JOSH: Hello, and welcome to Clearer Thinking with Spencer Greenberg, the podcast about ideas that matter. I'm Josh castle, the producer of the podcast. And I'm so glad you've joined us today. In this episode, Spencer speaks with Tom Quisel, about online dating and relationships, predictors of job performance, self-experiments, and the dark forest hypothesis.

SPENCER: Tom, welcome. It's great to have you on.

TOM: Yeah, thanks, Spencer, great to be here.

SPENCER: So you have a huge amount of experience on sort of the inside of the online dating world. So I'd love to ask you about that first. Can you tell us a bit about what it's like behind the scenes in online dating on the kind of technical end and kind of things you see that maybe the end-user doesn't see?

TOM: I can speak to that a little bit. So first, kind of some personal experience – so I think when I first moved into online dating personally, I created an online dating profile. I started doing it, I think that was 2006 or ‘07. It is really interesting to me, because I did observe the sort of traditional dynamics of dating, which are kind of disappointed, like, I come from a pretty again, certainly have a pretty egalitarian view of men and women. And this is just focusing on straight dating for the moment. And so you kind of hope that it's the metric, yeah, maybe a woman asked the man out, or maybe a man asked a woman out, and it's all kind of the same deal. But of course, as we all know, like traditionally, and even still today, it's not like that. So typically, like, a man will express their interest and ask a woman out, of course, that's not always the case. So when I saw online dating, I thought, man, this is a really cool opportunity to completely break that mold, like this is all the context has gone, all that traditional trappings that's eliminated here. So surely, this will be basically a egalitarian, and men and women will behave basically the same on a dating site. And yeah, like it was pretty crazy. Like I couldn't have been more wrong. I think most people have seen, if you've participated in straight dating on a dating app, like on Tinder, like men will spam out like spam out low-quality messages, and women will, for the most part, ignore those messages. And of course, there are exceptions, but that's the general rule. Like in one of the places I worked OkCupid, the pattern was typically men send 10 messages to a woman for every one message that a woman sent to a man. I mean, first contacts.

SPENCER: It's a huge discrepancy.

TOM: Yeah, right. And so you see something like that, and you're like, oh, shoot, we're still stuck in this trap. And it's really interesting to think about how it comes about. And so my best guess for it is that it's like a self-reinforcing cycle. So you start out, kind of at some random position. And then one guy comes on who's used to being ignored. And so he spams out a lot of messages. And then this first woman comes on and has this experience of receiving low-quality messages she's not interested in and then gets into this place of expecting messages to be low quality, getting a lot of them and therefore valuing them less, which leads to a response rate going down, men notice as well, the low response rate, they say, “Well, the answer to this is quantity, not quality, just a really unfortunate conclusion.” And then they start stemming more. And so you end up in this kind of spiral to the bottom. And that's my best guess, for what's going on there.

SPENCER: So what do you think of the kind of evolutionary psychology explanations that some people want to give that basically, it's very, very costly for a woman to get pregnant. Whereas it's like very low cost in an evolutionary sense for men to get someone pregnant, because he doesn't have to care for the baby, you know, like we're talking 50,000 years ago, obviously. And that that might create sort of a different dynamic, and that because it's so costly to get pregnant, that they're much more selective with their partners, whereas men because low cost, they're less selective.

TOM: Yeah. I feel like Richard Dawkins talked about that pretty convincingly in The Selfish Gene, and it makes sense to me. But that said, it seems like a pretty big leap from that to the way people send messages on a dating app. I mean, that's not getting pregnant. Maybe it's at the root of it. And like that kind of led to the tradition. And then the tradition has led to these latest dynamics that we see now, but it seems like a little bit of a stretch to me.

SPENCER: So how do online dating websites try to deal with this dynamic?

TOM: So yeah, well, in my time (04:23) relocated, a lot of our focus was kind of, in some sense, trying to nudge and shape behavior to be healthier and more productive. Your really interesting thing is how businesses define a KPI or key performance indicator. And so one naive KPI you might come up with for a dating app is number of messages sent but of course, not kind of scenario. The way you you maximize your KPI and feel like you're succeeding in your business is having tons of messages sent which could in fact, be a result of men sending more messages, women respond in less and actually have an overall unhealthy ecosystem. So we thought about this problem a bit and came up with some different indicators. One thing we focused on was kind of humorously was first contacts and three-way conversation. So this is a conversation where A sends a message to B, B sends a message to A and then A response back. And the idea was that this is a basic indicator of a quality interaction, rather than just a stay on that's not pleasant or enjoyable for other sides. And then that only solves kind of half the problem. The other half is more focused on dealing with this dynamic we're talking about. So there, you see that one of the underlying issues is that the way that men choose to send messages. And of course, women do it too, but of course, it's more dominated by men. And you end up with the men focusing in many of their messages on just the most attractive women. So it kind of ends up being another one of these Pareto distributions, where you have few people at the very top, who got most of the messages, and then the tails off very quickly. So you end up with many women, and also many men who really don't receive any first contacts. You can think of this in the same way as wealth inequality. And in fact, it leads to many of the same undesirable consequences. And so this was another KPI we tracked, we looked at the fraction of users who received a contact each week to at least one contact a week. It's like a very primitive measure of evenness or of how equitable messaging is across the site. And so we put a fair bit of effort and to try and to increase that.

SPENCER: I think this is a really great example of how each individual, while they're feeling like they're acting in their own interest actually ends up creating a system that doesn't work well for most people. Because like, let's say, a man feels like he's not getting a lot of responses. So he starts sending out more messages that create lots of spam in the system, causing women to respond even less. All the men doing that creates the system where people, in general, aren't happy. And similarly, you know, a man trying to send a message to one of the best looking women on the site might feel like, oh, yeah, well, that woman's really good looking, I want to send her a message. But then everyone doing that causes this crazy, Pareto distribution of messages being sent to like a small number of women get, you know, a very large proportion of all the messages. And then of course, they're not gonna respond to most of them. So it's just a great example of kind of how stuff can go to share it with everyone, even if everyone is acting in their own rational self-interest.

TOM: Yeah, exactly. Yes. Isn't it that fascinating where individual actions add up to this kind of emergent properties of the system as a whole are super undesirable? So we started looking at a few different statistics about how users send messages. We looked at the number of first contacts a user would send out per day. We look at the response rate to the messages they sent out and so that kind of gives you a sense for the quality of the messages they're sending, and how well targeted they are as well. Basically, it tells you how wanted is this messages message. And we also looked at the response rate that they gave back the messages that I received, as well as some information about attractiveness based on how users vote on other users. So basically, like using Tinder-style information, where you see left swipes and right swipes, you look at the ratio of left and right swipes to tell you roughly how attractive other people on the site think a person is. And so you can start to see with all this information, you can start to pick apart how much each person is contributing to these undesirable overall dynamics. And so we'd start doing things like identifying people who send really low-quality messages that rarely get responses. And we would tend to recommend only users who get fewer messages to that user. And so that way, you take some of the pressure off in the spammy experience, off of the most attractive people on the site. And you add some more attention onto the people who are less attractive.

SPENCER: That's really interesting. Like, it's a pretty good solution. Because they're sending lots of messages, you can redirect them to people that aren't getting enough messages. I suppose they're spammy messages, but maybe better than getting none, right?

TOM: Right, exactly. Yeah, that part is a little bit uncertain. But you know, at the end of the day, it's humans on both sides, right? Like, sure, maybe someone sends a low-quality message but that doesn't mean they're like a bad person. Like, it could still be very worth making that connection.

SPENCER: That's really cool. Any other tricks you want to share with us about how you kind of optimize the system to make it work better for everyone?

TOM: Yeah, well, like veering a little bit into machine learning. I think this is a really interesting example. And I'm curious about your take on this. Because you're more of an expert in this area. You can imagine trying to build a recommendation system for a dating app. And it's basically like Google, like it's showing results, you know, in the search area, showing users to another user – and the goal is presumably to maximize the number of contacts, right, similar to maximizing clicks for a search engine. So you can imagine doing this in a really naive way. You have a million users or whatever. You lay out all the attributes about all the users and you try to predict given this user who is likely to message – sounds like a pretty standard machine learning setup. What you would find, unfortunately, is that the user, of course, is going to contact the people who are the most attractive, the people who are the most contacted, and basically, the people who are most desirable in any number of different dimensions you can measure. And so, as you can imagine, if you launch this recommendation system, you basically have one optimal user who's the most desirable. And maybe there's some variation, like, not everyone has the same taste, but there's an overall tendency. And this person will get recommended at the very top of the list, maybe because they respond a ton. And then immediately, they would get flooded with messages, they would have a terrible experience several messaging them another terrible experience, and their response rate would go almost zero.

