HILARY GREAVES: (00:03) Thanks very much to all of you for coming and thanks in particular to Professor Ingela Alger for delivering today's lecture. So I'm Hilary Greaves. I'm the director of the Global Priorities Institute, here in Oxford. I'll just kick off by very briefly saying a little bit about GPI and then I will introduce our speaker.

(00:22) The Global Priorities Institute was established here in Oxford in 2018. Our mission, if you like, is to produce and to facilitate that other-produced research that's especially crucial for agents who are completely impartial and who are trying to do the most possible good subject to the fixed resources they have under their control. So the question is, if you've got some pot of let's say, money or it could be time and you want to deploy it in such a way as to do the most possible good and it has to be completely impartial between everything that is appropriate to be impartial between, so time, species, people, geographical regions, you name it, then what's the best way of doing that? So this project is of course, in principle, an extremely big question. But the reason GPI exists is that we think, despite the scale of that question and the number of issues it raises, there are ways you can make real progress on it by carefully deploying the tools of various academic disciplines. And here at GPI, we focus in particular, for the moment at least, on economics and philosophy, the two disciplines in which we have teams. And at GPI we have a particular focus on the further future, so we're especially interested in the question of whether there are sufficiently tractable ways of influencing, nudging the course of the very far future in positive directions. So we want to know what are the main challenges to doing that thing, what research would help to evaluate those challenges and help us to decide whether it's even worth trying to do that thing or whether it's too intractable, and insofar as one is trying to influence the course of the very far future, what research would help guide one in one's attempt to do that. So this leads to among other things to the purpose of the lecture series that brings us here today.

(01:58) The series is named in honor of Professor Sir Tony Atkinson, who as most people here I imagine will know was a towering figure, both here in Oxford and internationally. Atkinson's work focused in particular on the notion of inequality and also more generally, on issues to do with increasing welfare. So we're very honored to be able to host this distinguished lecture series in his name and we're very happy to be back here in person for the first time since 2019, for the lecture.

(02:25)  On a logistical note, this lecture's going to run for one hour and there'll be 25 minutes of Q&A at the end. Please confine any questions during the talk to essential clarifications and reserve substantive things for the end. Please be aware that the talk will be recorded for posting online, so among other things, if you do raise a question during the lecture itself, that will be possibly online recorded.

(02:50) Okay. So without further ado then, let me introduce today's speaker… I'm sorry. For the people who are online, there is a link that you can use to raise your questions at the end or alternatively, you can use the raise hand function. You can also use the raise hand function to indicate that you'd like to ask for clarification during the talk. Did I get that right?

ROSSA O'KEEFE-O'DONOVAN: (03:09) Sorry. Online you have to submit questions using a link. And if you're in the audience here, you can use that link as well, if you prefer to type a question, and then I will read out the question at the end. And if you just want to ask a question in a regular way, just put your hand up if you're in a room. Thanks.

HILARY GREAVES: (03:24) Okay. So without further ado then, Ingela Alger is a CNRS Senior Researcher in economics at Toulouse School of Economics. She's a CEPR Research Fellow. She is also the director of the Institute for Advanced Study in Toulouse, that's the IAST, and this year she received the CNRS silver medal, which is awarded to a small number of distinguished scientific researchers in France. Her research focuses on the long-run formation of human preferences, in particular, when these are transmitted from generation to generation and are subject to selection. She's particularly interested in preferences that might explain moral and altruistic behaviors, as well as behaviors within the family are topics that she'll touch on in her talk today. Ingela is going to be presenting on the Evolutionary Foundations of Morality and Altruism, Recent Advances. I hand over the class to you.

INGELA ALGER: (04:12) Thank you.

(04:21) Thank you so much for the invitation to GPI at Oxford. It's a pleasure to discover what GPI is about. I mean, it's a young Institute, but I've already heard of it several years ago, so it's a big splash already and I think this combination of economics and philosophy is very interesting. So just to tell you a few words about that. It's also an Interdisciplinary research institute and I think that indeed, many of the important questions today require an interdisciplinary approach. And so, I look forward to future interactions, not just a few months with researchers like I've had at GPI especially because we do not have a lot of philosophers coming our way perhaps in terms of future...

