Ulysse Pence

Life As Science

Introduction

October 30th, 2018

We have formal methodologies for testing hypotheses about universal fundamentals and harnessing subsequent discoveries for practical application, yet we rarely apply any rigour to the science of ourselves: “What am I like?”, “What can I do?” “What resources and properties can I exploit?”

My life is a science. I seek deeper knowledge about the underlying rules and truths of myself and what I am capable of. In this pursuit, I employ a practical framework for experimenting on myself and moving towards desirable outcomes.

The subjects of experimentation can be anything about myself. It ranges from diet and health to riding electric scooters to settling on a reading schedule to internalizing useful heuristics.

My most impactful experiments so far:

Some of my stranger experiments:

How does it work? I will describe the inspiration for the framework and then the details of how it fits into my workflow. If you are only interested in the practical details, skip the next section.

Origin story

I attended a rationality workshop in early 2018 that introduced me to several new concepts and techniques for better understanding the world and getting stuff done.

One of the concepts that resonated with me was “Copernicus and Chaos”, which introduced me to the exploit-explore dichotomy of actions you can take to get what you want. Generally, “exploit” actions use knowledge and skills you already have to produce value, while “explore” actions surface new ideas and activities that may be better than what you already know, but don't produce any value when you do them. I would guess that the actions of most adults (including myself), could be labeled “exploit” in the scope of my life.

“Copernicus and Chaos” also suggests that, without more information, it's usually safe to assume that an unknown quantity falls near the median. Combined with the “explore” / “exploit” ideas, these two principles form the Chaos Heuristic:

Odds are, you're nowhere near [your potential], and you don't know what you don't know. So go exploring!

This heuristic inspired me to start integrating the rationality techniques I'd learned in the workshop into my daily routines; however, at the time, I was reading work by Charlie Munger and Eliezer Yudkowsky, so I shortly began reconsidering and “attempting to falsify” my beliefs, aversions, and preferrences. Every morning, I prompted myself:

These questions begot many ideas for changes I could try. For example:

This went on for several months and expanded in scope, until it felt like the process was too ad hoc and inconsistent. So I formalized the best parts of my improvised system and created an experimentation framework to track everything.

Experimentation Framework

Lab notebook

What would an experiment be without a lab notebook!? For each new experiment, I include the following items:

Here are some of the methods I use:

Experiment Lifecycle

Each experiment starts out in the “Conducting” phase (tracking which phase each experiment is in is essentially a Kanban board). I fill out the experiment lab notebook's context, outcome, and methods sections and then add it to the “Conducting” phase. I also schedule an “evaluation” date in the future where I will seriously consider if the experiment has converged on its outcome. Depending on the experiment, the evaluation might be a couple weeks in the future or several months.

Every week until the evaluation, I spend about 20 minutes reading through my notes (which are timestamped) on each “Conducting” experiment and tweak the experiment methods as needed with the hope that it will make the experiment converge on the desired outcome faster.

Finally, during the evaluation, if the experiment is not doing well, I must decide whether to throw it away or totally revamp it and set a new evaluation date. If the experiment has largely converged on the outcome, I mark it as “Full-Time” and review it every month or so.

Examples

Pomodoro cycles

I used The Pomodoro Technique in the past to break up tasks into manageable 25 minute chunks. On an older computer, I had a program that reminded me to go walk around for a few minutes every hour. Both of these things were beneficial to me and I would like to do them again. This is the context.

The outcome I wanted was to (1) break up larger tasks into several pomodoro work sessions and (2) after each pomodoro, to perform a quick checklist of things. For now, the only item on the checklist would be to get up and walk around for 1-2 minutes.

The initial methods I devised for this were specific to my setup. I created 2 TAPs (TAPs are referenced above):

That was pretty much it. It only took a few repetitions to make this a habit. However, I noted that when I went to walk around, I would often think of a household chore that had to get done and wander off to do it, which was undesirable. During the next review for this experiment, I changed my “walk around” checklist item to specify that I would stay near my desk.

After 2 weeks, my calendar reminded me that it was time to evaluate the experiment. Looking back at my notes and reviewing my memory, I felt like I had converged on the value I wanted out of this experiment, so I marked it “Full-Time”.

Course correction

Improving my decision making and processes for getting stuff done is an essential part of growth. I noticed that there was occasionally a pattern to certain bad situations and failed projects. This is the context.

The outcome I wanted was to have a feedback loop to notice these problems more explicitly.

The initial method I created to accomplish this was to use a decision journal:

This experiment is ongoing and I've only recorded 2 journal entries thus far, so I haven't been able to perform a meaningful review or evaluation.

Maintenance day

In “The Productivity Project”, Chris Bailey suggests doing a lot of “maintenance tasks” (e.g. cleaning, calling people, buying groceries) together on the same day. I liked this idea, since I had found it disruptive to do many of these tasks interspersed among my other tasks. This is the context.

The outcome I wanted was to perform weekly maintenance tasks on the same day and prefer this to doing them throughout the week.

In order to implement a Maintenance Day, my methods were pretty simple:

I noted after a couple weeks that some I often ran out of time for some tasks. Additionally, while I preferred doing house-related tasks on the same day, making calls (e.g. doctor's office, DMV/RMV, bank) on the same day wore me out. During the next review for this experiment, I removed the calls from Maintenance Day. Additionally, I stopped scheduling some tasks specifically for Monday and, instead, just made sure they got done by the end of the week.

Two months later, during the experiment's evaluation, I decided to retire the experiment because I believed regularly dedicating a whole day to low-impact tasks was ultimately wasteful and it seemed only marginally preferrable to doing the tasks throughout the week.

Specifics

Whichever tools you think will work best for implementing the experimentation framework probably will work best, but I want to describe a basic setup. According to my description above, the key features you need are:

  1. A place to store lab notebooks. Since each lab notebook grows, a physical notebook might work, but a digital one makes a lot more sense e.g. a file directory with separate files or a notes program.
  2. A central status board to track what the status of each experiment is (i.e. “Conducting”, “Full-Time”, “Retired”). A physical kanban board, an issue tracking service like Trello, or a simple text document all work fine here.
  3. (optional) A place to write down future experiment ideas.
  4. A calendar or agenda program to schedule regular reviews and experiment evaluations.

In my case, I use org-mode, which is a collection of outlining / agenda features in the Emacs text-editing program. Key features include collapsing and expanding list items in the outline, scheduling specific list items into a centralized agenda, and linking from one file to another like on Wikipedia.

In my central “Experiments.org” file (#2), I track the state of each experiment and keep a list of future experiments (#3). Each experiment in the central experiments file links to its full lab notebook (#1).

When I start a new experiment, I copy a template lab notebook which has all the notebook sections described above (outcome, methods, reviews, notes) and save it as a new one. Next, I fill in the outcome and methods. Finally, I link to this new lab notebook from the “Conducting” section of my central experiments file and schedule the experiment evaluation for some date in the future.

I have a weekly reminder to review experiments that I'm “Conducting” and a monthly reminder to review experiments that are “Full-Time”.

To hear more, subscribe to my newsletter or follow me on Twitter.