SPENCER: Let me make sure I understand. You're saying basically, if you train a machine learning algorithm, to predict who this person is gonna want to message, you're gonna end up saying to a lot of people, oh, you're all gonna like the same person because you know, they're really attractive and really responsive. And so ultimately not going to be good for the whole system.

TOM: Exactly. Yeah.

SPENCER: So how do you think about adjusting the problem you're trying to solve? Or do you feel like it's just it's actually just quite challenging to use machine learning in this context?

TOM: Yeah, I mean, in my time in, okay, at OkCupid, we stepped away from using machine learning at that top level, and use the lower levels to predict different factors, which were then incorporated by hand into the matching results. I actually don't have a good answer for it. I mean, I think what's so fascinating about it is that the kind of traditional machine learning setups, typically assume, or gothic behavior is kind of like static, like, yesterday is going to look like today, it's going to look like tomorrow in terms of the behavior of the system, but that doesn't take into account is that the actions and the changes that the system itself brings into the recommendation system brings into the overall environment, change what the recommendation system should do. And so since it's not trained on its own, you know, on the results of its own activity, it's not doing the optimal thing,

SPENCER: Right. And with online dating, if you did so using a machine-learning algorithm to predict who someone's gonna like, and then recommend that person to them, you get in that weird situation where then that's influencing user behavior, and then you're feeding that data back in as training data, but the data was actually generated by your machine learning algorithm. So you get this weird, the output becomes the input. And you can get all kinds of weird effects from that.

TOM: Yeah, exactly. And that there's this kind of inherent negative feedback loop. If the machine learning algorithm succeeds in any way of finding someone who really has a good recommendation, then, because there's a certain amount of homogeneity between people, its success is going to bring out failure, because any one person getting too many messages is going to lead to that person having a terrible experience.

SPENCER: Going back to the question of attractiveness, to what extent do you agree about what's attractive?

TOM: Oh, yeah, that's a really good question. It's pretty high. I don't have a specific number, but it's fairly universal. It's, of course, not perfectly universal, but high correlation. And I think one interesting kind of related tidbit is looking at what goes into attractiveness. And OkCupid did this interesting experiment years ago, where for a while, profiles have been shown, and then users could score them, rate one to five. And there were separate scoring, one to five ratings. One was for attractiveness, and one was for personality. And then there was also the photo shown, as well as some summary texts that the person wrote about themselves. So over time, OkCupid removed the two different kinds of scorings and move just to a single scoring, which is just how desirable is this person for you. And then, as an experiment, OkCupid removed the text about the person in some cases. And so this is an experiment that's testing, what is the effect of having this profile text on the ultimate result of the attractiveness? And another way to put that is, how much is attractiveness, a result of personality, and kind of something someone writes about themselves, and how much is it a result of the photo itself? So we ran this experiment and found, I think it was like a 0.9, 0.6 correlation between the two. In other words, seeing someone's profile text or not, had almost no impact on whether one user thought another was attractive.

SPENCER: So they're basically ignoring the text essentially.

TOM: Yeah, this kind of like gets to – I found this like, over and over again, and on online dating is people like aren't under the scrutiny of other people. It's just, they're just behaving however they behave, and not individuals. But in aggregate, people start to look kind of bad, like, in this case, people. And I mean, it includes myself, no one is exempt from these kinds of statistics. But yeah, like people look pretty shallow, right? When it's almost entirely the photo that's determining the result of that experiment.

SPENCER: Now, I wonder if you could design a system where you kind of de-emphasize the image and get people to actually make a judgment based on the text. Do you think that if you were to do that, that would actually lead to better dating outcomes?

TOM: Yeah. So I think one really interesting example of trying to upend this dynamic is Bumble, which has been highly successful. And their route was just, well, men seem to spam out a lot of low-quality messages in general, like, let's have them not do that. And only that women can do the contacting. And I think that's pretty brilliant. And it's been effective. But back to your question about photos. Yeah. So one time when I was there, OkCupid had a Love is Blind day. So this was a day where we hid photos on all profiles across the site.

SPENCER: Wow, that's awesome.

TOM: Just text. Yeah. And the idea is, you know, it's kind of like, I mean, we really did want to encourage users to think more about the personality of people and less about attractiveness, and just see if they could have a positive impact on the culture. And what we found was, I think it was halfway through the day, there was such a massive negative impact on metrics, the number of messages was way down. The number of successful interactions is way down. I mean, it almost looked like a ghost town. We were so panicked by this result that we cut it halfway through and turn the turn all the photos back on.

SPENCE: Oh, my gosh.

TOM: Yeah. So definitely not a success in that department. So yeah, I don't know that without photos, you could build a dating system that humans would really enjoy. Like, I think at the end of the day, looks are so really important to people.

SPENCER: You know, another fascinating thing about the OkCupid approach is the color matching algorithm. For those who haven't seen it, and actually not sure how it works now, but at least back in the day, the way it worked is you would answer a bunch of questions about what you're like, but then you reach the question, you'd also answer what your ideal match is like. So for example, say like – do you smoke, you know, yes or no? And then like, what would your ideal match answer to this? And you say, “Oh, I don't smoke and my ideal match also wouldn't smoke.” And then by answering, you know, tons of these questions, it is you could come up with a kind of compatibility score, the extent to which you each tick each other's boxes that you say you want in a match, and then you kind of rank order people based on the extent to which there's a bilateral match. Did I describe that right, Tom?

TOM: Yeah. That's a really good description. Yeah. And the kind of way we would describe it as is we would say is each person can build their own matching algorithm based on what they care about.

SPENCER: Yeah, I thought that was such a cool approach. And one of the things that makes me so sad about the online dating world, is it seems like it's gone away, generally from kind of clever matching things, trying to help you find your soulmate, and more towards just complete superficiality, embracing the fact that people just care about photos and just, you know, swipe left or right, which I have to imagine actually makes dating much harder.

**TOM:**Yeah, I think that's a really interesting question. So yeah, like a few comments, like one is that it's interesting to kind of put this in the broader picture of how attention has changed over time, and how content has changed over time. And this movement from the kind of longer form, content, and experiences to shorter, more gratifying activities. Like Tinder, swiping left and right is probably the pinnacle of that. And that was a really interesting thing that we observed in the OkCupid messaging data over time. And this is published in Christian Rudder's book Dataclysm. A lot of this is, I totally recommend that book. And what we found was, I think it was between 2004 and 2014. We looked at the average message length each year. So just like other messages sent, what was the average length this year, from 2004 to 2014, the average length of messages in that year. And what we found was a pretty market downward trend. So users in 2004, were sending messages that were two or three times longer, on average than ones in 2014.

SPENCER: Oh, wow. That's a huge difference.

TOM: Yeah, totally. So I think there is this general movement towards more instant gratification, kind of faster responses. And I think what's interesting in dating is that an online dating particulars, you can put a huge amount of effort into trying to do a good job on the dating site. So not just showing photos, also trying to capture someone's personality, try to capture compatibility at a deeper level. But at the end of the day, there's a lot of information missing from the virtual experience. And like at the end of the day, dates are in person. And relationships are for the most part in person. And so what really matters is compatibility in that scenario. And so I would say that there's a certain element of chemistry, and other harder-to-find pieces that just aren't captured on a dating app or on a dating site, no matter how hard you try. So it's a little bit of a case of diminishing returns. So I think that that may also be driving this tendency towards, let's just very quickly identify some super shallow matching, like, “Oh, we both find each other attractive.” And then from there, let's just move as quickly as possible to a date where we can kind of measure some of these harder to quantify in-person things like chemistry.

