Algorithms and Strategies

This is a list of heuristics for how to act and think in a variety of situations. No one can be a perfect act utilitarian1, it isn’t possible to calculate what is best for everyone: however instead of sufficing with some basic rules it makes sense to update the rules of your ethics over time so that you can handle more and more situations quickly. Even with moral uncertainty, it makes sense to update the heuristics guiding your actions over time to increase the odds you do what is right. This list is far from all inclusive, but it reflects patterns of thought I notice myself having that I don’t always notice in conversations with others who have similar goals.

Training:

  • Never grow complacent. If you want to be satisfied in life, make sure satisfaction depends on developing your abilities or doing something useful.
  • There are many sources of  happiness. The more pleasures you know, the more masters you have. Aim to serve only the masters which will help others and make you a better person and your happiness in the moment won’t come at the expense of the future (aka don’t bullshit yourself).
  • Always optimize: it gives you something to do, exercises your brain, and gives you more free time for whenever you do need to rest.
  • Pre-commitment is a great way to actually do things. Make sure you will be embarrassed and hated if you don’t follow through with the things that you think are important to do. That being said, don’t allow guilt to be your only motivation and don’t pre-commit when you need the ability to change course. You should deter bad action, but if you find yourself constantly punishing yourself then your self punishment isn’t working.
  • Thoughts are sticky. What you think about at one time determines what ideas will become mentally available in the near future, and also how your memories will be stored. This is probably an additional part of how spaced repetition works: that which you are trying to recall will gain more possible triggers for being remembered by having been associated with multiple different previous mental environments. Beware, there are many ways for sticky thoughts to wreck your productivity. Getting a stupid song actually stuck in your head probably does reduce your mental capacity, getting caught in a depressive spiral of thought is not useful if the spiral makes no options mentally available for escape from the spiral and changing your future. Model the way you think at different times, and use this information to not only increase productivity, but also the odds that you will change your mind when it makes sense to do so.

Rationality/Productivity:

  • Increasing rationality should often priority #1, because it often makes you better at all other priorities faster than other priorities possibly could. When rationality starts having diminishing returns, it will tell you, and you’ll be in a good position to know what to do next.
  • The will to believe and the will to learn are opposites. Let your beliefs update precisely as much as the evidence demands.
  • Emotions evolved for a reason. Notice what your emotions help you do, and what they make you worse at. Embrace attitudes which help ensure that your emotions are evoked at the times they are useful.
  • Make the garden weed itself. The less things that you must consciously think about, the more brain power you can bring to bear against the problems at hand. Minimize the actions you do which produce future chores for yourself.
  • Exercise does more than make you look better and be healthier. You can actually do things faster and think better, but this requires you to actually take advantage of the new abilities which you gain from exercise.
  • If there isn’t a reason you need to move slowly, always run and bike to the places you are going whenever doing so saves time. It’s free exercise: it gives you more free time, unlike going to a gym.
  • Ergonomics matter! By rearranging and improving your clothing, bags, computer, living space, and work-space, you can save a lot of time, avoid injuries, and avoid vices. The more actions per minute you are able to perform, the more time you have for actions which require slow concentration.
  • Do not spend too much time organizing your things: your focus on organization should be to save time, and maintain whatever level of hygiene and appearance is needed for your situation.
    • Simply stack papers and emails as you use them: the easiest to find ones will gradually become the ones you use the most
    • Standardize where you put important things, so you don’t lose them and can save time sorting. This applies not just to your living spaces, but also your pockets and bags. The mind is great at remembering things spatially.
  • Argue kindly. Identify areas of agreement, make your opponent’s argument better than they do, acknowledge what you have learned from them, then criticize. Do not just argue for one side, evaluate which side you should argue for.
  • To help others change their minds, get them to explain their beliefs thoroughly. When someone tries to explain something which they don’t actually know, they will feel confused and be more likely to investigate their own gaps in knowledge.
  • When teaching others about things they are biased against, use the Socratic method and appear neutral. Blunt disagreement will rarely produce the useful feelings of uncertainty in those you want to help learn.
  • If you notice fallacies being used by others, self-reflect to see if you have done the same.
  • Properly done, thought experiments are useful for determining things you should do in real life. The closer the experiment is to real life, the better intuitions you will generate.
  • Beware of correlated errors. You might believe something for many reasons, but all your reasons could be wrong for one reason.
  • Simple and sometimes even sloppy algorithms outperform intuitions at prediction when based on available statistics. Intuition must be disciplined to be useful.
  • Intuition can not be trusted when there is a lack of stable regularities in the environment. Statistics deal better with noisy environments than intuitive causal models.
  • Opportunity for expertise = (quality of feedback) x (timeliness of feedback). Accelerate your learning by setting up environments with iterated rapid and specific feedback.
  • Regression to the mean is not really a separate idea from the idea of a correlation coefficient. If correlation is imperfect, there will be regression to the mean. The brain matches intensity, but does not naturally regress toward the mean based on the correlation coefficient.
  • Pop culture is often more influential than religion or ideology. Becoming more rational can sometimes be harmful at first, because of increasing consistency with bad ideology.
  • Being in a state of flow can help you get things done fast, perform very well, and feel happy, however this comes at the cost of tunnel vision.
  • Do not get too narrow in optimizing, explore many of your options before you exploit the best ones. Decide how long to explore based on how much time you have to act.
  • The more you learn the less you feel like you know, because it becomes harder to create plausible narratives to explain and account for all of the data you have. With few data points, you can connect them however you want, the more data points there are, the less options you have. The mind seeks plausibility, but plausibility is not probability: beware of the narrative fallacy.
  • With the exception of mathematics, hard questions often lead the brain to make a substitution and answer an easier question. If you feel that a hard question is easy, you may not have actually answered it.
  • How to make a good estimate method 1, use biases to mitigate bias:
    • Step 1: Find  the average value and anchor on it
    • Step 2: Estimate the value based on your impression of the evidence
    • Step 3: Find a correlation value between evidence and value
    • Step 4: Move away from the average toward your impression value based on correlation coefficient
  • How to make a good estimate method 2:
  • Depression = sadness activation + deactivation of motivation, drive, and decision making. Taking actions, however minor, that increase executive function may help.
  • When discussing a topic, try to get opinions and impressions anonymously before people lose them to groupthink.
  • The less useless stuff you buy, the less useless work you’ll need to do paying for it and moving it.
  • When planning on what to do in the future, seek to minimize regret. This is a great way to mitigate planning fallacy and get yourself out of focusing only on the current moment.
  • When having a great conversations, remember interestingness bias: what people talk about is not necessarily the most reflective of reality even if what they say is true. Don’t plan based on stories you have heard from others without reflecting on the accuracy and applicability of such stories.
  • When a problem is impossible to solve given the constraints you are under, try relaxing constraints to generate ideas for partial solutions which do the least harm. If you are the one setting the constraints for others, consider setting proportional penalties instead of absolute command and control types of rules so others have many options for problem solving.
  • Instead of complete removing people from your life or suffering through someone consuming too much of your time, consider exponential backoff as a way to quickly find the level at which it is useful to spend time with people you’t can’t handle for extended periods of time.
  • Often the best way to get rid of competitive pressure toward losing things of value is to find best overall solution for those you are competing with. It is easier to cooperate when everyone knows what defection looks like and how to punish it.
  • When working with others, be computationally kind: give people your preferences, some solutions, and default options so you don’t waste their time forcing them to think unnecessarily about things they don’t need to.

