All Self-Models, All The Time

Hello everyone :).

FINALLY, BLOG TIME. It’s been three weeks. There will be no organization—it’ll just be a series of rambles that mostly flow into each other and then abruptly switch as I hit a “okay, time to reference my notes” moment. Here we go!

Hm :). I recently finished a long and intense social experience that has been occupying much of my attention. It’s been a fantastic learning experience, and I’ve learned a prodigious amount of information about how I think, how I interact with people, how to manage my emotions, and how others think. (I also learned a lot of other interesting more-academic things, but I’m focusing on the introspectively-flavored ideas here.)

There are a lot of observations to be made, and I think they’ll continue filtering through in blog posts to come, but I’ll just start with a few that came up. First, I manage myself in a lot of different ways—I like to have control over my own emotions—but there is a definite Way To Manage Monica that other skilled users employ. I was just talking to my mother about this, and she was telling me some of the strategies she’s had to employ since I was young.

Apparently, I’ve always been pretty narrowly goal-driven. In essence, I don’t have problems with motivation, because once I’ve decided I’m going to do something I’m going to do it, which means pushing myself through scary things and bugging the people around me until they also make things happen. These goals have to be pretty clear, and once I find what I believe to be the most efficient path to get there, I will not be deterred.

This caused problems when I was younger, because what I believed to be the most efficient path was not usually the most efficient path. Since receiving that feedback, I’ve tried to get around it by asking others around me for what I should be doing. Now when I’m trying to figure out how to achieve a goal, I’ll ask around to figure out what everyone else thinks is the most efficient path, and that path will get pushed to the forefront. Crowdsourcing decisions, works great.

In order to crowdsource decisions like this effectively, what you want is to get a bunch of people around you who know more than you do and also care about you. This is… surprisingly difficult. It seems to have some very interesting network dynamics, in that there are a lot of clusters of people in which everyone knows each other, but the transitions between clusters are not very strong. Upon coming to Berkeley, I got very lucky, and came upon some key people with regards to this network business. One person approves of me and has an excellent network and connections, and their endorsement has meant that I now have access to more than ten people who all are great sources of information, friendship, and idea-bouncing. One person cares about me, is very knowledgeable about a culture I’m interested in, and is great at teaching—so while their network doesn’t fully overlap with where I’m trying to go, they’ve been fantastic with introducing me to the relevant ideas and my personal growth. And the last person cares about me, has this gigantic network that they’re very generous in sharing people across, and has basically been handing me opportunities for connecting whenever I’m interested. (And also introducing relevant ideas and personal growth—many thanks; you know who you are :)).

Networking like this has been so interesting, because I normally build up social networks that are aimed at friendship only, rather than networks that contain friendship / work / personal interests / research relationships. This is the proper bubble-type material, because I can see why people find a cluster like this, which hits a lot of different needs, and then don’t really go outside the cluster. My current clusters are pretty separate by virtue of my environments—but here I’m encountering people with similar research interests and life philosophies, and it’s super fun.

Besides providing a stimulating and wonderful group of people to hang out with, I rely on networks to keep me on track in goal setting and goal attainment. I do a lot of feedback checking, implicit and explicit, to make sure that I’m working efficiently towards useful goals. My mother has recently suggested that a possible next step forward for me is to try to keep myself on track, and generate ideas, without needing feedback.

Not that feedback isn’t useful and essential, but she suggests that I might be able to simulate what to do, and what is and isn’t a good idea, using information I already have, rather than asking it from others. One way that she suggests doing this, that I think might work for me, is to visualize problems from other people’s perspectives: to instead of constantly observing what I’m doing and what I’m thinking, to try to observe what other people are doing, the rationale behind their processes, and emulate. She suggests this as a good idea in the research realm, and also socially, because simulating what individuals would do rather than applying a general rule to the situation (which is what I usually do) will lead to better predictions and results.

To me, this is suggesting two possible next steps. 1) Spend less time thinking about your own processes, and try to model other people’s and incorporate. (My mother: this has the additional possible benefit about being less defensive about your own strategies, since you’re not really evaluating your own, you’re just incorporating.) 2) Abstract rules are great and all, but you need to insert some more individual-level models when you interact with people.

The latter is more problematic than the former, because it involves a strictly increased amount of computational effort. General rules that apply to situations are important. One needs to characterize a situation, analyze previous efforts in that situation, simulate future results that could arise, and then use that information to inform actions in the current situation. I do this sort of situational analysis offline and cache these sorts of general rules for future use, so my decisions in real-time are reasonably quick.

So I can’t cut back on general rule creation for situational modeling. But now it is suggested that I bias my general rule in accordance with my models of individuals. It has to be done at the same speed to be effective in practice, because conversations take place very quickly. This means, essentially, that in addition to general situational modeling, I have to have cached individual models of people. And individual models of people are hard.