(05:23) Yes. So, Tony Atkinson... This is, I think, the most that I got to him. Actually it was at... I was giving a talk at the World Bank and a day before that Tony Atkinson was talking, tackling the challenges of the 21ˢᵗ century, remotely, however. And then suddenly, he passed away a year later. So of course, I'm extremely honored and... Yes?

FEMALE SPEAKER: (05:51) Sorry…

INGELA ALGER: (05:53) Yes, I'll try. Okay. So it's very humbling to be speaking in honor of Sir Tony. So into the talk. So economists are essentially in the business of formulating policy recommendations. So and what is the goal of these policy recommendations? The goal is simply to try a way to maximize human welfare given some constraints, given the available resources. Now of course, if you ask a man in the street about how we can even proceed if we want to achieve this goal, it's an extremely daunting task. Where do you even begin thinking about answering this question?

(06:41) And so fortunately, in economics we have a very powerful set of results and a powerful paradigm to stand on to try to answer, to bring an answer to this question. As we know, in the 1950s, we have the first welfare theorem, which states sufficient conditions for a decentralized market to bring about an equilibrium outcome that is efficient. So, if markets are complete, if there is perfect information, perfect competition and if information is healthy, that any market equilibrium is Pareto-efficient. And so standing on this benchmark then for the next few decades, economists got very busy trying to understand how or which kind of inefficiencies would arise if these assumptions were not satisfied. So in these decades, we saw many, many public policy recommendations. So recommendations on policies that seek to mitigate inefficiencies that stem from incomplete markets, market power and/or information asymmetries and imperfections. So one remarkable feature of the first fundamental welfare theory is of course, that individuals need not want to achieve the greater good, it is achieved even if individuals are perfectly, materially self-interested. This goes back to the famous butcher, brewer and baker sentence of Adam Smith. However, we also know that Adam Smith, before he wrote "The Wealth of Nations", wrote "The Theory of Moral Sentiments" and it's clear that he did not believe that individuals are always purely selfish. However, he realized that it was sufficient to bring about efficiency and intuition and which was formalized in the 1950s. So, he understood that human nature is much more complex than selfishness and this was also true for many scholars in the 19ᵗʰ century, early 20ᵗʰ century.

(09:16) However, in formulating the first welfare theorem, the complexity of human behavior was somehow pushed to the background. And so, to some extent, I think that we should think of the first welfare theorem as having this 4ᵗʰ condition, which is that individuals care only to some extent, only about their consumption utility and utility from leisure activities. And so, this led economists to formulate policy recommendations that were mostly based on material incentives; if individuals care only about maximizing income and maximizing the consumption utility derived from income, then material incentives seem to be sufficient and perhaps also necessary to induce correcting behaviors.

(10:13) But if we really want to have a sense of what are desirable policies or you want to reach the goal of maximizing human welfare and if we want to identify policies that are effective, not backfiring for example, then we need to have an accurate account of human behavior.

(10:36) And as we know, as you probably know, for the past 20/25/30 years or so there has been a large number of alternative roles of humans that have been analyzed, proposed. So going from "Altruism" by Gary Becker, "Warm Glow" proposed by Jim Andreoni and so forth. We have a long list, a menu of such alternative preferences that can generate either pro-social behaviors or conformity and so forth. So many of these theories were inspired either from psychology or sociology. A lot of experimental evidence has been built to test some of these theories. However, I think it's fair to say that there is not yet a consensus of what a policymaker should expect here. An economist works to make a policy recommendation and also state, "Okay, I recommend this policy because the distribution of preference in the population that you are governing is this distribution."

(11:56) But I don't think that there is a consensus on what the distributions would be and how we could actually be able even to accurately measure them. So, what I'm going to talk about here is that we are seeking to go one step beyond this literature by asking which preferences should we expect from first principles? So ideally, what I want is that such a theory should shed light on which preferences are more plausible than others so we can try to sort of narrow this set of hypotheses that I think are plausible. And ideally, we should also be able to understand... It should help us understand why we should expect this and that preference, so the mechanisms behind this.

(12:54) Okay. So, longtermism at GPI we've heard about, so instead of looking forward however, I'm going to peek backwards in time and maybe potentially several million years. So homo sapiens is where we are now and who knows what future is reserved for us. But we are standing on, say, the shoulders of many, many millions of ancestors and so, this is a fact. And so, what can we say about how this evolutionary history... Has it shaped us? What does it mean?