Ethics:

  • Ethics should drive your actions, what other things are important enough to matter?
  • Ethical reaction time can be very important because many opportunities to do good are unexpected. Although effective charity is active rather than reactive, there are many situations in which you can take effective actions by being more responsive than those around you. Positioning yourself to encounter, recognize, and act on these opportunities is important.
  • Responsibility lies with power, not just error: no one blames a deer for stepping in front of a hunter’s bullet. If you can change something for the better, you should whether or not it is your formal responsibility to do so. AKA: Heroic responsibility
  • Do not discriminate on the basis of any factors irrelevant to your ethics. Just avoiding long known problems like racism and sexism is not enough to avoid treating individuals or entire groups of people in unjustifiably terrible ways.
  • For a utilitarian: freedom, equality, rights, desert, and rules can only be derivative values. If your use of these principles is not based on what is good: what actually makes conscious beings better off, you are not using them appropriately.
  • Don’t increase personal freedoms which are only harmful. Checks and balances are useful for yourself, not just the government. Power is neutral, and can be abused by almost any type of individual or group.
  • To be for equality in a meaningful way, you must have a specific thing which you wish to make equal. Equal outcomes, equal consideration, and equal starting conditions are often contradictory goals when you are constrained by other factors.
  • Having the legal right to do certain things, does not mean it is always right to do certain things.
  • What people deserve matters only for reinforcing or punishing behavior. Do the actions which cause the most good given the current situation, not the past. Consequentialism is forward looking.
  • Good rules are heuristics for what causes a net good, most should have exceptions you can notice. Once you have formalized the exceptions: update the rule.
  • Symbolic action is pointless other than for selfish benefit unless it actually causes good things to happen.
  • Think about and be aware of your own moral uncertainty. Try not to take actions which are likely to be very harmful according to moral systems you think are plausibly the best, but which aren’t currently your favorite.

————————————————————————–

Notes:

  1.  In Do the Right Thing, Stuart Russell argues that rationality is a property of “programs” within a finite architecture, and their behavior over time in the task environment, rather than a property of individual decisions. In other words, computers can’t be act utilitarians either. Return to top
  2. There are many sources for the list, but as I’ve blended ideas together overtime tracking down sources would be rather tiresome.
  3. I will likely remove some rules from this list over time, to increase the density of information that may actually be useful.
  4. Algorithms to Live By is a rather useful book for making better decisions and saving yourself time with many familiar situations.
Advertisements