I recently executed my most effortful modeling attempt of a person—one that took far more focused attention than I’ve ever tried before. Usually when I know someone well enough to predict their opinions, it’s because I’ve hung around them long enough that I’m implicitly picked things up. This is talking about observations about attitudes over years. In this case, motivated by the goal of making myself feel better, I sat down a few times for more than ten hours total over a couple of weeks, and went: okay, what was their reasoning?

Most of that time was sorting out what I was feeling enough to be able to isolate rational thoughts. This was using my usual techniques: all right, Monica, what are you feeling, why are you feeling it, what was the cause, what was the other person’s reasoning, what was the situation and how do you need to react next time. However, this was a situation in which the other person was complex. It’s always annoying when someone does very nice things and also things I also very much don’t like, and they’re the SAME PERSON, and I can’t classify them neatly and they take SO much work. (I feel like I’ve complained about this before. Greys are hard.)

In this case, this is a very nice person who did something I didn’t like, who gave me all of the relevant implicit information but not all of the relevant explicit information, and so I needed to do inference over them and figure out what the heck they were thinking. I knew they’d only act in accordance with a strict moral code that they apply to life in general, they’re very thoughtful and careful, and they prefer honesty over all values except the moral dimension. Thus, all self-interest, information I had, and omissions of information would have to fit within this framework.

It took forever. Again, it would have gone much faster if I hadn’t been trying to shove my emotions around while I was doing it (“Is it fair to feel this? Is it useful to feel this?”). (By the way, I know that I’m allowed to feel whatever emotions I want, and that my emotions are valid. That’s the first thing I do whenever I go into one of these sessions—acknowledge what the feelings are, why I’m having them, and why it’s reasonable that I’m feeling this way. It’s just that after that bit I launch into the fair and usefulness questions, because we feel all sorts of things that we really shouldn’t be feeling if we want to be treated how we’d like to be treated by others, so it’s important to me to try to be fair in what emotions I have about people so that people will try to be fair in what emotions they have about me. Also, I like feeling happy, and assuming the best about the people I choose to spend time with is both confirmation bias and also helps on the happiness dimension.) But having strong emotions is also useful in a sense, because you have to justify hard, believe really strongly to have a good reason to shut those feelings down.

I got it done in the end, and was to the point where I was like: okay, we’re going to pretend we’re this person now. I’m going to pull up a series of written interactions we’ve had, and pretend I’m this person, and try to figure out what I was feeling when I wrote them.

And you know what? My new model of them was thinking COMPLETELY different things than I had thought they were when I was reading those messages. It was a bizarre experience, because I’d had certain feelings attached to certain messages, and guesses about the motivations behind them, but when I was doing this exercise my new guesses were colored with a completely different valence. A valence that fit better with what I knew about this person, that integrated my perspective on this person across time much better than I had been before. I had had a model that I’d developed implicitly. Then that model started performing badly, so I developed a new model explicitly. Both probably took the same amount of effort, but one was spread out over time and didn’t require focused attention. Explicit models are hard. And I had a handicap, given that I had a huge set of constraints from this person. They would act according to a set of values that I knew about, any self-interest or omissions would have to fit within this framework, they had to be able to rationally justify it, and outside of any personal complications they meant the best for me. How crazy it is that I had a hard time doing inference over someone who has laid out that many concrete, 100% rules for me to follow?

I’m told that I can at least cite the fact that they’re a complex person. I hadn’t been thinking of it that way, but I guess it’s easier to try to guess at someone else’s thoughts when they haven’t already run it through their own brain and shoved it around until it made sense to them. (I was told that this is sometimes a problem with me too—that you can’t fully trust me because I can make myself believe anything. This might be true to some extent, but I don’t really believe this. I won’t say things unless I believe them to some degree, and I do push emotions around to back certain ideas, but I can’t convince myself of everything. I base my rationalizations on how I feel afterwards, and things that don’t feel true on a gut level (implicit level) don’t feel right, and they don’t stick. If I’m starting to feel breakthrough of emotions that I thought I had sorted, I’ll know I didn’t convince myself of the right thing and I’ll try again until it works.)

(It occurs to me that it’s funny that I use emotions as a cue for regulating emotions. But… it works something like this, I think: I can generate positive emotions out of a situation when I understand someone’s perspective and can take a solid guess about how to solve the situation in the future. Negative emotions come about when I don’t understand what happened, or I think the person didn’t have a justified reason to make me feel bad, in that I thought they were capable of controlling whatever negativity they were expressing towards me and chose to not do so. If I can come up with a good enough rationalization for their behavior—if I can figure out what they were struggling with that caused them to do something I perceived negatively—that overcomes whatever I was feeling, then that does seem, in some sense, to be the best perspective. Because people generally mean well. This is especially true of the people I hang out with, who have given me so much evidence that they mean well. And any perspective that intuitively feels like it fits in with what I know about them, that feels like it’d be a valid struggle, that indicates that they were trying, when I know they’re the sort of person who tries—that’s a reformalization that feels right. We have these implicit models of people that we build up, and they’re usually pretty accurate, and when you can line up weird could-be-perceived-as-hurtful actions with that general model, then that’s arguably a better model of the person with more integrated evidence. Of course, I’m super biased right now, because I like to think the best of friends, acting in this sort of biased way makes pretty much every interaction better even if you’re giving people too much credit, it makes me very uncomfortable to be rude to people on purpose, I have a rule about honesty so I have to have tactful things that I believe are true in my head lined up, almost always when I’m mean I’m doing it accidentally and it sucks when people don’t take that into account, confirmation bias because I choose the people I spend time with, and I really like positive emotions. So maybe reframing things like this isn’t the “right” or true perspective, the perspective that fits the evidence the best, and it’s just a perspective I like because of all of the above reasons. I don’t know. Thoughts? It is just useful / nice to try to think why people do the things they do, or is it actually the perspective that’s most predictive?)

