Having made myself unemployed and uprooted, I now need to make some decisions about what to do next. Since I spent much of the last two years reading about bias and rationality, it makes sense to try to apply what I've learned to this decision.
I'm publishing this because I've seen very few concrete examples of in-the-wild rational decision making. I don't particularly expect this to be a shining example of such, but it is at least an example :)
The inputs to a rational decision are terminal goals. Rationality has nothing to say about what those goals should be, only how to effectively pursue them. My goals are to be happy, excited and useful.
To make a good decision I have to explore the search space, make predictions about the outcome of each option, evaluate how well that outcome matches my goals and then pick the best outcome. So, the likely reasons for making poor decisions are:
- Over-narrow search space (eg took me a good while to realize that I don't have to be a programmer forever)
- Poor predictions of outcomes (eg imagining freedom of being a student but not imagining being poor, paper-submission grind, losing interest in the subject)
- Poor evaluation of outcomes (eg poor understanding of happiness)
- Stopping too early (eg because of desire for certainty, or being overwhelmed by choice)
How can I protect against these mistakes?
Poor evaluation of outcomes
Specific reasons I can recall for making poor evaluations of outcomes are:
- Over-weighting unimportant concerns (eg was very resistant to moving to US because I dislike the politics, but this had barely any impact on my day-to-day life)
- Under-weighting or forgetting important concerns (eg once took an exciting job that required living in a remote town with a two hour journey to meet friends)
To correct these, I need to figure out what concerns are important ahead of time.
Inside view: in what circumstances have I been happy in the past? This is easily distorted by individual strong memories, so I broke it down into roughly 6 month periods and for each period tried to recall several typical work days, several typical weekends and who I regularly spent time with. It's pretty personal, so I won't include the details here, but once I wrote it all down there were some clear patterns, some of which surprised me:
- Excitement was predicted by sense of agency. Not clear why I have periods of high and low agency or in which direction the causality lies, but making choices that restrict my perceived agency (eg living on a visa) is probably risky.
- Happiness was predicted by friendships with excited and agenty people. Occasionally those are people I work with, but more often not.
- Skillful exercise seems to be important for happiness. I don't have a good handle on how exactly to explain this - something like a physical equivalent of the need for cognition - I need to move. Gyms and most sports do not satisfy this need. Climbing and parkour both do.
- The reason I have left every job is frustration with instrumentally irrational strategic decisions.
Outside view. What does pysch research have to say?
- Basic needs - healthy, no financial worries, low environmental stress (eg quiet enough to sleep well)
- Quality of close social relationships is the best predictor of happiness once basic needs are met
- Motivation is determined largely by autonomy, mastery and purpose.
So far I had mostly been thinking about what kind of work I want to do, but this suggests that it is at least as important to figure out how to maximize my chances of making good friends and collaborators. I notice that almost all my close friends came from repeated meetings within some not-purely-social framework (eg work, climbing, RC alumni) so it's worth focusing on those.
Poor predictions of outcomes
Specific reasons I can recall making poor predictions about similar situations in the past:
- Believing self-reports (especially in interviews)
- Focusing on the stereotypical outcome rather than the likely outcome (eg working in a game studio was mostly tedious debugging of stupid engine bugs)
- Over-weighting inside view (eg this project is bound to succeed! ...even though vast majority of similar projects have failed in the past)
- Making decisions while hot
The last can be avoided by pre-committing to not making a decision until some set point in time.
The first three, and many other mistakes like them, are examples of attribute substitution. One way to combat this is to follow an explicit, external process. For each value that I've decided is important I've come up with a short list of questions.
- Is the city pleasant to be in?
- Is there good public transport?
- Is there easy access to nature?
- Is the weather good? (impact of weather on happiness is probably over-weighted in general, but bad weather is an obstacle to climbing and parkour)
- Can I afford to live comfortably?
- Will I have to commute? (impact of commutes on happiness is heavily under-weighted)
Culture (specifically the density of the kinds of people I want to meet)
- What kind of people/lifestyles are accepted/respected? (eg SF bipolar attitude to startups, HK weights high-status jobs)
- What are the main industries?
- Are there good universities / research groups in the area?
- Do I already know people in the area?