SPENCER: It seems like the ultimate dating site like, you know, if you could build the best possible one, it would essentially predict the actual compatibility you would have with different people really accurately. And in addition, It would get from you your real intentions about what you're looking for, like, “Do you want a long-term relationship? Do you want children? Do you just want short-term relationships?” Whatever. And with the combination of incredibly good accuracy of predicting a call, you're gonna be compatible with plus this, like actual knowledge about what you're looking for, that would be the best possible dating site because it would give people the most of like, what they really want. And it seems like we've actually just gone in the opposite direction from that. It seems like with things like Tinder, we're moving towards more and more sort of random matchups that are highly superficial. I mean, do you feel like we're in a backslide in terms of getting to the ideal?

TOM: You know, I really don't, I think at the end of the day, people are moving to these newer, faster apps, because they love them, because it works on some level. And, and I think it depends on how much information you think is, is present in the app versus on the in the in-person experience. Like, obviously, if you could only learn 10% of the information that's important to you in a virtual experience, then it makes sense to just jump past that as quickly as possible to get the other 90% of the information. The other thing is like the scenario you're describing is like do you want kids or not? That sort of assumes you're looking for, like a long-term relationship. And I think that's a that's not a lot of what people are looking for. You might be open to that in the long term. But I think usually you're just looking for someone, or tonight or for the next month, or just to meet someone and see where it goes. And so I think that if that's your criteria, it's a little bit harder to say like, how do you really solve that, aside from just looking at someone's photo? See, I

SPENCER: So I guess I view it as a trade-off between short-term interest and long-term interest. I think that apps like Tinder are a lot more fun for people to use short term. And they kind of gamify the experience and give immediate rewards. But I actually think they make worse dates in the long term compared to something like OkCupid. But you don't know that's true?

TOM: No, that's a good point. Yeah, again, I think the benefit of something like Tinder is you can very quickly get that other missing half or whatever of the information.

SPENCER: If you could just meet a bunch of people quickly, you're saying.

TOM: Yeah, exactly. And often, you can rule out a lot of people that way. So if you're on OkCupid, and you spend a ton of time finding your perfect match, and then it turns out, you don't like the way they smell or whatever, then you just wasted all that time. But I think you also made a really good point about short-term versus long-term. I think an ideal system might be a little bit like Tinder at the start. And then it would provide a framework, later on, let's say you make it through a few weeks or a month, and you're still really happy with the person provide a framework for ensuring that you get to know each other at a deeper level and are compatible at a deeper level before you go too far. Like the kids question, for example, can be super important. And like, Yeah, I mean, you see relationships where people kind of skirt around that issue for a long time and it kind of badly or can end up well, but it may be better to face it in an organized format. That's like a potential hybrid idea. But I haven't seen anything like that.

SPENCER: Well, yeah, it's I mean, it's really tragic when people have some fundamental deal breakers, but they kind of fall in love with each other. And then they kind of the deal-breakers come out over time. And it's like, oh, wait, we're actually highly incompatible, but we're extremely attached to each other. And now you kind of either just live with each other despite the deal-breakers or you have this painful breakup.

TOM: Yeah, exactly. And one other thing I think I should touch on about the way people are going about this question of how to build an ideal app is I think people don't know themselves very well. Not all the time. Often, I think what someone says is different from what they want, or just the way these questions are phrased. Like they're in such isolation, like, you see a question like, is having a political match more important to you or is having an activity partner match more important to you? And depending on kind of what you think of as what politics is, and what activities are, you could get that question a bunch of different answers. And so you can see that would not be very informative. And I think like one really extreme example of this, which also is kind of like a funny follow-on story of astrology signs. So I think OkCupid and most dating apps support being able to post what your astrology sign is, like Libra or Scorpio or whatever. And OkCupid did this analysis where we were like, man, this is a really interesting thing. We have all these people matching, and messaging. And we have all this astrology information like this is a great chance to actually answer the question like is astrology important for dating, right? Because like when you go look at the horoscope or people talk about astrology, that's one of the most important elements of dating and kind of who you're compatible with.

SPENCER: Right. If your sign actually mattered, it should affect your match score on OkCupid with different people, right, based on their sign.

TOM: Exactly. Yeah, yeah. Because people are creating their own matching algorithms and what questions are important to them, and saying what answers they want. Yeah, you totally expect that it would show up and match percentages across the site. And so what we did, and this isn't an OkCupid blog post, you know, computed a grid. So for people who are one of the 12 signs on one side, what's the average match percentage with people who are one of the 12 signs on the other side. So you've got like a total of 144 little cells, each representing a different pair.

SPENCER: Like Scorpio versus areas or whatever, right?

TOM: Exactly. Yeah. And so if any of the predictions are right, then you'd expect like, oh, well, like the match percentages will be higher in some of these cells and others. What's interesting about this approach is it's not assuming some particular system that astrology is putting on top of it, right? It's not assuming that astrology believers think that these two signs are compatible. It's just like, is there any signal here at all of any kind? And what we found was that every single cell had the same average math score, I think it was 61%, something like that.

SPENCER: It's kind of in one fell swoop. It's kind of the most convincing refutation of this, of astrology for dating I've ever seen.

TOM: Yeah, exactly. Yeah. And we did a similar analysis, we're like, well, maybe it doesn't affect mass percentage, but maybe people are aware of it when they send messages like, you think that it would, it might determine who people decide to contact. And it didn't show up there either. I presented this small talk for some friends and a few of them were horoscope believers. And this is part of a larger presentation. I wasn't just there to trust people's beliefs. And to be clear, I respect people's beliefs, like if they have a positive impact on their lives. I think that's great. And there are many perspectives to look at the world, right? And this is just one of them. But yeah, so anyway, I give this presentation to some of my friends who are astrology believers, and have asked afterward, I asked them, “Well, hey, what did you think about that?” And they said, “Oh, yeah, that was interesting.” But then they just went back to looking at the world through the astrology lens as usual. It's just a really interesting example in how seeing evidence isn't necessarily that important or useful for changing your mind.

SPENCER: Well, maybe one of the things where there might be a negative correlation between believing in astrology and being the sort of persons convinced by that kind of data.

TOM: Yeah, that's true. Yeah. And then it's kind of like, well, how do you get someone out of that circle?

SPENCER: Yeah, well, maybe you have them predict in advance, like, okay, you explain the setup. Like, if astrology works, what would you expect to see? And then maybe they'll be more invested when you show them the result.

TOM: Yeah. Well, you know, in fact, I did have one conversation with someone separately about this a year or two later. And they said, Oh, well, actually, there's more complexity than just that. Do you know, there's like more than just your astrological sign?

SPENCER: Your astrological sign and your moon sign or whatever? Yeah. Okay. All kinds of things.

TOM: Well your analysis and look at that. So that's where the real signal is.

SPENCER: But still people all the time, use just the regular sign like you hear people say constantly like, “Oh, that person's an Aries, therefore, you know, I shouldn't date them or whatever.

TOM: Exactly. Yeah, that's what I've seen mostly. You know, I've never heard someone go into the details of the four different signs and predict something from that.

SPENCER: So recently, your experience has changed a lot in the online dating world because now you're actually working on an app for gay men. Is that right?

TOM: Yeah, that's right. Yeah. So I've moved over to work in Grindr now, which is the largest gay dating app.

SPENCER: So what I'm really curious to know is how is gay male dating different from primarily heterosexual dating?

TOM: Yeah. So what you see as you shift from primarily straight dating up to a primarily gay male app, are a few shifts in messaging behavior. So first off, each user tends to send out more contacts and new people per day. And then you'll also see that those contacts, a higher fraction of them receive responses, then finally, you see that the messaging conversations tend to be longer in terms of message count, on average.

SPENCER: What do you attribute that to?

TOM: My take is that it has to do with different initial conditions. And by that, I mean that when gay men drop into dating, or online dating, or even specifically Grindr, for the first time, their experience, and how they want to be communicated with matches pretty well how they're actually communicated with. So that is, they'd like to receive a certain kind of message, they'd like to receive it at a certain frequency. And they'd like to go likely to go to a certain place. And there's pretty good alignment there. And so gay men end up with pretty positive experience out of the gate, and end up feeling pretty happy to respond to most of the messages they receive, and also have a pretty positive experience and how they send messages because they get a lot of responses. And so they continue in that same strategy that they start out with, whereas in straight dating, I think it is touched on before – when a man first drops into online dating, and this, of course, is just a general trend. That's not going to be in every case, but often a man drops and sends a bunch of messages, doesn't get a response and then adjust their strategy and there's a few ways man adjusts his strategy, but I would worry that but often the way they adjust their strategies to some more messages of lower quality. So obviously that's not desirable for either side. For either men or women, and then similarly, women drop-in, and they're receiving far more messages than they would have liked. And so as a result, they start responding to very few of those messages and feeling generally dissatisfied with the messages they get, because so few of them meet the bar that they would expect. And so you have this kind of downward spiral that's happening as a result of that dynamic. Whereas you don't see that downward spiral in gay dating, at least on Grindr. Plus, that's what I attribute the differences in those two populations in terms of online dating strategies.