(One final tangent: I was also thinking that my preference towards honesty—meaning that I can’t say anything that I do not believe in any sense—means that I have to have my head in order more than most people. I also strongly value honesty and kindness to others, but unlike the person described above, my ordering is flipped, so I’d rather give a fuller picture of what I feel than omit information to keep from hurting someone. (This might just be computational cost. I don’t like trying to keep track of who knows what, and if everyone knows the same thing that’s easier on my memory.) This means, however, that since I want to not hurt people, I have to say positive truthful things, which means I better have some positive truthful things readily available. I think that if you ran your life where you could make something positive up on the spot, instead of genuinely having those positive feelings, there’d be less need for all of the unnecessary shenanigans that I do.)

Anyhow, I blame the fact that it took so long for me to sort this person’s model out mainly on the fact that I was having emotions, and kept on having to deal with those objections and settle them one by one. It actually makes sense to me that it’d be good to think about people more deeply, develop individualized models, before anything unbalanced me. It seems like a lot of work, and like it could result in some more negative uncoverings of personality than how I normally visualize people. Both of which I’m not a huge fan of, but then again, I’m supposed to continue pushing forward on my reluctance to learn information that I don’t want to know.

(This has recently been coming up a lot in my head—the feeling that “I don’t want to know this information because I’ll have to update the darn model again, it’ll set of my emotions and then I’ll have to deal with those again, and is this really worth it?” But hey, this kind of attitude leads to bad things down the road. It’s the cause of a lot of bad science, it’s a truism (can’t hide your head in the sand), and it’s important to want to know the things that are hard to know. It makes us keep from making mistakes in the long run, it gives us more complete perspective, it gives us more accurate predictions, it prevents me from becoming a kind of person I don’t want to become, and it’s important. I just have to keep on remembering that.)

So maybe I can spend some time thinking about people and caching those models. It’s certainly a good trait to develop if I want to make people share my beliefs in the future. I think some people can do it intuitively—know what you care about, and frame conversations in that way—but hey, I’m me, and how I do things that aren’t intuitive is to plan them to absurd extents. Definitely could be worth investing the energy.

Ha! It occurs to me that I actually was independently talking to two friends about modeling individuals. One of them says she’s really good at hurting people when she’s angry—can find true things about them and throw them back. I asked the two of them what would hurt me most, and they did a decent job. One of the things they said: “You say that a lot of your social skills are learned and applied—I think it’d hurt you if someone said you were being fake.” Another one: “You analyze emotional situations very distantly—if people said you weren’t compassionate.” And then they got a really good one: “If you weren’t saying smart enough things.” They didn’t mention this one, but it’s even better: if people I want to be interesting to say I’m not interesting.

Treating these one by one: first, social skills and being perceived as fake. This one wouldn’t hurt very much, because I’ve actually been told by a lot of people that the thing they like most about me is that I’m “genuine” or “sincere” or “honest” or otherwise come across as non-deceitful. It doesn’t bother me that my social skills are learned either, or that I think of social situations as performance, because I don’t think there’s a “true self” I’m aiming to emulate, and life is about learning how best to incorporate your tendencies and preferences with the skills that are most helpful in life. It’d also be quite weird to me if someone thought I was being fake, because even when I’m disagreeing with someone politely, I think I’m pretty blatantly obvious. (This happened to me recently, when I forgot to control what my face was doing. I was like, uh huh, I get your point, I don’t agree with your point, let’s move on, and they were like: I can totally see you don’t agree with my point so I’m just going to keep trying to convince you, and at some point one of my friends said: everyone look at Monica’s face, and I thought: oh dear, that wasn’t managed well.) (I’m never quite sure what to do in those situations. If I’m genuinely interested, I’ll argue back if I disagree, but I feel like I should be politer than I am when I don’t and want the conversation to just move on. This is one of those politeness vs. honesty trade-offs that I haven’t figured out.)

Second point: not being compassionate. Definitely used to be an insecurity, no longer an insecurity due to the dutiful efforts of my friend Tiffany. Also, I see it in myself: I genuinely do care about my friends and try to do my best by them. Processing emotional problems abstractly has little relation to that, though I thought they were more linked at the time, and the way I analyze things does tend to throw people off.

