Category Archives: Uncategorized

Cognition Is Social

Oddly enough, this stylized fact bears fruit on several different levels.

E pluribus unum

Intelligence emerges non-additively from the interaction of many less-intelligent agents. A group can have intelligence greater than any of its individual parts, or even their sum. Of course, the opposite can be true as well.

This phenomenon has many forms and names:

  1. Brian Eno’s idea of scenius: “an ecology of talent…supporting each other, looking at each other’s work, copying from each other, stealing ideas and contributing ideas.” (see takes from Kevin Kelly and Austin Kleon).
  2. Marvin Minsky’s Society of Mind modeled human intelligence as a society of agents competing for resources and power, and collaborating to achieve shared goals.
  3. Hayek’s point that computation is difficult and information scattered, so to solve the social optimization problem, our only hope is to use markets to aggregate many local decisions.
  4. The hive mind of twitter (the cool kids are calling it an “egregore”), the world of memes, of the tyranny of ideas.
  5. Individual minds often work better in a group setting. We respond to rewards, social rewards are powerful, and if you are rewarded for “good” thoughts then you generate more of them.

Counterpoint: “nine women can’t make a baby in one month.”

E unibus pluram

The human intelligence explosion can’t be explained by normal processes of natural selection. What new threats or opportunities emerged in the savannah to create such an intense selection pressure? Answer: other humans.

Runaway evolutionary processes are usually explained by an arms race of some sort. Humans are intensely social creatures, and in our evolutionary environment the most interesting thing is other people. Being able to anticipate what other people will do and incorporate it into your actions is an immense edge. All of our social actions (flattery, threats, flirtation, requests, manipulation) rely on having accurate mental models of other people. So success depends on two traits: the ease with which you can simulate other people, and the difficulty of being simulated. The arms race goes exponential because of the “coincidence” that these two are both grounded in “intelligence”.

You can tap into this with a simple technique: Decide who you would ask for help, and imagine what they’d say.

[epistemic status for this section: shakier than I’d like. I didn’t invent this idea, but I haven’t done the work to firmly ground it in the originating research.]


Find a scene, think in public, work with the garage door up, use your social imagination productively.

Stability and Play

I often think about how heavy my head is. When I let it rest on my hand, or lie down, I feel its weight and think about how much work my neck is doing to keep it upright throughout the day. The neck is, generally, an incredible system which keeps the head stable and allows for fully three-dimensional movement (pitch, roll, yaw) through a fairly wide range.

When I’m stressed or unhappy I usually feel that emotion to be physically located in the back of my neck, and sometimes treating the physical symptoms can alleviate the emotional distress. Which is to say in the normal course of events my neck is working hard.

The other day I was doing some yoga with a youtube video, and I got to a part of the session where I was told to move my head in circles to stretch out my neck. The instructor said something like “remember to keep your core activated”, I tightened up, and felt something completely new.

I became aware of a freedom in my neck that I had never felt before, a looseness that let me stretch farther, with less risk and worry, than I ever had. The muscles relaxed, the low-grade worry about keeping my head from toppling over dissipated, and I could playfully explore all kinds of new movements.

This all came about, of course, because I had given myself more physical stability with the rest of my body. This is a general pattern: there is a certain amount of stability necessary across every system, and stability in one place can ripple out, allowing other parts of the system to relax and play freely. Stability here is not static, but a control loop of sensing the outside world and making microadjustments to maintain whatever equilibrium is important.

Note that this isn’t a “nerve center” controlling the reaction of different components: stability within one component can trickle to the rest without direct control.

There are other kinds of stability we can think about, too.

Corporate research labs are kind of like this: Xerox PARC, Bell Labs, Google X. Only a company that is absolutely printing money, whose core business is rock-solid, can allow for such an unstructured approach to R&D. Here stability is basically cashflow; play is expensive.