SPENCER: I wonder if there's also a safety issue, like maybe men feel more safe meeting another random man, whereas maybe a woman is more wary and wants to do more of that in first before like, meeting up with a stranger?

TOM: Yeah, I think that's a great point. I mean, gay hookups are not without danger, as well. And to be clear, neither kind is dangerous. I think that's one of the interesting things is, people often come into online digital and think, “Oh, well, how could you possibly meet with someone like you've only seen them online?” Like, who knows who they could be? Like, that's just the same as when you meet someone in other contexts, right? You meet someone at the bar, you don't know who they are. So it's really just a similar level of safety and danger online or offline.

SPENCER: I guess coming through a friend's recommendation, maybe it's somewhat safer.

TOM : Well, maybe.

SPENCER: It's really fascinating how the cultural shift on online dating has occurred, where I remember when I was young, it was literally viewed as weird to be on online dating. And now it's just incredibly normalized.

TOM: Yeah, that is such a good point. It's been a remarkable shift. And it's continuing. It's, you know, in different age groups and cultures and societies, it's pretty much across the board, it's moving towards being more accepted. And they're ones like older populations, for example, where it's still not the majority, but it's, but it's growing quickly.

[promo]

SPENCER: So I want to transition to another topic that you've had a lot of experience with, which is management, to love to hear some of your thoughts about what makes for effective management and effective leadership more generally.

TOM: Yeah, for sure. It's always interesting to see the resources that when you have to similar terms, like leadership and management, like all the different ways experts come up with differentiating them, and they kind of all don't agree, and then you're like, well, what do these two words mean? And like, how is it that suppose it experts, I'll have different views on it? It's like a fascinating little microcosm of language. But in any case, I think like at a high level, like leadership focuses more on the people and management focuses a little bit more on the process and kind of the tools involved. So I think like, one of the core tenants for me is I've focused on servant leadership, like throughout my career. And so what does that mean? It means that you're really there to support everyone else who's actually doing work. Like if you think about it, like the leader typically isn't doing the real work that counts that launches the product that moves things forward themselves. And really, they're there to remove blockers and improve quality of life and ensure that the rest of the company and the people on the team are mentally and physically, in every way possible, ready, be their best selves, and do their best work.

SPENCER: That makes sense. I would also add, there's a strong coordination mechanism that a leader has, which is making sure that like everyone who needs to talk to each other is actually talking to each other, that there's the right information flows, and that everything stays pointed at the vision that needs to go out instead of kind of everyone rowing in their own directions.

TOM: Yeah, exactly. And there's a trade-off between alignment and autonomy. That's always interesting,

SPENCER: Right. So it's like, you want to make sure everyone moves the product towards your, your vision or the big the broader vision, but at the same time, you want everyone to feel empowered to like, make their own decisions, and solve their own problems, right?

TOM: Yeah, exactly. Yeah. And it's not just about people feeling empowered, but people who are kind of locally close to the problems are actually have better information and are more effective at making decisions in those areas and someone who's in a different place than they are organization sending ideas or orders down. I think there really is a sweet spot there where you communicate alignment in general ideas and general goals. And I think a big tenant of this approach is providing context. So you're giving everyone in the company, the information that they need to make really quality decisions, without specifically saying what people should do.

SPENCER: Right, you should be passing along goal rather than a series of steps that they need to take.

TOM: Yep, exactly. And one of the ways I really like to think about it as your goal as a leader should be to make yourself unnecessary so that basically the organization could function perfectly well, even if you weren't there. Because everyone on your team has learned all the ways that you contribute value and has internalized all that you've taught and mentored effectively, and they're good to go. Of course, you never quite get there, but I think that's a really great way to look at it.

SPENCER: And so what do you do if you get to that point when [unclear] and they can do just as well without you? Guess you find some other way to help them be even better?

TOM: Yeah, right. There's always some way you can look ahead and find other challenges that are lurking in the shadows, had us thinking multiple moves in advance and exploring more options.

SPENCER: One thing I found really useful to communicate to people that I manage, is that a nonnegligible portion of my job is making sure that they are happy and effective. And like, by communicating that – that's like, this is my job to make sure you're happy and effective. I think that helps them rely on me more, because then if they're actually having a problem, they're like, “Oh, Spencer's job is to help me with this. Okay, I'm gonna ask Spencer to help me.” Whereas I think that the default assumption a lot of people have is that, your manager's job is to tell you what to do, not to be responsible for your wellbeing in the work, not to be responsible for your productivity and the work, etc.

TOM: Yeah, exactly. Yeah. And I think that's such a good point. I was in a leadership role early on in my career when I was at OkCupid. And I felt like in many ways, I was pretty inexperienced, and made a number of the kind of really rookie mistakes. And some of the challenges, you would realize, like, “Oh, well, someone's upset, I didn't phrase this thing right.” Like, what's the right technique? What's the right skill for communicating effectively and getting the right result from people. And it really took me a while to realize and reading a really good book, which I completely recommend for anyone addressing management or even just an improving kind of your own relationships in life. It's a book called Leadership and Self-deception and I think what's so fascinating about this book, is it makes the point that there's a missing ingredient, that without it, all of these techniques and skills you learn like a lead with a compliment, and then kind of give some feedback after that. Or look people in the eye when you talk to them, just, you know, with all these little techniques, you're like, “How do I become a better leader? I'm gonna learn all these techniques.” That without this ingredient, all those techniques are not that useful. And conversely, with this ingredient, all those techniques aren't done necessary.

SPENCER: Okay, so you got to tell us, what's the gradient?

TOM: Alright, so the key is, how you're thinking about the world and how you're thinking about the people around you. Are you coming into a situation where you're focused on proving something about yourself and holding up your own self justifying beliefs? Like maybe you believe something like you're the smartest person or a really good leader? And are you coming into an interaction kind of trying to justify that and seek evidence from that? Or are you coming into an interaction in a way that sees you and the other person, as people, and you both have needs, and they're both equally important. And you're here to work together. And with that ingredient, if people can see that you really respect them, care for them, or thinking about their needs, it makes all the rest of it so much easier.

SPENCER: So can you unpack that a little bit? So suppose that the person realizes that they have needs, the person they're managing has needs, what is this unlock for them?

TOM: Yeah, right. Well, I think what it creates is a bedrock of respect and mutual caring. And I think it's really helpful to kind of think about how you feel in different interactions. Like if you receive a piece of criticism or feedback from someone – if you're feeling a little bit defensive, feeling like they're judging you a little bit, and they give you that feedback; typically, I would respond in a defensive way, like that's, that's not true employee view. But if it's coming from someone you deeply respect and care about, and you know that they care about you, then it's an entirely different interaction.

SPENCER: Right, because you say, well, there's this person actually cares about me. So if they're giving me negative feedback, it's probably because I like I really have this issue and I could get better.

TOM: Yeah. And, like a very simple example. It's probably something most people experience. You know, you sit down on a bus, and it's starting to fill up with people and you're alone and you're like, “Oh, I kind of don't want someone to sit next to me.” So you'll put like a newspaper or your bag next to you on the bus. And so people come, they see the bag, and they look for other seats. And this is like a really good example of you're focusing on your own needs. And you're actually in a bit of a state of self-deception, because you are seeing your own needs is much more important than someone else's needs. Because obviously, your desire to not have someone sit next to you, is a much more minor desire need than someone else's desire to not have to stand on the bus, right?

SPENCER: Is it self-deception? Or is that just being selfish?