Third point: not saying intelligent things! Was an insecurity until very recently, flared up upon arriving at Berkeley, is being dealt with very well by Smitha and Carson (Carson’s launched this campaign against it, it’s great), and is now less of a problem than it’s been ever in my life. This is also not really an issue unless you happen to be one of the people who I’m trying to be intelligent around, so there’s a limited number of people who could throw this one effectively. Otherwise I’ll just take it as a “you’re communicating badly” critique, which is fine, and that’s something I’m aware of.

Interestingness—still working on this one. You’re interesting if you’re working on interesting work, so I’m pretty much guaranteed as long as I continue doing cool stuff. A lot of this is person-dependent too, because different people like different topics and you want the topics that you’re interested in to line up with what other people like to hear about. But this is pretty much out of my control. In any sort of long-term relationship, I’m going to talk about what I find interesting, and analyze situations how I always analyze them, and that’s not going to change much. If people like it and think it’s interesting then that’s great, and if they don’t that’s fine too. I’m thinking of this as sort of an: you can’t really do anything about it except continue to do things you yourself find interesting and cool, and being excited about what you do.

(There are other insecurities, I’m sure, and I’m of course not a big fan of being criticized, but constructive criticism is all about pushing through that roadblock of not wanting to know information when we really should want to know all of the information to make better decisions in the future. So please deliver criticism! I can always choose how to evaluate it. I feel like a lot of the time I’ll have reasons for what I do, or it’s too much work to change, but awareness is the first step towards progress that needs to be made.)

Right. I’m now going to track back to a point my mother was making about goals and feedback. I’m a big fan of feedback now. A lot of the rationalization above was based on feedback. But here was my mom’s point: as you continue on in life, your circles for effective feedback will get smaller and smaller. And it’s good practice to learn how to think through arguments on your own, when you already have all of the existing information.

True point. I should be able to logic through lots of things without needing external feedback, though external feedback is great and should always be collected when it is willingly offered. I don’t know, though. How I deal with this right now is to place a lot of value on social networks, so that I have a distributed group of people who care about me and whose opinions I value, and who will have time for me by virtue of there being a lot of them. This works for social situations, though I don’t know how many crazy social situations will be totally new to me in the future (still many, I’m sure, but the rate is decreasing). It doesn’t really work in science, when eventually I’ll be responsible for coming up with ideas entirely on my own. It’s not even that far off—I’m currently getting a lot of help on my research projects in grad school, but I’ll need to figure out my own things pretty soon. But to solve this problem is to solve the general problem of how to come up with good ideas. And the answer to that, as I’m already working towards, is: know more and think better. And I can’t rush that, or make a goal and work in a clear way towards that. Thinking better about research is the sort of thing that’s a process that’s learned from practice and observation, and everyone knows it’s the most important thing in grad school, so I don’t need to make any sort of special effort towards that, I don’t think.

Last track back! Way back in the beginning I mentioned that there are certain Ways To Manage Monica. All of my advisors pick it up, actually, and I just firsthand observed a newcomer do their best to figure out how to manage me. In retrospect—and in normal-spect to the people who know me—that process was incredibly amusing. Here goes.

1) Zero ambiguity. I am apparently terrible at shades of grey. I like clear goals, with clear ways to accomplish them, am inflexible with different approaches, do not think about other approaches. I will frequently ask for feedback and help. I enjoy when people are direct in their intentions and feedback.

Counter: I will work hard, and I will act on feedback. I will communicate what I’m doing and feeling. Once I am aware of different approaches I will incorporate them. I am almost always willing to change how I’m thinking about a problem or situation if it’s indicated that I should do so.

(But yeah, I’ve been advised I should probably try to relax about the clear goals thing, or more importantly clear approach thing. Be willing to tolerate more uncertainty, be more flexible in the best answer to a problem. But this is a tradeoff skill—I actually was recently given feedback that I was acting too independently in research and needed to learn to ask for help more. And I know from past experience that if I go too long on my own without checking in with someone, I’ll waste a lot of time pursuing unimportant subgoals I set up. There’s not a correct general solution, and I don’t think my approach is one that necessarily needs to change right now, until I’m shown that the other options are much more optimal.)

2) I really like learning stuff. If you can frame it as a topic that will be useful in the future, I’ll learn anything. “Useful” can mean a fundamental skill, research skill, social skill, or something that matters to you when I think you think about important things.

3) I like support: pair me with older students. I like working independently but with supervision.

4) I like having multiple projects at once. I like projects with clear broader-picture goals. I like accomplishing things, so a mixture of long-term and short-term stuff is good.

5) If I’m supposed to be doing something that involves “going with the flow”, I’m going to be pretty bad at it, because those are usually scary things that I’m only doing because I have a specific goal in mind that I’m keeping in sharp focus. I’m probably at my most analytical at those times because I’m unsure about what’s going on. What works is, apparently, distraction. You’re going to have to manage it for me, though, because if it’s new I’m running at full cylinders and not in a mental place where I’m going to be moving my emotions towards “relaxed”.