(13:44) So we're going to use evolutionary logic to formulate this theory of an endogenous preference formation. So what is evolution? It's simply competition for survival and for reproduction. So not all individuals who are born survive and not all who survive reproduce. And so, the Darwinian logic is very simple and it's implacable. Those alive today, we have all ancestors who were successful at surviving and reproducing. It's kind of mind boggling. We have, each of us we have billions of ancestors who go far back. So our preferences should reflect this. And so, we're going to use this simple logic to develop a theory for the evolutionary foundations of preferences and in particular preferences that are relevant for economic policy.

(14:47) Okay. So according to evolutionary theory then, reproductive success is the name of the game. So well then, shouldn't it simply be the fact that humans should be expected to be equipped with traits that make them maximize their own reproductive success? That will seem to be perfectly logical and salient, right? So end of story. So that's the main theoretical challenge of this literature is to answer this question. Is it the case that evolutionary logic should lead us to simply seek to maximize own reproductive success? Or is it something more complicated? And understand the mechanisms or why we should expect preferences to be like this or not like this?

(15:48) So, roadmap for the talk. I'm first going to spend a couple of slides describing the general framework that we use in this literature and then I want to talk about two insights and the implications of these. If I have time, also a third insight and then conclusion.

(16:09) So the framework is based on simple evolutionary logic. We take the analysis of the level of the population. There is a process of mutation and selection going on this population. So think of there being a sequence of generations; in each generation there was a certain distribution of preferences; there may also be sometimes novel or mutant preferences that enter into the population. Then individuals are somehow matched to interact with each other. Each individual's preferences guides his or her behavior according to standard economic theory. The equilibrium or the behaviors result in equilibrium material payoffs and these material payoffs in turn will determine the reproductive success and those with a high reproductive success will have offspring and so, the differential reproductive success then determines the distribution of preference of the next generation and so forth. So, think of this as a wheel, you turn it on off, you turn it on neutral sometimes and then you see if there is some preference that actually withstands the invasion of these mutants and that is then we will call evolutionarily stable. So I would like to just to note here that these mathematical models that we use are silent as to whether the transmission of traits from one generation to the next is biological or cultural. So there is no assumption there. There is no assumption there that these traits need to be biologically coded. It could be, but not necessarily.

(18:23) Continuing on the general framework. The first goal with these models will be to determine which preferences this process of mutation and selection leads to. And here I'm citing some seminal papers in this literature. The second goal will be to understand how features in the environment in which the population evolves affects the evolutionary viability of preferences. And so, there are many more modeling choices to be made. For example, how are individuals matched to interact? Under what informational... What is the informational context? Can they observe each other's preferences or not? And what is the set of potential preferences? So you will see a couple of variations on these assumptions.

(19:19) So Insight #1 – Evolution by natural selection may favor weaker intra-family altruism in harsher environments. So this is based on a work with Jörgen Weibull, who has been my co-author on essentially, on this work. And so here we're going to look at interactions within the families, between members of the same family and because members of the same family they have the opportunity to observe each other over long periods of time. It is then reasonable to think that those interactions take place under complete information. They know each other very well. They can foresee what the others will do. And so here in this work we also take as a given the fact that individuals are equipped with preferences whereby they attach some positive weight to the other family members' reproductive success. So w here is the notation for reproductive success. And so here, utility of an individual with degree of altruism 𝛼 attaches a weight 1 to own reproductive success, given the behavior strategy 𝑥 played by self and given that the other family member plays strategy 𝑦 and then this weight 𝛼 to the other family members' reproductive success who plays 𝑦 when self plays 𝑥. So this is here, in this paper, the set of potential preferences is going to be this integral (-𝟷, 𝟷) when I look for evolutionarily stable values of 𝛼. Citing this paper here, which is also sort of a canonical paper more generally on preference evolution when interactions take place under complete information.