TOM: Right. And I think that's a good point. And I think it really is self deception. Because the way you come to the conclusion that that is a good course of action is by concluding that your needs are more important than someone else's, and kind of coming into it assuming that which is a distorted view. Actually, you're both people, you actually both have needs. Now, if you approach it a different way, and you sort of say, “Well, we both have needs, and they're equal, and understand that and I honor the other person, and I respect them. However, I get really anxious if someone sits next to me, so I'm going to put a bag here.” But that totally makes sense. That's really the key is it's about how you approach the action, and what your what your model is, and how you're thinking behind it, rather than the actual action you're taking.

SPENCER: So what do you think of this kind of idea of like a soft management style versus a hard management style, where you know, soft management style would be like, you know, you're very gentle with your employees. If you're going to critique them, maybe you sandwiched between two compliments, you know, harder management style, maybe you're just more like blunt, if you someone does a bad job, you tell them immediately, you don't try to soften around the edges. Do you think that one of those is like inherently better? Or do you think it really is contextual? Or based on people's personalities?

TOM: Yeah, well, I think one of the really big revolutions, for me reading this book and thinking about this approach is, is it's not as important to question because you can behave in a hard way, and still be deeply caring for the other person and be considering their needs the whole way through.

SPENCER: As long as they understand that you're looking out for them to care about them. Is that what you're saying?

TOM: Exactly, but I think what is so interesting about this is that you don't have to explicitly say, “Hey, I care for you, and I'm looking out for you.” If you're going around in your interactions, and you're sincerely feeling that care for the other person, it just comes through, it comes through in your body language and the way you say things, and people pick up on it. Just like you know, someone's like really angry at you, and they're trying to pretend they're not angry, you can kind of tell, right? This is the same kind of situation, even if someone's being super nice to you, you can kind of tell, maybe they're a little bit upset at you or don't respect you, and the opposite. Like even if they're being a little harsh with you, you can kind of tell what their internal state is beneath that surface. But I guess the really big point here is, that's what really matters is the way you're viewing the other person when you're interacting with them.

SPENCER: So basically, I think what you're saying is if you get to a state where you genuinely care about the needs of the people you're working with, the people you're managing, that will come through in your communication. And that matters less whether you make eye contact, or use a hot, soft style, or hard style. But basically, as long as they trust that you're like looking out for their interests, it will make all the leadership go better.

TOM: That's exactly right. So then you kind of get to a final question of like, well, what is a more effective approach, like being hard or soft, and I think I naturally tend towards soft like, I really worry about how other people are feeling and don't want to cause pain. That's the approach that I found works the best is starting with warmth. So you start out with the assumption that someone is really coming from a good place, they just made a mistake this one time, or they just need some care. And that's typically that's great. But then if a problem continues and repeats, then you get a little bit harder over time to help underline the seriousness of the situation and nudge towards them improving, which is really the goal. And like, you know, if you're given suggestions for how someone can improve, and it's super clear to them that they really care about that you really care about them. And that's quite effective and helpful.

SPENCER: What do you think about showing anger? I think people differ on this. You know, some people think, you know, anger is a useful tool, and some people think you should never show in anger in a work context.

TOM: Yeah, right. I'm a little bit fortunate there. Like personally, I tend to not get very angry or stressed out or worked up about situations, like I can tend to be pretty cool. But of course, on the other side that has downsides. If it's time to celebrate, I'm not really great at celebrating either. But that said, I think it's like not clear. One thing that is clear, though, is like coming into an interaction with repressed anger is not effective. So yeah, if you're feeling anger towards someone, I would think it's not super effective if it means that you're losing control. But if it's a way of underlining kind of how much you care about a situation and how much you care about a person, and I think it makes more sense. So I think it's really the kind of in control question.

SPENCER: One of the things I find interesting about anger as a part of communication, is that it communicates seriousness right? It communicates that you feel like someone is either doing something bad that they could have done better, or doing something that like blocks, your own goals, etc. And in some cases, it does change behavior, right. Like sometimes when someone expresses anger, the other person actually reacts to it and behaves differently in the future. But it's based on a certain limited view of what went wrong. So for example, if you know, a manager expresses anger, because an employee screwed up, that will change the employee's behavior if A, they know how to fix in the future, and B, they find that anger actually motivating to fix it. But it very well could be that the employee doesn't know how to fix it in the future. And so then the anger punishes them, but they actually don't, they're not empowered to do better next time. And now, maybe they're just even more stressed out because they don't know how they can avoid that future anger.

TOM: Yeah, right. And I think it's so important to have a strong relationship underneath all that. So that if you do express anger towards someone else, they understand that it's because you care about the situation about them, and that you're not just being a jerk. Because if it feels like someone is unfairly angry towards you, I think that's really demotivating. Like, it's not just demotivator, it can be scarring. It's really unpleasant and tense interactions, and it destroys trust with the person in the future. But if it feels like instead, the model of the person who is experiencing the anger is that, wow, I know, this person respects and cares for me. And what they're showing me is that this is really important. Whatever it is, I didn't realize this was so important. And now I'm wondering if this is important, better than nothing can be really motivating.

SPENCER: That's a good way to put it. I would also add that I think it can be very iffy, when a manager jumps to blame too quickly because a lot of times the manager doesn't really see what's happening underneath. So for example, you as a manager might become aware that this person didn't do the thing that they were asked to do. But on the other hand, what you might not see is that they were actually blocked because someone else didn't do their work. Or maybe the actual request that you gave them was actually ambiguous, and they misinterpreted it, but actually is your own fault. So I think there's really a lot of value in kind of holding off on blaming someone and really investigating what actually happened and trying to understand it. Because a significant percentage of the time, it's actually not the person's fault, and then blaming them for it can be just very demotivating and very unfair.

TOM: Oh, yeah, that is such a good point. Yeah. And in fact, I'm sure you've had that experience. I've had the experience of kind of looking at a situation saying, “Look, you mess this up, it really wasn't fair.” Like they didn't mess it up. It was for some other reason.

SPENCER: Totally.

TOM: I feel terrible after that. And you can tell, it's like when did the relationship a little bit and there's a lot of work then to put into to rebuild that relationship. And so, and I think that's why I tend to not go the anger route, or the blame route. And I think blame is kind of a different concept we can talk about a little bit, but like going the anger route, it can be risky. And so I'd rather not take that risk for the most part. So yeah, my preferred approach is really to when something goes wrong, approach it with curiosity. And again, it's so key that the curiosity is sincere. Like, imagine your manager asking you, “Why did you mess this up again, dude?” Like, you can tell it's not serious. Like it's demeaning, and you're not going to respond well to that. But if it's sincere curiosity, then it's like a chance for you to explain them and that feels really good.

SPENCER: Absolutely. So what about hiring? What do you think people often do wrong when hiring and what do you think they could do to make it better?

TOM: Yeah, for sure. I mean, hiring is such a fascinating topic. And the reason is, I think it's the most important thing an organization does, really any organization like government, business, nonprofit – like put it in the words of the book, Good to Great, – deciding who is on the bus, like who is going the same direction that you're going with you think is the most important decision that a company makes. Like if you have the right people, then they can solve the problems and help set the right direction for the company. So it's like the most important thing. And then also, there's really not very good science behind it, right? You have these strange results, like the Google study that the test results that they had in the hiring process, were not predictive of later performance in jobs. And yeah, so it's just so strange, that's so important, and yet so poorly understood.

SPENCER: One thing I would just say about that Google result. First of all, it's kind of shocking. But second of all, I think we should downgrade our surprise a little bit because they're hiring very few of the people who are interviewing. So it's not saying that their process doesn't weed out a bunch of bad people. It's just saying, by the time you pass through the whole process and get accepted, there's no additional correlation last.

TOM: Yeah. And that's exactly right. Like, that's such a fascinating situation where a system is designed to solve one problem, and then you measure the effectiveness of that system on an entirely different metric. So right, I agree. You would usually expect something like a linear relationship or something.

SPENCER: It's still strange little disturbing.

TOM: Yeah, exactly. And I didn't dig too much into that study. I don't really know if the methodology is great or not.

SPENCER: Well, the fact that Google released it made me trust it more, because it doesn't exactly make them look at.