6) I’m real direct. Like really direct. Especially if it’s a new situation and I don’t have my stock phrases or emotions that I should be feeling sorted out. I’m mostly going to be really quiet, but if you require some immediate response from me it’s going to be strange. (This blog is filtered Monica. It gets even more uncalibrated from proper opinions when it’s unfiltered.)

7) Fairly open-minded and willing to listen. There are lots of crazy ideas out here on the West Coast, but if they work, I’m down. Willing to try new things and challenge myself. Invest fully in things I’ve decided to try. My mom says she used to have some goal she’d like me to work on, and if she timed it right so that it was a year or two out, she could convince me of it and then she’d just have to scramble to make sure that everything was out of my way.

And with that, I’ll close up that section. Next!

I was talking with a practitioner of polyamory (what? That’s not a word, Microsoft Word? Shame) yesterday, and they made me justify why I wanted to get married. It was surprisingly difficult. I mean, most beliefs are surprisingly difficult to explain, especially when they’re the default opinion so you’ve never had to explain them before, but they kept on saying: and do you need marriage for that? And I kept on answering nooooooo

Take companionship. That’s the main reason why I want to get married in the end. There’s someone who is societally committed to you and I think many people would agree that it’s nice to have someone who cares about you and who you can care about in return, who is going to be around for a while.

But they were like: yes, having relationships is good, but why marriage specifically? Why do you need institutional approval?

And I said: uh, so they have to stick around? But that’s a terrible argument. Half of US marriages dissolve as far as I know, and while it might be nice to have an additional barrier to get over in order to make people stay together, if they don’t want to be together, they should not be together.

So then I said: so I can have the public recognition for it? And then they said: but why do you care about public recognition? And I said: um. I think I’d just like it. Like I like having cool friends and people know that I have cool friends.

But then I was relating this to my housemate last night, and they said: but do you really care what other people think about you? I think I’m cool when I’m with a bunch of people I think are cool and we’re doing cool things together. And I actually care far less about what strangers think than a lot of people, and in conversations I tend to get fixated on one person and then completely ignore everything else, and yeah, there’s something to be said about public recognition by strangers, but it’s not my main motivation.

So why do I want recognition among my friends then? So I can say: look I accomplished this goal? (…um. That might, embarrassingly, be a part of it. I like goals.) Because it’s a pretty convenient thing to do, in that pretty much no one will ever ask you to justify why you’re married? (…This is also probably a good part of it. No one asks you why you have kids either, but they certainly ask you why you DON’T have kids. I guarantee that lack of marriage is going to be another one of those things people ask about.) It’s something I kind of like the idea of, because it slightly encourages stability, and because it signals nice things about my ability to find a nice partner? Because it’s maybe sort of romantic? (eh, that one feels weak to me personally). Because it’s… because it’s stable. Right now, my reason for wanting marriage is because it’s stable, and I want that stable companionship thing as a general goal. Also I’m sure there’s some influence from the stupider reasons above, but the stability one feels most true to me.

The argument is then: you can have stable partnerships without getting married. And I say: yes, this is true. The next argument is: and with polyamory, you can have more stable partnerships. You don’t have just one true friend, you don’t necessarily need one true romantic / sexual partner. And I say: well, that sounds reasonable. But complicated. Also, we only have a limited amount of time and emotional investment, and if we’re spreading it across multiple partners and also friends, we’re going to run out at some point. But I agree that’s a reasonable point. Though it seems like it’d be really complicated, and I feel like a normal partnership would be plenty complicated on its own.

So after that conversation I concluded: I’d be open to polyamory, and I don’t think I’ll ever do polyamory because I have a preference not to (because of wanting to avoid complications, a desire for stability, and some silly signaling reasons). Also I think we should make people justify why they want to get married, because there have GOT to be better arguments than this. (Well, the kids argument helps, but I’m not into that one.)

Note: the other thing I learned is that one of the biggest problems with polyamory is jealousy. I get that—sometimes I have stupid thoughts like: “no, that’s my friend”, or “but I was your best friend”, but I fully recognize those are not useful or helpful thoughts to have and I deal with those innate responses appropriately. But I can see how it’d be a problem in relationships. One of my housemates was telling me recently that they were worried that their first partner was cheating on them all the time, and because they always brought it up the two of them broke up. And I was like: oh, wow. That’s a whole different level—sometimes I read about those things, but I can’t imagine not having that trust in a relationship.

So much fun conversation happening while walking along the waterfront with friends. And around the center kitchen table at night, and sitting in circles at parties, and in loud dining establishments in front of food. So many fun conversations.

I was thinking about my long-term goals recently. I’ve wanted to be a professor since I evaluated a whole bunch of options and decided that I’d most like to be a professor. (Isn’t that a great origin story :P). But being in the Bay Area, you realize that there are more options for science majors than being a professor! (Oh man, science PHDs. I’m going to have a PhD, be a doctor. That’s so strange. Then again, it’s not happening for a long time, so not to worry.) (Some part of me always kind of thinks I’ll be dead by then, most likely by bicycle accident. I have no idea why I have this strong feeling that I’m not going to get to accomplish anything because of something stupid that happens to this physical body of mine. I’m also really biased against small groups making progress on big problems, and I know that one’s incorrect. I was just reading something about “What would you do if you knew you were going to die in the next 10 years?” and I was like: uh, probably the same thing I’m doing now? No, that’s not true, if I was GUARANTEED to die after 10 years I’d do… something differently. But that just made me think about how I’m planning for living after 10 years, but I have this weird sense about not being in an accident before I get there.)