- Is there an active climbing/parkour/rationality/tech community?
- Does it attract many RC alumni?
- What opportunities would I have to learn new subjects outside of software?
- Who would I be learning from? (eg is there any mentoring/training?)
- How masterful are the people I would be working with?
- How much time would I be actively devoting to learning? (eg via experiments, classes or training vs just learning on the job)
- How much choice would I have in what to work on? (eg do I need permission to switch projects, or to start a new project?)
- How much power would I have to make important decisions?
- How much would I be judged on results over process? (eg can I go home early if I'm not productive? can I take a walk in the park to think during work hours?)
- How often would I be able to take time off to travel?
- How much free time would I have for personal projects?
- How flexible are the hours?
Purpose (and rationality)
- What is the goal is this work?
- What are the odds of success?
- How is progress towards the goal being measured? (ie how do you know if this working?)
- What is the process for making decisions towards that goal? (eg is it based on evidence? is there any attempt to avoid bias?)
- Are there pressures towards / risks of instrumentally irrational decisions? (eg publish-or-perish, pointy-haired boss)
When answering these, I want to focus on the outside view and limit the impact of self-reporting, motivated recall and reference group effects. For example, rather than asking 'do people get to choose their own projects' I could ask 'of the people I met today, which of them are currently working on a project that they chose'. In theory this will lead to more accurate answers.
I'm not going to attempt to make some kind of scoring system - I'm using this structure mainly to ensure that I actually pay attention to all of the important areas.
Over-narrow search space
The search space has to be narrowed to be tractable, but I need to be careful not to make unnecessary assumptions.
- I am insistent on doing something that I believe to be useful. It doesn't have to be world-changing, just not taking part in a zero-sum game or being actively harmful. (I am willing to trade spending some of my time playing a zero-sum game (eg trading) if it helps with spending the remaining time usefully).
- I need exposure to fresh ideas and problems. I could afford to live in a cheap country for years without working, but I wouldn't do anything worthwhile in isolation.
- I've never managed to build a healthy social life when living in a non-english-speaking country, nor while living nomadically, so those are out.
- I'm wary of the loss of agency and certainty that comes from living on a visa - I won't rule it out entirely but I will focus my attention on options in the EU.
To avoid succumbing to inertia I need to actively investigate choices that are completely different from what I've done in the past.
- I could get a full-time or part-time job, go to school, go back to consulting, start a business etc.
- Learning to program is a sunk cost. The fact that I have that skill opens certain options and affects my chances of success in others, but I shouldn't be biased towards choices that involve programming just to justify the time invested.
So the search space is something useful, in an english-speaking area, most likely in the EU. Now I need to generate ideas.
- I will reach out to existing contacts for ideas, and also encourage anyone reading this to contact me with ideas.
- For tech companies, I'll talk to some of the more effective recruiters, especially RC. I can also look at news, journals, conferences, meetups etc to make a list of companies worth investigating.
- For jobs outside of tech, even programming jobs at non-tech-focused companies, I have much less idea how to proceed. Most job listing sites are awful and I don't have a good idea of what kind of non-programming jobs I could work towards. 80000 hours doesn't push me in any new directions.
- I'll list major US/EU universities and skim the list of research groups in each. If any are particularly interesting I'll try to contact someone already working there and find out more, and visit the department if possible.
- Both consulting and starting a business alone are likely to be fairly isolating. Working with a small group could work. I'll raise the idea with a few people and see if any promising ideas emerge.
- I have some tentative ideas for research projects which might have a good effort-reward trade-off and don't fit well into academia or industry. I will flesh out a couple of them to consider in combination with part-time work. I can also search for companies which do related work.
Having an explicit process for exploring options prevents motivated stopping and helps expose potential gaps, such as the fact that I don't have a starting point for finding out what non-tech jobs are open to me.
Stopping too early
The easiest to solve - I will publicly pre-commit to not making a decision until March 26th at the earliest. During that time I will continue to actively generate ideas, meet new people and visit companies/labs even if I already have a compelling option. (Some companies/universities might demand a decision sooner, but given the wealth of choices I think the risk of making a hasty decision outweighs the risk of exploding an offer).
On March 26th I'll post a condensed description of the process and any decision I may have reached.