TOM: Yeah. And it's so fascinating. Like when you look at the way different companies hire, like hiring strategies are radically different. So I can talk a little bit about the approach that I've taken. And unfortunately, it's not particularly evidence-based, you know, I haven't been running an RCT or anything. It's really just through personal experience over time, which is flawed. But one of the things I find really fascinating is, once applicants have come in the door, you have resumes to review. And in my experience, resumes are where most of the people are rejected at this point. So it's kind of the most discriminating step. And yet, I think there's a pretty universal agreement like this is talked about and work rules by one of Google's I forgot what role he had.

SPENCER: It was like kind of people management, or whatever you call it.

TOM: Yeah, and other resources, that resumes have almost no information and no predictive power over whether someone's going to be a good hire.

SPENCER: I had a funny personal example of this, where one time for a role, I rated all the resumes coming in. And I also had another person who was in that same role, write them all. And there was essentially no correlation. I was just like, this is ridiculous, like this is not to say that it doesn't mean resumes are ever predictive, I think there are some jobs, where you really do need to have certain kinds of experience to do them well. And in those cases, resumes can be good as like a filter, like, okay, the person has the relevant experience that we really need them to have. Okay, that's like a starting point. But I think for a lot of jobs, they just have really low correlation with what you'd want.

TOM: Yeah, and it's just such a weird contradiction, that there's this really big filter at this point where almost no information is known. And yet, it's hard to get around. And one of the techniques I found for getting around it is by – I mean, so the kind of hiring that I do is focused on software engineering. And so what you're really looking for is someone who can do really effective work, they'll build good software systems communicate well, someone who's kind, etc. There's a kind of a list of factors that I would say are important. And central to that is that they're able to get stuff done. So you can devise a relatively short test that tries to mimic as closely as possible, what the person will do day to day, maybe on a kind of like a little bit of a harder day at your job, and then create a test that aims to measure their effectiveness at that, and then send that to a wider audience. So you do very little screening at the resume level, and you send this stuff out to a wider audience, and then you really get into a measure for their skills, rather than what they're they're claiming on the resume. I found that's really effective.

SPENCER: Right. So I hear that, and it's called a work sample test. And I think the key there is the, as you mentioned, that more closely resembles like the actual work that person would do in real life at the job, the better predictor it is. It make sense. It's like, oh, you want to know if they're good at x, like have them do X, right? And it's also really good, I think, to make it standardized. So that because otherwise, you could get a situation where like, one candidate gets like a harder work sample another, it gets easier. And then it's not apples to apples comparisons. Personally, I like to use work samples tests. I like to make them a little bit longer, but pay for them. So basically, it's like a paid work sample. And maybe it will be like in their spare time over two weeks or something, but they're getting paid for it. So it's not that big a deal if like, doesn't work out, and then they understand that they're going to be evaluated based on it. And just another thing I'll add about that is, I find that it's really good to be super clear about what you're looking for because the question is not will this person somehow guess what you're looking for? It's like, can they do what you're looking for when they know what it is, if that makes sense.

TOM: Yeah, it's it shouldn't be kind of a guessing game. And I think that's such a fantastic idea. Like one of the big weaknesses of that approach, especially if you've got like a shorter kind of work sample, like 30 minutes or an hour is that the person is under time stress that isn't really realistic. Like that may make her a few times in their work, like really, really to go to fix out quickly or something. But for the most part, that's not a good reflection of their day to day work. So yeah, if you can do it over an even longer period. In fact, my favorite interview is a 30-day contract period. But of course, it's kind of the best candidates and potential teammates are typically not interested in that kind of arrangement.

SPENCER: You basically hire them for 30 days, but then with the idea that can be re evaluated at the end of it?

TOM: Exactly. Yeah. And of course, you need to do some, a little bit of filtering before that, because there's always this balance of how much time do you and your team invest in the hiring process versus kind of the benefit you'll get out of it and the extra work that the person will do for your company?

SPENCER: From the literature I've looked at on what seems to best predict job performance across a wide range of jobs – not in super specific demands, but you know, across many, many jobs – the things that come up again and again in these studies are IQ, conscientiousness, which means that personality trait from the Big Five personality test, performance on work sample tests. And then you have like a bunch of like, much weaker factors, like, you know, interviews work a little bit, and you know, maybe resumes work a little bit and things like that. But it's kind of fascinating, the things that come up again, again, are those, you know, three – IQ, conscientiousness, and work samples. But one thing to say about conscientiousness is actually it's really hard to measure because you might think, oh, just give them you know, big five personality test. But actually the questions on it is super obvious what they're asking about. And I think even a quite honest person, if you say to them, you know, “Are you disorganized?” They're on a job application. They're gonna be really reluctant to say, “Yes, I'm really disorganized.” Or “Yes, I'm not punctual,” or, you know, things like that.

TOM: Right. Yeah. It seems like it may not translate from whatever environments it was validated into interview context.

SPENCER: Yeah, a lot of times, they're validated in low-stakes environments where there's nothing on the line. And another problem with that is that the most like sociopathic people are willing to just completely lie, because it's so obvious what they're looking for, they're gonna actually have a boost. So you're gonna actually like, give advantage to the liars? Which seems like a bad incentive to create.

TOM: Yeah, exactly. You know, I find that list interesting. Like two big comments, for me. One is, obviously IQ is a kind of a problematic measure. There are some pretty unfortunate correlations of socioeconomic status, and ethnicity, and some very real reasons that people in different cultural backgrounds would just be fundamentally at a disadvantage. Even obviously, they're equally intelligent, the tests would not reflect that as much. And so I actually believe it's, I think, in many states, that's not acceptable to use an IQ test in a job interview. And I think related to that, there's also a little bit of a risk of a conformity trap. So this is like kind of looking for people similar to yourself in interviews. And I think it's like really interesting to design interviews in a way that avoids this. And to kind of be more clear about it, most people feel look at themselves and say, like, “I'm successful, like, I'm interviewing people for this role.” This is what success looks like. It's this thing and that thing that I do, and I'm going to look for those traits and other people, when in fact, success may look very different from kind of the way that the person is judging, acts and behaves. That's a really interesting challenge and an important thing to be aware of in the interview process.

SPENCER: And I think likability plays into it to a great degree, where of course, it is important that you like the people you work with. You don't have to like want to be their best friends. And I think in job interviews, people are extremely influenced by the likeability of someone and sort of the person's social skills. And I think this creates actually a lot of biases in hiring that lead people to hire people who are really not the best for the job.

TOM: Yeah, that's such a good point. I found that surprising that something about not exactly likeability, but three things you pointed out: IQ, conscientiousness and work sample result.

SPENCER: Yeah, in a lot of these studies, exactly.

TOM: What's interesting about that list is, is that I don't see anything about interpersonal skills there.

SPENCER: Yeah, it's a really interesting point. I mean, I would say that there are jobs that really interpersonal skills are key, right? Like if you're someone who has to work with customers a lot, right, things like that. But then there's also a bunch of other jobs, where just interpersonal skills matter, just dramatically less.

TOM: Oh, that's so interesting. I really come out thinking that that's an essential ingredient. And to be more clear, what I think is an essential ingredient is that someone is kind. And when they come into a work environment, they're there not just for themselves, and to kind of make their life better and accomplish their goals. But they're really there for the people around them. And they're collaborative. And they support the mission of the company and the needs of their teammates. It's hard to find a role where that's not important, unless the person really just doesn't have to collaborate or interact. But I think like any sufficiently complicated, like, certainly software system, there's a tremendous amount of collaboration that goes in and communication. So I think that ends up being super important, although it's really hard to measure, as well.

SPENCER: One of the interesting challenges is, you might have someone actually makes the whole team better, but they're not necessarily on an individual level, going to have higher job performance by whatever metric you use. And so maybe these personality traits or skills that help the whole team don't necessarily get factored into the studies.

TOM: Yeah, I completely agree. Yeah.

[promo]

SPENCER: So let's switch topics again. I want to talk to you now about self-experiments. So tell me a bit about how you use those in your own life and what your approach used to them.

TOM: Yeah, sure. Personally, I think it's different for different people. I think self-control comes easily to some people; harder to others, like it comes harder to me. And so I really found that, for me, lasting behavior change is pretty much the biggest challenge I face in terms of self-improvement at least. Do you feel the same way Spencer?