Another is the advice given to startups that they shouldn’t try to innovate on corporate structure, because their core mission is already so difficult. The scarce resources here are time and attention for the people involved, but also some sense of risk. If your central idea is already likely to fail, you want as little risk in the rest of the system as possible.

Similarly, young writers often try to live bohemian lives, but are usually advised to build routine in their lives so they can be reckless in their art. There’s been an analogous idea floating around the techno-rationalist worlds that getting married early frees you to do greater career work, and I think the hypothesis is similar. (The econ blogosphere has been talking for a while about the marriage premium but the emphasis there is slightly different.)

Stability Mapping

When you are trying to understand, explore, or improve a system, why not try “stability mapping”? Which parts of the system generate stability, and which use it up or rely on it? Which kinds of stability are in surplus, and which are in short supply? How does stability percolate from one part of the system to the next?

Right now, organizations of every sort are reeling with VUCA (new term for me, see slide 14 here) . This analysis suggests one promising approach: identify the sources of stability that still remain and build around the kinds of stability they give you.

Fear of Criticism

When I was younger I hated to read any comments that teachers made on my work. It was manageable for math and science, where there was usually a right answer, but for any kind of free-form writing it was absolutely terrifying. When I got an essay back I would quickly look at the grade. If it was good enough, I would skim the written notes, scanning for praise, and trying to ignore as much as possible any ideas for improvements or notes on weaknesses. If it was bad, I would shove it into my backpack and never look at it again.

In school these assignments would be structured with submissions for intermediate outlines and drafts. And the purpose of this structure is to teach you, correctly, that even your best first effort can always be improved. But because I hated the process so much I would subvert it. I would usually just ignore any comments on the draft and rewrite it myself, or just re-submit the first draft as the final assignment.

Why was this so hard?

All my self esteem came from seeing myself as “smart”. I saw being smart as something static: you have it to a certain level, and it’s always the same. If I tried hard at something, and it wasn’t perfect, that deviation was because I wasn’t smart enough. Therefore the only meaningful output of doing well was praise, validating that I was smart, and the risk was discovering that I wasn’t as smart as I hoped or believed. This was terrifying, because it was the only self-worth I had, and I couldn’t imagine any other source of it.

I love, unironically, the Mos Def song Fear Not Of Man where he says (paraphrasing):

“You’re valuable, and not because you have a lot of money, or because someone thinks you’re sexy. You’re valuable because you were created by God.”

I don’t think you need to be religious to have a strong sense of self-worth, but the problem here is real and can really mess you up. It’s a bit hard to admit that this line is so important to me, because the underlying sentiment sounds so earnest and dumb. But in general, the most obvious truths sound stupid and are sometimes for that reason inaccessible to smart people.

Some helpful ideas

I’ve come a long way since I was younger, but this is still sometimes a challenge and I haven’t found a silver bullet.

First, a few quick tactics. These help me disengage from feedback, and allow me to see it as improving the work rather than reflecting on me as a human being:

  • Explicitly emphasize, to yourself and others, both before creating something and when asking for review, that done is better than perfect and you’ve prioritized getting something out there over getting it exactly right.
  • Use language like “I’m still tightening it up / playing with the structure / figuring out the right ordering”. Sometimes that’s even true.
  • Separate out the parts you want feedback on. “The language is still rough, I’m more looking for feedback on whether the core ideas make sense.”

One big change has come from seeing other people fail in the same way. In the workplace, I’ve come across a number of coworkers who send around a perfectly formatted, delightfully designed piece of work with fundamental errors in logic that several other people would have caught immediately in a draft or outline. So you can think of incorporating feedback early as a way to avoid wasting your own time.

This also became much easier to me when the feedback came from “the outside world” rather than “other people”. Working on my first startup, it became obvious that all the finely-crafted arguments in the world wouldn’t be as helpful as just running an experiment and looking at the data. Internalizing how useful that process was, and considering personal feedback to be an extension of it, went a long way. However, this situation is still in some ways closer to math homework than writing an essay. If I can put metrics on something, and think of it the feedback as “objective”, it’s easier to receive. Not everything in the world is amenable to this approach.

