. That’s why we are quite good at predicting how much longer a person can live for we know the general lifespan of people
    7. Our predictions tell us a lot about who we are because they’re based on our experiences.
    8. If you can’t explain things simply you don’t understand it well enough
    9. If you can’t solve a problem, relax the constraints and try to solve an easier version of the same problem to see if it gives you any clues or jumping off points for how to solve the real problem
    10. Exponential back off is a technique you can use when things fail or you don’t know how to proceed. For example, if people cancel their plans with you last minute wait a week to reschedule. If they cancel again, wait two weeks. Then four, etc…
    11. The first and only rule of hierarchy is that the hierarchy must be preserved
    12. The innovators dream is not a eureka moment but rather a situation that makes you say, “huh, that’s funny.”
    13. Seek games in which honesty is the ultimate policy and then just be yourself – Vickers Auction – where the winning bid pays only the second highest bid price
    14. Sometimes even the optimal strategy will yield bad outcomes which is why you must focus on process over outcome
    15. Sometimes good enough is simply good enough
What I got out of it
  1. Some good techniques and thought processes for how to make better decisions