One Year Review:
So it’s been about a year since I started writing blog posts on rationality. Throughout this mental journey, I’ve changed my mind about a lot of things, and sometimes even flip-flopped on some concepts. In this concluding post, I’ll try to show the development of my ideas. Lastly, I’ll focus on some concrete benefits MLU has brought me, and where I think I’ll head for 2017.
(I’ll be also be doing a writeup of my current thoughts on rationality in a much clearer and shorter post, as this post is longer than usual.)
Part 1/6: Starting Out
I began the year with the concepts of meta-levels and precommitment as my preeminent tools. I had mantras like “Optimizing doesn’t always feel optimal, but you know this so power through regardless!”, which focused on trying to anchor myself to the values held by my past self.
Godel, Escher, Bach had just blown my mind a few months earlier, and everything looked like a nail for my meta-level hammer. Most of my thoughts anchored around “Going Meta”, a term I used for popping out of situations and noticing oneself in the moment. It’s basically glorified self-awareness:
- Saying to yourself, “I notice I am going to eat a cookie,” before I actually eat the cookies, to give yourself time to assert executive control.
- Noticing you are in class and asking yourself, “What have I learned so far?” rather than just following along with the lesson.
- Realizing you are in a conversation that is derailing and thinking, “What is the main point of this tangential discussion again? Is it even important?” rather than getting bogged down in the object-level discussion.
I thought this would solve everything.
Around the same time, I’d started reading Nate Soares’ blog Minding Our Way. His post on desperation struck a very powerful chord with me. I began thinking about blind spots and the absurdity heuristic. “Just how far would we be willing to go, to get something done?” I wondered, “Even if the idea seemed ludicrous?”
Filled with a desire to OPTIMIZE LITERALLY EVERYTHING, I tapped into a well of willpower and tried to power through my tasks. Using sheer force of will, I thought it would be possible to get anything done.
That… didn’t work as well as I thought.
Part 2/6: New Ideas
From there, I read more of Nate’s blog, and I encountered his revolutionary idea of “dropping the whip”. It might be possible, I saw, to be productive and like it. I could perhaps just become the type of person who likes doing work! Then I’d just do work! And like it!
At the time, I had started to burn out from my stint with pure willpower. I took up a more consistency-based approach, where I’d mutter to myself, “Make Productivity a Habit!” The reasoning there was that the more I did work, the more of a work-doing person I’d become, which meant that, as a work-doing person, I’d easily more work, etc. Positive feedback loops and habits and all that good stuff.
I also started thinking about distractions in terms of sunk-costs (which still seems really insightful). Recursive meta-awareness was another interesting idea: I know that just knowing about biases isn’t enough to help me. But is knowing that enough? What if I go one more level up? Or one more…?
Also, CFAR’s Propagating Urges was an interesting concept that I’d interpreted to mean basically “positive brainwashing”. Through some initially forced positive affect (smiling while doing work, good music, etc.) I managed to get myself much more engaged in math and programming.
Part 3/6: Creating Skills
As my thoughts on fighting procrastination and staying focused began to coalesce into something more clear, I started looking at things in terms of techniques rather than concepts.
Giving these mental notions quaint names, I started to view rationality as a unified toolkit.
I think it’s here that the idea of a mental action really started to cement. It was less about using psychological research as an intuition pump to precommit, but more about finding instructions that could be adapted to different situations.
I’d found my first mental “algorithms”, where it was about executing a certain maneuver in my mind—asking myself a certain question, doing a particular action, looking at things a different way, etc.
Examples of techniques I’d written down (* = I no longer endorse this):
- * “Celebrate Small Victories”: Use small gestures like a fist pump after completing tasks to create your own positive reinforcement.
- * “Generalized Action”: When overwhelmed with the specific situation, abstract from it, and then do the thing that seems best for the entire class of situations, of which yours is only an instance. (EX: If you’re nervous because you’ve just met Elon Musk, just do what you’d do if you met any investor/innovator, instead of worrying about your hair.)
- “Inner Simulator”: Ask yourself “What would X do?” when you don’t know what to do, but X is someone in your mind who would know what to do. [Misinterpreted from a CFAR technique of the same name].
- “Optimize Your Mindstate:” If you’re feeling in a “not-work” state, do a thing that is low-effort but puts you in a more of a “yes-work” mode, so you don’t have to immediately force yourself. EX: Walking outside, stretching, getting water, etc.
These skills made some things easier, and I updated my viewpoint again. But this was still on the conscious level; I had to remember the tools to use them. I’d yet to think about really integrating them into my workflow.
Part 4/6: More New Ideas
My mental models received a significant boost after EuroSPARC and EA Global, where I considered things at greater length than before. Maybe I would have bumped into these ideas on my own, but these were great opportunities for me to just sit and think about things.