SPENCER: Hmm, I'm not sure that that is the biggest challenge I face. I feel like for me, a lot of it is trying to figure out what strategy works for some tricky to solve problem. And so I might need to like throw 10 or 20 strategies at it, but it's less about changing my own behavior and more about like, what actually produces this outcome. I'm really not sure about it.

TOM: Yeah, that's super interesting. So I think I think you're in a place of like, kind of more self disciplined than I am. And yeah, it's like, so fascinating to see the different ways people behave and are configured. So one approach I found really helpful for me, is to run regular experiments, and kind of having this regular cadence helps me have that self discipline challenge. And so maybe I'll do one a week. And so as I kind of lay out the steps of the process that I use. The first step, I'll set the goal. So identify something about myself, that has been troubling me the most recently, or this week, that seems like kind of the highest value to improve. And so like, one thing I've been focusing on recently, personally, is that I find I often distract myself with Slack or email. So it'll be like focusing on some important thought, heavy task. And then I'll just be like, “Oh, that's hard.” I'll go check my slack real quick, and make sure it's not a message I missed.

SPENCER: Got it. And so how do you think about designing those experiments to be as effective as possible?

TOM: So once I set the goal, then I'll look at a few different possible interventions, but also, I go a step deeper. Because what I've learned recently, and I think is really interesting here is often, when you're behaving in ways that are a little bit self-undermining like that, I'm accomplishing less, I'm creating less value due to this behavior. It's not just about tactical changes and thinking, maybe I'll close my tabs or something. It's not just about that. It's about what big assumptions, and what was my story and thought process was leading me to have that behavior. And so I found it to be quite helpful to dig into that behavior and understand the assumptions a little bit better.

SPENCER: Can you give an example?

TOM: Yeah, sure. Well, so with my distraction, with Slack example, one of the assumptions that I think is driving that is the feeling that I need to be there for other people. You know, like, someone has maybe asked me a question, and they're blocking, waiting for my response, or maybe a sudden meeting came up, and I'm not going to make it because I'm not responsive, and on the set on something important. So I worry about those things. And I kind of like worry about how other people are perceiving me as a result of not engaging with Slack. And so that tempts me to engage more.

SPENCER: Yeah, I think illustrates a really important principle, which is often overlooked, which is that when we have “bad behaviors” that we find undesirable that we want to change, very often, they actually are serving some purpose for us. And I think to deny that and just say, “Oh, I'm just being dumb, I'm being bad is actually a problem,” It's actually better to first recognize what is this doing for me? What purpose is this serving? Is it helping me alleviate stress? Or is it distracting me for something I don't want to think about, or what have you? And then once you've kind of identified that that can help you figure out what to do about that?

TOM: Yeah, exactly. And I think what is so cool about this approach once you've identified the assumption is it can be really tricky to root out these big assumptions. Like they're kind of fundamental to how you behave. And they're so deep in there, they're like driving your behaviors, and you don't even realize it. And so it's a long process to engage with them. And one of the most effective ways I think, to engage with them is to generate and examine evidence that contradicts the assumption. And so that can drive the design of the experiment. So for example, one of the first things I did when looking to address this self-distraction problem was, I just blocked out periods of time on my calendar, and decided to close my tabs, and I wouldn't interact with email or slack during this period of time. And I laid out what did I what I would collect. So I just write down each time I did it, how it went, how I felt. And critically, this is what happened during this period of time. Did it support the big assumption? Or did it contradict the big assumption?

SPENCER: And what was the big assumption here that you were checking slack because you felt responsible for people?

TOM: Yeah, that's fairly personal. It's like, if I'm not there for people, unresponsive people may think that I'm not doing my job well, that I'm not supporting them. And like, yeah, it's kind of like gets into self-worth a little bit, like viewing myself as not worthy. So yeah, the question is, “Well, after I do this, I don't respond for a while, what are the actual consequences? What happens? Are the fears that I'm predicting happening? Is someone really upset with me, for example, that I wasn't responsive?” So, I did this for a week. And like, kind of unsurprisingly, I found that not responding for a few hours on Slack, had no serious consequences. People weren't upset for having to wait for a little bit, especially because I was able to set some expectations with the calendar event that people could see.

SPENCER: Yeah, I like that a lot. It's similar to a technique used in cognitive therapy, where you have people predict what the outcome will be if something that they're avoiding. And then you basically get them to agree to do an experiment where they try the thing out, they predict in advance what's gonna happen, and then after they evaluate against their predictions. And I may work especially well when the therapist thinks that the person's explanations are unlikely to be met. And so basically, the behavior is being reinforced by sort of a false belief that's preventing them from doing it.

TOM: Yeah. I mean, what's interesting here, too, is it's also okay if the assumption is correct. I mean, it may be that you learn that it's actually a very important behavior. Or maybe you learn something even more fundamental, like, maybe not about yourself, but about your business, like, oh, we do a lot of surprise meetings. And actually, that's stressing me out. And that's probably stressing other people out. So that's the root cause to dig into.

SPENCER: Yeah, interesting. You mentioned self-worth, because I think a surprising number of “bad” or “negative” behaviors have something to do with your sense of self-worth, obviously, not all of them. There are lots of reasons we do things. But one thing that I found really fascinating is, I ran a study where we collected tons of different measures of how happy a person is. So it's basically just asking the question, like every way we can think of, you know, happy or with your relationships, you know, how happy are you with the environment that you're living in? How satisfied are you overall are you with your achievement? Like all kinds of things. And then we asked this kind of statistical question, which is, “If you could only use one of these -” like, I think it was over 90 questions, right? “If you could only use one of them, and you had to use that one question to predict all the others, which one question would you use to best predict all the others?” You want to make a guess? It really surprised me with it with the answer.

TOM: Yeah, what's the list of questions?

SPENCER: This is everything you could possibly imagine asking you about how happy you are. I was like over 90 different questions about how happy you are.

TOM: Yeah, I don't have a great guess. I'm, I'm super curious, though.

SPENCER: The question was, to what extent do you feel confident and positive about yourself? So that result really got me thinking more about how a sense of self might be really core to our happiness in a way that I hadn't previously appreciated?

TOM: Yeah, that's a really interesting point. It kind of gets back to the leadership conversation, too, about whether someone was going around in life, focusing a little bit more on finding reasons to support their inner story of self-worth, like, “I'm really smart,” or “Everyone likes me,” and then blaming people who contradict that story, versus going around the world, less distorted way, perceiving people as equally valuable.

SPENCER: Alright, Tom, so I want to do a final bonus topic with you, which is aliens. So tell me some of your views on aliens? Do you think that aliens exist in the universe?

TOM: Alright, cool. Well, yeah, this thing just falls into like the important ideas category or surprising ideas. This came from The Three-Body Problem book. Did you happen to read that?

SPENCER: No, no, I haven't read it.

TOM: Okay, cool. I recommend it. It's really interesting, kind of sci-fi story. It's pretty dark. But what's fascinating about it is it's written by a Chinese author and then translated into English. And it's been a pretty big hit in the US. I think it's one of the only examples of that kind of cultural flow, which I think is really cool. Yeah. So it introduces this idea of Dark Forest theory. And so let's kind of back up like, what does this refer to? So, like, one of the big questions everyone wonders is, are aliens out there? And one of the attempts to answer that question, is the Drake equation.

SPENCER: Would you break it down for us? What's the Drake equation?

TOM: Yeah, so I'm not going to list like all the factors, but it's kind of like the multiple a bunch of different ratios that in the end gets to the number of alien civilizations, you'd expect to be transmitted radio waves out in the galaxy, or the universe. And it's just kind of the product of a bunch of factors you would expect.

SPENCER: Right number of civilizations with which humans could communicate, that's like what trying to calculate and then it's like, takes into account things like the mean rate of star formation times the fraction of stars or planets times the mean number of plants that could support life per star with plants, etc.. Basically,it kind of breaks it down to the series of products that you multiply together, right?

TOM: Yeah, that's exactly right. And of course, like a bunch of these factors, especially ones like given that you have a habitable planet, what fraction of those habitable planets will actually develop life? Like, that's a very hard thing to estimate. So it's really only an approximation, but different groups have plugged in different plausible numbers and you come out with you a usually significantly higher number than one. And that's been especially true recently, as we've discovered more planets around stars than we thought.

SPENCER: I number higher than one that suggests that there should be a civilization of aliens that we can get in contact with, right?