(21:38) In the 2010 paper that I started, we have the following interaction that can be analyzed. So, think of a pair of siblings simultaneously choose productive efforts. Think of an agricultural society, so in the spring they sow and so forth and they do what they need to do on the farm then they choose a level of productive efforts. Then summer arrives and each siblings stays... Think of them as grown-up siblings that each have their own farm and an output is realized. This is random, so there will be weather shocks, but also the probability that a high output occurs depends also on the level of effort that the individual exerted. And then they actually observe each other's outputs and they may choose to make a transfer to the other or not. And then this amount of food they have and so forth, then gives rise to fertility differential and reproductive success. And these decisions will then be taken, given the altruistic preferences that they have, and in particular, the altruism will lead to lower or higher transfers – the higher is the altruism, the higher is the transfer from a rich sibling to a poor sibling. And this also comes into play when they choose the productive efforts. They can foresee how much they will help each other out. The more altruistic an individual is, the more he or she will be happy to help out the sibling and also to feel sorry for the sibling when the sibling has to help him or her out. So all this comes into play and we can figure out what the equilibrium efforts and transfers are for given levels of altruism. And we can then use this to determine what are evolutionarily stable degrees of altruism.

(23:53) I can go to the evolutionarily stable degrees of altruism and I'm going to show you how these depend on the environment in which this population evolves. And I'm going to have two parameters that measure the environment. One is 𝜆, 𝜆 is here the ratio of the low to the high output. So you can think of this as measuring output variability. Think of 𝘠^H as being nice, red, juicy tomatoes; you can grow such tomatoes in various environments. However, in some environments it's harder to get them and you fail a lot. Whereas in other environments, the outcome of those tomato plants is less sensitive. The second parameter is 𝜃. So 𝑥 here stands for effort and 𝜃 is the marginal return to effort. So you can think of this 𝜃 as the highest 𝜃 with fewer calories you need to put in to produce more calories. So ( 𝜆, 𝜃 ), this pair of parameters, is the environment. And I would say that an environment ( 𝜆', 𝜃' ) is harsher than some other environment. If either the output variability is more pronounced so that if you fail, the consequences are dire. And also the marginal return to effort is smaller, so you need to work harder to produce the same output.

(25:35) Okay. So here we see a picture that shows as a function of the environment. So here we have the 𝜆. So lower 𝜆 is a harsher environment and 𝜃, the marginal return to effort. So again 𝜃, lower 𝜃 is a harsher environment. And so, the vertical axis we see here, it's the value of the evolutionary stable degree of altruism. So in, let's say, these generous environments over here, we get an 𝛼 which is close to 0.5 and this is what we actually would have expected from evolutionary biology because siblings are related by a factor of ½. So essentially, if you're a carrier of some rare gene in the population, then the probability that your sibling also carries those rare genes is ½. So evolutionary biology tells us by Hamilton's rule, that you may have heard of, that you should expect 𝛼 to be about ½. However, now that we have this effect of altruism on production levels as well, we actually get this other effect, which takes the degree of altruism down way below 0.5 in certain environments. So here in this very harsh environment, it's only about 0.2. So there's a stark difference there. And so, what this says is that we may expect intra-family altruism to vary with the environment and in this model, the intra-family altruism is lower in harsher environments. Have you heard of Sweden Gate? No. So recently this whole thing of social media that Swedes are not generous towards strangers. They don't invite them to eat. However, I say that you shouldn't worry if you're a stranger because in Sweden, you would not necessarily even invite your brother to the table. Don't worry, they don't treat you differently.

(28:09) So in sum, what do we get from this analysis? For pre-industrial times in agricultural societies, our model predicts weaker altruism in harsher climates. So one could say that this is actually in line with the notion that individualism rose in northwestern Europe prior to elsewhere. And this is also what Max Weber wanted to... In line also with a statement by Max Weber that you can find in the "Religion of China", he says... So you all know about Max Weber's Protestant work ethic, however, here's another statement,

"The great achievement of Protestantism was to shatter the fetters of the sib. These religions established a common ethical way of life in opposition to the community of blood, even to a large extent, in opposition to the family."

So I'd say this prediction is kind of in line with the idea that in parts of the world where Protestantism was readily adopted, that is northwestern Europe, we have this weaker family.

(29:26) So more generally, what this model says is that it predicts that the strength of family ties depends on the environment. And I suppose that maybe this is a bit provocative, but I'd like to raise a question that, well what does this say about the relevance of the economics model of individual utility maximization? How relevant is that for parts of the world where we may expect family ties, intra-family altruism to be extremely strong? I think it also raises the questions for economic development. It's a chicken and egg question. Sometimes we hear that in developing countries, informal risk sharing is there because there are no formal markets, no formal insurance markets. However, you could also think that it could have been that the lack of formal insurance markets may be due to extensive intra-family resource sharing that actually lowers the demand for formal insurance.