Right now, I’m just become aware of the rationalist community, which seems to be one group of people (same culture) spread across a couple of organizations. There’s Center for Human Compatible Artificial Intelligence (HCAI), Center for Applied Rationality (CFAR), 80000 Hours, Effective Altruism, Machine Intelligence Research Institute (MIRI), the websites LessWrong and SlateStarCodex, and others that I haven’t checked out yet but will go up on my website at some point*. There’s some interesting on 80000 hours about what the most effective jobs are, and I’ve been hearing a LOT about AI safety recently, which is doing research into artificial intelligence to make sure it’s not going to have catastrophic consequences for humans in the future. Today I have an interview with someone at CFAR, and I’m super super excited to be attending one of their learning-how-think-better workshops in March.

*(Update: the Center for Human Compatible Artificial Intelligence, 80000 Hours, and Effective Altruism are not quite part of the rationalist community. I’ve been told they’re related, but not quite. I’m happy to hear about how everything is connected, since I’m currently trying to figure out how all of the organizations and people are related!)

Anyhow, I was thinking about what I want in a job, and what I like in a job, and there are more options than I thought that fit those criteria when I’m done with my PhD. I like communicating science and teaching. I like presenting to public audiences. I like writing. I like coding. I love learning. I like being independent, and thinking all day. I want my work to intersect with the artificial intelligence / machine learning / computer science communities. I want to work on social inference. I want to do something that advances human progress. I want to think about people. I want to do something that earns me enough money to live on and donate. I want to work in interesting work. I want to be surrounded by interesting people. I want to help younger people advance in their goals.

Academia seems to be a great place for all of that. Selected industries, though it’s hard to work on things like social inference in industry. Organizations like those in the rationalist communities are very intriguing. I don’t know nearly enough about how they work to know if working in those arenas could be something that’s fun to do. It doesn’t really matter at this point, since the skills I’m learning (thinking, hard skills, knowledge, presenting, writing, managing, etc.) are going to be useful for all of the possible jobs I’m interested in, but just thought it was fun to realize that not everything necessarily has to be about the professor goal. (Though, just to be clear, until about 5 years from now when I actually have to decide, I’m all about the professor goal.)

AI safety is a big deal among some of the people I’ve been hanging out with. I was recently witness to a quite interesting conversation, in which one of the people who does AI safety things for a living was having a conversation with an undergraduate. It went something like this:

Undergraduate: I’m interested in the intersection of inequality and technology. I’m going to be working on a project relating [x].

Researcher: Ooh, that’s an excellent interest. Have you thought about doing a focus on AI and inequality? This is going to be a huge field. I have people from the government calling me for advice, and I’m like: I’m a mathematician, I don’t know. Your interests are perfect; if you can develop those skills you’ll be perfectly set up in the future.

Undergraduate: No, I wanted to focus on [x].

Researcher: No, but seriously, in three years, you are going to be in such a good place. This is where you want to focus. This is a really important issue, it’s going to be big.

Undergraduate: Well, I want something that’s pretty well-defined, that I’m going to have support for right now, that has a bunch of literature so I can write my thesis on. I’d really rather work on [x].

Researcher: But in the long-term, this is going to be important, and you have a great opportunity to develop those skills now.

Undergraduate: I’d really rather work on this, I think this is the most important thing for me right now.

Researcher: Okay.

It was fascinating. For one thing, it was fascinating because I’ve been asking about what career I should be doing for the last few months, and it turns out there’s an entire website that lays it all out for you (80,000 hours) which people had sort of directed me towards but not in a like: if you want the answer to your question, read this kind of way. (Exception of Smitha—thank you Smitha! And to those people who also did this but I didn’t catch because there were a lot of new terms being thrown around in the beginning. But still, Smitha sending it was the first explicit READ THIS I understood.) The problem is that people don’t know about the website, but this researcher definitely knew about it, and I kind of wanted them to say: no, undergraduate, you really should look into this website, it’s very useful, after his attempt to verbally convince the undergrad didn’t work.

It was secondly fascinating because the undergrad didn’t listen. There are things I won’t listen to, categorical things I’ll dismiss out of hand that I shouldn’t. Career advice is not one of those things for me. Neither is anything that comes out of the mouth of this particular researcher. I mean, this researcher is accomplished. They don’t talk about in person, but it’s all over the various websites they’re a part of, and even in conversation you can tell they’re one of those people who is extremely knowledgeable and also a great communicator. It was just… strange. I mean, I’ve done the same thing that undergraduate has done, even occasionally for career advice, but if it comes from someone who I know knows what they’re doing, then I’ll at least consider it. I think it has a lot to do with respect for the person giving advice, and how you know whether to award it. I’d watched this particular researcher interact with people more, plus I’d looked them up, plus I have a lot of respect for qualifications (probably more than I should) so I guess the undergraduate just didn’t quite have this bias.