In particular, learning about CFAR’s Trigger Action Plans got me thinking more explicitly. I found I had an aversion to making detailed plans, and it turned out I had a gut feeling that if I wrote a plan down in detail, I would “have to” do it.
This reinforced my idea of “actually sitting down and getting something done”. Earlier, I had spent some focused cycles trying to chart my overall map of rationality-related concepts, hoping to find some overlying themes.
The best I could get was to cast most of my techniques as “reframes” that served to either trigger different modes of thinking, or change how I viewed the situation. And because the whole point of the techniques was to do things, the whole thing could be interpreted as a roundabout way of saying “stuff that happens in your head influences reality”.
Kind of like a bastardized version of “mind over matter”.
Other ideas that have proven useful:
- Cross-reference ideas by relating other things to them so they don’t become mere free-floating nodes. The deeper that something is embedded with your other mental networks, the easier it will be to access later on.
- Examples are fantastic. Always look for opportunities to ask for examples, or provide examples. It’s way to easy to get bogged down with abstract language or the super high-level picture.
- Reference class forecasting is one of the most powerful prediction tools we have. Adapting it to personal endeavors yields realistic estimates. Rescale your internal expectations, and don’t expect progress immediately.
- Procrastination and other poor habits are addiction cycles, reinforced by temporary alleviation. Breaking out of them is difficult, because hyperbolic discounting and preference reversal are a thing.
All the information felt like an overload in a lot of ways; there was simply too much to internalize. I’ve tried to focus only on the ones that are the most immediately helpful, but there are other avenues I haven’t explored, like many of Leverage Research’s ideas on building models.
Part 5/6: Present Thoughts
It became clear that having techniques isn’t enough. Over and over, I saw that remembering to use the techniques proved to be the biggest hurdle. At first, I thought creating more explicit algorithms was the solution.
Looking deeper in the mechanisms behind these techniques, I tried to create detailed step-by-step algorithms for two techniques. But that didn’t solve the problem; they still kept slipping out of my mind.
The obvious answer was to integrate them as habits. I didn’t see it that easily, though. Habits felt like they were part of the self-consistency school of thought, where rationality was simply “a thing you do”. I felt like the paradigm of using these algorithms was at odds with habituation. Wasn’t relying on techniques to do things the exact opposite of “just” doing them?
The answer I’ve come to now is that these two categories are best used to tackle two different types of irrationality, which resolves the apparent contradiction nicely:
- Akrasia is often a disconnect between multiple desires. It leads to unfortunate situations like the “action-intention” gap, where what you do doesn’t match up with what you (explicitly) want. Self-consistency and resolving internal disagreements are great for pushing this kind of problem aside.
- Real life has many situations where it’s necessary to make good choices— remembering to bring your coat, deciding whether or not to purchase Ikea furniture, etc. Many of these come in the form of split-second opportunities. Focusing on habituation and the 5-second-level techniques can make it a lot easier to optimize in these scenarios.
There are a few other things, like where I think are promising fields to integrate into rationality, but I’ll save those for the specific write-up.
Part 6/6: Evaluation / Future Plans
What were the main benefits of MLU for me?
1) Writing Strength:
Altogether, I’ve written some 30,000+ words this year. Although word count does not seem to be a good indicator of quality, I do feel like my writing has improved over this time. Writing quality isn’t well measured, but I can confidently say many of my earliest essays are much worse than my later ones.
Anecdote: When I first started writing these, I cross-posted an essay to LessWrong in January, where it attracted a lot of criticism. Looking back, all the points were quite valid; my ideas were all over the place.
Also, it was terrible.
2) Compressed Ideas:
Having gone through the actual process of writing these essays, I can shuffle around complex ideas in my head with labels like “Optimize Your Mindstate” without having to unpack the abstraction. Also, writing these ideas down makes it easier to cache. It feels like an extension of my thoughts that’s neatly tagged with the MLU label in my head.
Counterfactuals are always hard to evaluate, but I think that had I not stuck with a regular update schedule for MLU, I would be far less self-aware about the mental things that surround trying to implement rationality.
Those seem like the two biggest ones. There were probably a lot of smaller things, like the beneficial outreach provided by essays like Rationality 101, but I honestly didn’t focus on outreach this year aside from cross-posting to Facebook.
What started out as an ambitious project to write helpful intro rationality articles turned into something far more exploratory and introspective. Moving forward, though, I’d like to write more things like Rationality 101, which introduce rationality concepts in a clear way to new people.
For 2017, I’ll be focusing a lot more on:
- Writing / Doing habit formation things (lots more about this coming soon).
- Giving examples in my writing.
- Increasing readability (subheadings and pictures).
I’ll be maintaining the same weekly Friday release schedule for posts in 2017. See you then!