More broadly, it was motivating to adopt the frame of curiosity about the world rather than authoritative judgment on me and my work. Our language around this is very unhelpful. We use the term “feedback” to mean both “new information about an idea” and “another person’s judgment of whether your personality traits are helpful to your goals”. These two things are orders of magnitude apart in how much maturity they take to process, and usually you are dealing with the easier kind.

Another idea I’m exploring is that when you go to other people for feedback you are building a relationship with them. It’s like a grooming ritual for primates: by going to someone for help, you are building a stronger bond with them. You can then see their feedback as “an expression of care for you”. This can help disengage from the content of the feedback for long enough to read it and internalize it. This requires a relatively high level of trust and safety to work.

Unfortunately, caring a lot about the quality of a piece of work isn’t enough to break through this barrier. It can actually make it harder to receive feedback. The more tied up you are in whether something succeeds the harder it is to deal with critique, which might actually make you less likely to believe it will be successful. Alienation from your work can be helpful, if you can flip it back off when you need to. I suppose the ideal is to be 100% tied up in your work when you’re working, and then switch to an outside view where you and others think together about how it can be improved.

Two Theories of Japanese Culture

Two favorite theories from Japan and the Shackles of the Past:

  1. Japan never had a true Axial Age moment – the moment where Athenian philosophy, Buddhism, and Judaic monotheism (among others!) all began to separate the material world we live in from the “spiritual world” above. All human culture to that point was animistic – where spirituality and real gods imbued every aspect of the world and to talk of a separate spirit world didn’t make much sense at all. When Buddhism arrived in Japan it never fully co-opted the native Shinto animism. And it is precisely this “spirituality in everything” approach that characterizes the Japanese reverence for small design details, ceremonial acts, etc.
  2. Feudal Japan under the Tokugawa Shogunate (1600-1868) is widely known for the influence of the samurai warrior class and their spartan ethic of bushido. However, this was a peaceful period! In fact, the samurai were essentially useless at all for 250 years and they responded in two ways. First, they focused more than ever on an extreme version of bushido, seeking to one-up each other with asceticism and military technique. Second, though, they became decadent – noted for their patronage of everything from burlesque to prostitution — which developed customs of extravagant costume and theatrical presentation. It is precisely this dual nature that shocks one about Japan today – the ultra-serious business ethic coexisting easily with the hypersexualized otaku videogame culture, but it has a long history.

These are obviously oversimplifications and I barely know anything about Japan, certainly not enough to evaluate these two theories – but I enjoyed thinking about them.

Where are you from?

I’ve been thinking a lot lately about the question “where are you from”. It rankled me for a long time – in San Francisco when I first arrived it felt like a provocation, a way of checking if you were a native or a gentrifier. Then I made a friend who has spent a lot of time in Navajo nation building relationships with folks there. They ask the same question, apparently, because their relationship to the land itself is so much of a part of who they are, that knowing you grew up on a particular mountain or stream tells them about you in ways that go beyond “local culture”. So I started thinking about which parts of your background are important to get to know you? Which fields have you studied, who are your intellectual and ethical ancestors? Is there a “small talk” question that gets at that? “What did you major in” is probably close, for more intellectual folks who are still in or near college, but is still a pretty weak shadow of what I’m trying to get at. 
And this becomes even more relevant when you think about statelessness and movement as political issues. The 21st century doesn’t have a monopoly on this – human history is full of forced and voluntary migrations – but there are particular new spins on it. Cosmopolites (which I just read and is very good) talks about bidoons in the UAE, whose ancestors were nomadic desert people just like everyone else, but missed the citizenship initiatives of when those states were forming, and are now stateless – and the UAE’s efforts to buy them citizenships in Comoros, a poor island in the Indian Ocean. The most marginalized want citizenship as a way to secure their rights, to be legible to government and justice. Then of course the ultra rich are stateless in a different way – trying specifically to escape from government responsibilities like paying taxes, they collect residences and passports in the most convenient nations. Climate refugees may be one of the key stories of the 21st century and this question of “where are you from” may have a completely different meaning for them.