It was thirdly fascinating because this researcher wasn’t trying to convince me to go into a field they cared about. They don’t care too much about what I’m doing—they don’t think it’s stupid, but neither is it something they’re passionate about. So it wasn’t a general recruitment thing, it was a specific-to-this-undergraduate recruitment thing, and I wonder where the line lies, exactly, between how much effort one should put into convincing someone versus how much potential a new student has. This was an easy situation—it takes practically zero effort to put an idea in someone’s head—but it was kind of sad to watch the researcher abandon it and the undergraduate reject the proposal. I don’t know.

Another fun roundtable :). This time, we had four people: this same researcher, and three other graduate students. It was awesome—we each had 15 minutes to talk about what we were struggling with that was making us unproductive. I.e., everyone had free reign to talk about their problems for 15 minutes and then gets feedback. I was in heaven. (Deep discussions with lots of personal information paired with feedback is ENTIRELY my thing. Since that isn’t incredibly obvious.)

One person explained that they didn’t have a lot of introspective access. I was dumbfounded. Really? What does that even mean? To introspect, one looks back at one’s actions, and how one felt about… oh, okay, I get it. If you aren’t super good at tracing back why you’re feeling a certain way to the cause, or if you generally don’t think about feelings much, or think about what you’ve done wrong, then I could totally see that meaning “less introspective access”. And they made the point that this is an individual difference that varies across people, and I buy that as well, because I was reading one of my high school essays and it was talking about how everyone needed to introspect more so they could learn about themselves and make better decisions and I face-palmed in my head. So all right, not having a lot of introspective access.

How does one compensate for that? I lot of people don’t do introspection, but they work through problems by talking about them with people, and then getting feedback. External feedback’s much faster than sorting through stuff on one’s own, and often more accurate because you get an outsider’s perspective. But I don’t get the feeling that this person was doing a lot of external review either. But we were in this case! Because we were discussing productivity problems, and doing this external review.

After this person mentioned a bunch of things, the researcher said something I found pretty useful: “Say that you hypothetically benefitted from this conversation. If you did benefit, think about why you’d benefit and let’s backpropagate, figure out what kind of information we can give you now.” (Back-propagation is a method to train neural networks involving taking the final output and updating backwards. It’s what it sounds like in this context.)

It was a great way to make sure we talked about what that person wanted to hear, not whatever thing we found relevant. (On my turn I just let everyone talk about what they wanted though, since I was using it more as an exercise to have everyone talk about what they cared most about (and thus had expertise in) and I wasn’t super focused on solving my own problems. As promised, we did not solve my problems but I heard some pretty fun perspectives on how I should view my life.)

We did not make too much progress on the first case, but moved on the second person, who told us about motivation problems on a specific project. At some point, the researcher got up and started drawing bubbles on the board. I thought the answer to the actual motivation problem was pretty obvious: this person was not enthused about their current project, and should move onto a different project—but that’s not where the conversation went at all. When I listen to people telling me about things, I figure my role is to pull out what they’re actually thinking. People usually have some opinion, and they circle around it, and I usually end a conversation like this with: all right, so this is what I feel you care most about, for these reasons, so if you want to do that, you should maybe take these steps. I’ll insert my own opinion in there too, but I figure I’m mostly there to reflect people, let people talk themselves into what they believe.

Not this researcher :). This researcher got up and drew all the bubbles of what the person was talking about, all of the possible side reasons why they felt the way they did. In the end, they told me delightedly: see, the mind is a graph structure! And I thought: that’s a fun perspective to use to look at this. It really helps focus what you’re thinking, to write down all of the components and relate them to one another. And while I was more solution-focused, this was process-focused, understanding what elements were valued and how they played a part in decision-making. Both are good perspectives, I think, and would be helpful in developing next steps. One has a lot more information than the other :).

(I’ve actually been thinking about my role as a listener recently. I’ve found that the best way to not offend people is to just reflect them—isolate what they’re thinking, resay in their own words what they’ve been telling you. I completely suck at the influencing part of it. Like, if I have a strong opinion that I think could help make things better, I have no idea how to convince someone to do it if they don’t already want to do it / are interested when I first give the suggestion. Figuring out how to convince people to do something that would help them but is something they’re initially opposed to seems like an excellent excellent goal. I’ll continue looking for examples around me.)

Then it was my turn! I love this stuff. I started out with a few things about how I run my life: “I don’t have problems with motivation, because I’m really easily motivated by “shoulds”. I care about my research, but I also spend a lot of time thinking about identity-things and social interactions. I get distracted while working with social / personal thoughts, and will write them down before going back to work. I value social networks and will do things like go out for coffee with people rather than continuing to work. I have different blocks of time for different types of work—things like thinking work, coding, admin work, social things—which I try to distribute. I know everyone talks about doing better meta-level reasoning—trying to figure out how to best distribute their time—but I try not to do too much of that, because I already do a LOT of meta-reasoning about time allocation and I think more is just going to make it worse. It gets stressful constantly pinging if I’m doing the right thing or not. (If I could make my meta-reasoning smarter that’d be great though.)