TOM: Exactly. And so I think it's been really interesting recently seeing the results from these planetary surveys where you're finding that there are quite a few planets per star, and there are actually quite a few habitable planets per star that have planets potentially habitable. I mean, the evidence is pretty uncertain still, but you know, they can detect the signature of water, for example, on these planets. So that's really kind of increase the lower bound, like there's some pretty skeptical estimates that really assume that planets are very rare, and habitable planets are very rare. And so recently, we've been realizing, “Oh, they're really not that rare.” And so it kind of pushes up the lower bound on the Drake equation. You really get to the place where you're like, “Hmm, it seems pretty reasonable to think that there's life out there.” And in fact, that from a number of perspective, that seems likely this is called the Fermi Paradox, like life seems pretty likely. So why are we not seeing evidence of it out there?

SPENCER: Yeah, it's a great explanation. I'll just mention two kind of interesting things about this. Often people kind of just plug in a single number for each of these factors going into the Drake equation, which kind of just produces one number in the output. But kind of a more robust way to do this. And I've seen one paper doing this is you actually like assign a sort of a probability distribution to each of the factors in the equation, because each of them is actually very uncertain. And then you kind of multiply through this uncertainty to produce a distribution outcome. But I think you still end up with, you know, a significant chance, even if you do it that way, that there should be life out there. The second thing I'll just say is that I think one response to the idea of the Drake equation is that it applies this sort of the universe is kind of a fixed thing, where you know, every planet that develops has a certain fixed probability of producing life on it, because it can also be screwed up by kind of anthropic reasoning, where basically like, for example, suppose there were lots and lots of universes. I'm not saying that's true. But suppose it were true, then the only time you would have beings like looking around the universe being like, “I wonder if there are aliens here” would be in whatever fraction of universes actually produce life in them. So you could get a weird situation where actually life is just incredibly rare. But people only ponder it in the universes that haven't had it.

TOM: Yeah, right. And that makes perfect sense. I don't think the Drake equation is really that vulnerable to that, but certainly, like, any kind of reasoning that goes like, “Oh, yeah, humans are here. We happen. So can't be that unlikely, right?” But I think that's pretty fun.

SPENCER: Go on. And I want to hear more about your analysis.

TOM: I don't really have any big insights on those points. Like, I don't think it's kind of like a tantalizing question. Like, why isn't life out there. And I think you might be more likely to believe that it should be out there or not. But in any case, I found this this Dark Forest there to be just a really interesting take on it. And it kind of works from like, a few axioms to the conclusion that you would not expect to see life in the universe. And Dark Forest is kind of a metaphor, before I get into the logic of a number of if you can imagine a number of species in a dark forest, and it's super dangerous. And if any of them makes the sound or moves, another will hear and immediately attack. That's kind of the metaphor. So why would you think that the universe will be like this, or our galaxy anyway? And so you start out with a few axioms. One is that survival is the primary need of civilization. Another is that civilizations grow and expand, but the amount of resources in the universe are finite. So as you can see, these two set up conflict, one civilization success is other civilizations failure to some degree. So then you get into a little bit of game theory. So you imagine there are two civilizations that find out about one another, and then they kind of know these axioms about the other civilization. And so the question is, is this other civilization going to destroy ours? And you're not sure. And what's really concerning is like, even if they're nice and benevolent, like maybe your benevolent and they're benevolent, that kind of doesn't matter. Because if they think you might be malicious, then they're going to act first to try to destroy you before you can destroy them.

SPENCER: Reminds me of the kind of logic of nuclear weapons. It's like you're mutually assured destruction, right?

TOM: Yeah, exactly. So you get to this very concerning situation where you just, you really can't convince yourself with an alien civilization, that they're not going to destroy you. And this leads you to the inevitable conclusion that you need to throw them first before they destroy you. Because the cost of them obviously destroying your civilization is so high. And even if you have a low probability of it, you still need to act. Now added on top of this, there's a few challenges or a few things that make this even worse. One is the travel and information is very slow and states, this makes it very difficult to build trust, it also makes it very difficult to mount an attack quickly. And then you also have another pattern that we've observed that seems to hold pretty true, which is that technological capability of a civilization increases exponentially after a certain point. And so what this means is that if your civilization was just discovered another civilization out there, they are contemplating what to do, you look at the civilization, you say, “Well, they're very primitive right now. So maybe they're nothing to worry about.” But you can't ever convince yourself of that, because technology expands exponentially. And it would take, I don't know, 100 200,000 years for, let's say, your civilization does something like, captures a star and washes it out the other solar system to hit their star and cause a massive reaction or some attack like that. If you're contemplating whether to do this, even if the other civilization looks super primitive, by the time your attack arrive, they may have advanced so far, and so much because of this exponential increase in technology that they can deflect your attack, and someday even more effective one back at you.

SPENCER: You imagine like a distant civilization that has to send an attack across a huge number of light years, basically?

TOM: Yeah, right. Like, you know, it's a civilization that's 50 light years away from Earth, and it just received our first radio broadcasts from the 1938 Olympics.

SPENCER: It's interesting to think about those radio broadcasts are kind of expanding, and it's fear around our planet. And you know, every year there they go, you know, a bit further out into space, where another kind of civilization if it were there might be able to detect them.

TOM: Yeah, exactly. So because of all these factors together with civilization, even though it sees all these broadcasts are super primitive, they have no idea where our civilization will be by the time their attack arrives. And so in fact, delaying even one day may make the difference between the destruction of your civilization or not. So what it means is that any inferior civilization you encounter as a civilization, you need to destroy it immediately because you're in grave danger. And any superior civilization you encounter, you need to hide from them at all costs, completely mask any signals.

SPENCER: Interesting. So other civilizations are either already destroyed, or they're hiding, basically.

TOM: Yeah. And so then this, this kind of brings us to like the Dark Forest where there's a bunch of civilizations out there, maybe, but they're all hiding really, really hard. And that's why we don't see any. And it's also pretty concerning, right? Because we have not been hiding at all.

SPENCER: Yeah, I find it really interesting, the debate of should we try to contact alien life? And I think that there are some scientists that actually have attempted to do this, like broadcast signals, specifically trying to contact it. And you know, maybe the argument is, well, if they're really that advanced, maybe they'll, you know, be very nice, have a very evolved ethics or something like this, but also seems just like an extreme risk, like, we could argue about the probability that our signals would ever reach aliens, but let's assume they did reach aliens, it's like, it seemed just incredibly risky. Because if they're able to detect our signals, it seems just extremely likely, there'll be much more powerful than us. And then we're basically sort of at the mercy of like, how nice are they to us?

TOM: Yeah. So it kind of raises a question like, should we really be concerned about this and should be really be making similar to how we're thinking about the chance of an asteroid impact on earth? Should we be trying to protect against this danger?

SPENCER: Yeah. It's fascinating to think about, I ran a Twitter poll recently. I was just curious about what percentage of my Twitter followers actually believe that aliens exist somewhere in the universe. 84% of people said they think aliens exist, and only 16% said no. Which kind of surprised me because I feel like there was a time when the idea of aliens existing was like, really far out. And I feel like maybe it's now become more accepted that like, well if the universe is a really big place, why would we be so confident they don't exist somewhere?

TOM: Yeah, that's fascinating. Yeah, it makes me wonder why would that be? Is it the particular people who follow your Twitter? Or is it just kind of so many movies and popular media that has aliens? That's kind of like, yeah, of course, that's how it is!

SPENCER: I don't know. It's a good question. And then there's, of course, there was a long period where people kept having UFO sightings and this sort of thing. And you know, there would be these people who say, “Oh, no, I have really got, like, abducted by aliens.” And then, of course, the skeptics would come in and say, “Okay, this is all bullshit.” And so we went through like decades of that. And I feel like now, we don't hear about those cases anymore. So much like they're kind of much less of a big thing. But maybe there's more just general acceptance that like, “Oh, yeah, they're probably out there somewhere. They just haven't arrived on Earth yet.”

TOM: Yeah, that's an interesting thing that those reports might have delegitimize the idea and so prevented, “real scientists” from thinking about it very hard.

SPENCER: Tom, this was really fun. Thanks so much for coming on.

TOM: Yeah, absolutely. It was such a pleasure, Spencer.

[outro]