A very thoughtful recent blog post makes the point that institutions that seem “decentralized” or claim that as a value often exhibit centralizing tendencies over time. Some recommendations:

  • Be specific about what things you want to decentralize, and why. Regard decentralization as an ongoing process that can never be complete.
  • Find checks and balances, so that it is harder for any set of actors to achieve centralizing power. Use multiple forms of decentralization and participation.
  • Consider accountability: often what we really care about is accountability, and a centralized but accountable entity (such as a government antitrust enforcer) can limit the centralized and unaccountable power accumulation that we fear.

Pluribus Skepticism

Is Facebook’s new poker AI really the best in the world?

Facebook released a paper and blog post about a new AI called Pluribus that can beat human pros. The paper title (in Science!) calls it “superhuman”, and the popular media is using words like “unbeatable”.

But I think this is overblown.

If you look at the confidence intervals in the FB blog post above, you’ll see that while Pluribus was definitely better against the human pros on average, Linus Loeliger “was down 0.5 bb/100 (standard error of 1.0 bb/100).” The post also mentions that “Loeliger is considered by many to be the best player in the world at six-player no-limit Hold’em cash games.” Given that prior, and the data, I’d assign something like a 65-75% probability that Pluribus is actually better than Loeliger. That’s certainly impressive. But it’s not “superhuman”.

I don’t know enough about poker or the AIVAT technique they used for variation reduction to get much deeper into this. How do people quantify the skill difference across the pros now?

I’m also a bit skeptical about the compensation scheme that was adopted – if the human players were compensated for anything other than the exact inverse of the outcome metric they’re using, I’d find that shady – but the paper didn’t include those details.


Defensive Randomization

Machine learning is common and its use is growing. As time goes on, most of the options that you face in your life will be chosen by opaque algorithms that are optimizing for corporate profits. For example, the prices you see will be the highest price under which you’ll buy, as based on an enormous amount of data about you and your past decisions.

To counter these tendencies, I expect people to begin adopting “defensive randomization”, introducing noise into your decision-making and forcing corporate algorithms to experiment more broadly with the options they introduce to you. You could do this by simple coin flip, or introduce your own bots that make random (or targeted exploratory) decisions on your behalf. For example, you could have a bot log in to your Netflix account and search for a bunch of movies that are far away from Netflix’s recommendations for you.

One possible future is for these bots to share data between themselves — a guerilla network of computation that is reverse-engineering corporate algorithms and feeding them the information that will make your life more humane.

This is related to:

[mildly inspired by Maximilian Kasy’s Politics of Machine Learning]


Police Science

Very much enjoying Jackie Wang’s Carceral Capitalism.

Especially liked this thought in “This Is A Story About Nerds and Cops“:

Given that critics of the police associate law enforcement with the arbitrary use of force, racial domination, and the discretionary power to make decisions about who will live and who will die, the rebranding of policing in a way that foregrounds statistical impersonality and symbolically removes the agency of individual officers is a clever way to cast police activity as neutral, unbiased, and rational.

Contributing to pandas

Very proud to announce today that I had a pull request merged into the pandas library. In version 0.21, pandas will have a new feature: a way to read in line-delimited JSON in small pieces, which can be useful when working with large files or streams.

This is a fairly small change, technically, but a big deal for me.  Pandas is one of the most commonly used tools in the data science world. When I started at TrueAccord they bought me the book on pandas (Volume 2 coming out next month!). This was my first introduction to any programming language other than Stata, an odd proprietary language that languishes on among economists and epidemiologists. Now, writing software is a core part of my career.

Related, I highly recommend  The Success of Open Source, in which Steven Weber outlines the varied ways in which open source communities elicit and channel cooperation, and explores the complex set of motivations that leads people to contribute to open source.