We then talked through everyone’s thoughts on that. Someone suggested that I carve out a separate time block for writing down blog thoughts. I told him that that didn’t really work, since I actually have my best blog thoughts while I’m trying to do other work, and they’re hard to get rid of anyway. (Another friend again told me to try meditation for getting rid of invasive thoughts. I’ve tried it half-heartedly before, but I think that’s a good idea and would be helpful. But I don’t care deeply enough that I’ve gotten around to it, and have kind of resigned myself to how distractable I am. I value the blog thoughts, so they get integrated into my life.) We had some more discussion how to integrate various goals in our lives that have demands on limited time. Apparently, there’s a way to integrate your various life-goals so they’re working together rather than competing with each other. I agree—my work life, social life, and internal life all move along in parallel, for example, and I don’t think they compete with each other unduly. On the other hand, there are some goals that are so low-ranked in my mind they might as well be put in direct competition in work life. That was an interesting suggestion though, and I think one that could be implemented.

(Other random life-hack that was introduced: graduate students often don’t realize that they can change time for money and money for time. Like if you take ubers, then tutor for $150 an hour, you’re trading money for time and time for money at maximum value.)

And as I was listing all of my values and how I integrate them, I was thinking that I wouldn’t have know how to frame this stuff a few months ago. People have been pushing me here: getting me to state what I care about and why I do. Another thing that’s been very strange to realize is that I’ve reinvented a lot of the rationalist perspective without having ever heard of the rationalist community here. I’ve independently adopted a lot of the common strategies and organizations. I’m currently hanging out with family, so I’m noticing how much of it is family values, and also talking to my parents about how I was when I was younger, because while my sisters are also weird, I’m much weirder than they are in this respect.

The researcher was really lovely though. At some point during my discussion—my segment lasted more than a half an hour—they stopped me, and said:

“You need to sign up for a CFAR workshop. You’ve heard of CFAR?”

“I think so?” (tapping on the computer. Center for Applied Rationality. On my list to look at but not yet investigated.)

“You’re the type of person who would like this. I get the sense you’re a writer—it’s important to have that external memory. And you’re in psychology, and the way you plan. Actually, you should do it right now, I can send a message to the interviewer.”

And that led to me being able to go to a CFAR workshop in March, and I’m tremendously excited. Excited because I get to meet a whole bunch of other people who think like I do, get to learn some awesome techniques, get to stop reinventing the wheel, and am very very thankful because this researcher reached out to me. It’s people like them and my friends and advisor who go out of their way to offer me opportunities, and it’s incredible and I’m so thankful to them and thank you so, so much.

Change of topic! Back to the beginning, actually—you know how I was dealing with lots of emotions at some point? Mostly dealing with them on my own, because it was a pretty sensitive situation, but I talked to some other people about it and it was such a special experience.

I’m used to my own head. And in my own head, it’s important to be fair to people. That’s one of the most important things, actually—sorting out my thoughts by making sure I’m evaluating the situation in a way that I can understand what other people were thinking and why they did what they did.

Turns out, other people are much more the advocates of the “it’s valid for you to feel what you feel” attitude. It was incredibly touching, because through my process I have to keep more than one person in mind, but my friends were just focused on how I felt. I mean, I always do keep myself in mind first—the whole purpose of reframing situations is so that I can feel better—but it was nice to have the option to be unfair and have that be accepted. That’s not how a lot of things in life work and I appreciate the sentiment :).

Here’s a question: there are people who have pretty deep thoughts, but also a lot of the time just say whatever first comes into their heads, or directly contradictory thoughts for the humor value.

How do you extract information out of these people? Like, they’re fascinating, but I was to be like: TELL ME ALL YOUR THOUGHTS (but I’m really not sure you have them cached, so I don’t know how to access them).

Thought: you know when you have a great idea and you’re super enthused about it? But when you share it, people are not as enthused about it as you are? Sometimes I don’t want to share things because I want it to continue to be just as awesome as I think it is. The blog’s a lot like that, actually—an echo chamber for things I care about. Heh. At least you poor souls can choose whether or not to read it. Much much love for you :).

I adore the neuroscience cohort. We had a Christmas party on Friday, and it was so lovely. Everyone’s thoughtful and wonderful and way too fond of puns and gets along, and I didn’t think I’d have this again in a group this big but I do.

All right, I’m done. 9500 words of way too much self-regulation nonsense, but now it’s out there and sorted and we should progress back to normal slightly-more-variance content in future weeks. You guys are the best for reading, and for the people at home who have been incredible in all of the opportunities and kindnesses, thank you so much. Happy holidays to everyone and wishing you all the best!


First picture since arriving in Berkeley! Smitha, me, and Ryan

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s