(30:45) On now to Insight #2. So, evolution by natural selection favors a concern for universalization or a form of Kantian morality. So here, we turn to interactions between strangers. And so for interactions beyond the family, it's questionable to assume that we observe each other's preferences. So here we're going to adopt a model in which interactions take place under incomplete information. So each individual knows his or her preferences, but not those with whom they interact. And you remember this question, we need to always figure out what is the set of potential preferences. So in this work, we are going to take a minimalistic approach by saying that, well, a utility function could be any function where the domain of the function is the set of strategy profiles, that is, preferences only need to describe the preferences of this over the strategies played by self and the other. To check for the reasons you must pay strict attention to continuous utility functions, but the approaches, it could be anything. And so this is based on a work again with Jörgen and here, I'm mentioning these papers that also have general models of this kind of interaction. However, the way it differs is that we allow for another kind of matching process between interacting individuals, and this is what leads to this Kantian model that you do not see in their model.

(32:52) So before I state the result let me define what Homo moralis is. So, an individual is a Homo moralis with a degree of morality 𝜅, which is a number between [𝟶, 𝟷] if their utility function is of this form. So here, you recognize again, the first term is the individual's own reproductive success when playing strategy 𝑥 when the other plays 𝑦. Now here, this is different from altruism. Here we would no longer have the actual reproductive success of the other. Instead, it is the reproductive success the individual herself would obtain if her strategy 𝑥 was universalized so that the other also could play this, hypothetically.

(33:47) And I'm going to show results that come strongly towards this utility function as being favored by evolution by natural selection. So why do we call this morality a Kantian moral concern? So according to Kant, he said,

"Act only according to that maxim whereby you can will that it should become a universal law."

Now, if we look at this function here, you can think of it as saying, "Act according to that maxim whereby you can will that others should do likewise with probability 𝜅." So it's a partial Kantian concern when 𝜅 is below 𝟷, but you get the full Kantian concern when 𝜅 isn't.

(35:01) And so here is the result. The first part says that this kind of preferences of Homo moralis with a specific degree of morality 𝜅 = 𝑟, which I will tell you more about in a second, is evolutionarily stable against any preference type that is behaviorally distinguishable from such Homo moralis, and we have almost dimensionality of statement. The second part is, any type which is behaviorally distinguishable from Homo moralis with this particular value of 𝜅 is evolutionarily unstable, that is, will be displaced by some mutant utility function that comes into play.

(35:57) What is 𝑟 ? So 𝑟 is the coefficient of relatedness, which is a term that goes back to biologists Sewall Wright in 1931. The coefficient of relatedness is essentially the probability that interactants have a common ancestor not too far back in time. And so this can be calculated, but here we just take sort of a reduced approach and look at this as if mutants are extremely rare.

(36:34) So some intuition for why this result obtains – the Homo moralis preempt entry by mutants. How is that so? Because indeed, this baseline intuition that if evolution is all about reproductive success, we should be expected to simply maximize reproductive success. We should expect to see only this term here, but that's not the case. So how can this utility function lead to behaviors that preempt entry by mutants? The intuition is that Homo moralis with this specific degree of morality 𝜅 =  𝑟 will play a strategy that solves this fixed point product. So they're saying Nash equilibrium is the best response to itself according to these preferences. Now take a rare mutant coming to this population with some other preferences that are behaviorally distinguishable from Homo moralis, such a mutant would play some strategy, let's call it 𝑧, and remember that this 𝑟 here is the probability that a rare mutant is going to interact with another rare mutant. So here, we have the term that corresponds to when this happens. So a rare mutant playing 𝑧 with probability 𝑟 is  matched with another mutant also playing 𝑧. However, with probability (𝟷 - 𝑟) your rare mutant is matched with the Homo moralis, who is playing this strategy here. And so we see here that the rare mutant gets on average this reproductive success. But the Homo moralis is already playing that strategy that maximizes he reproductive success of a rare mutant, which leads us to this very strong result.

(39:00) So where does this positive coefficient of relatedness 𝑟 come from? It comes from... So as I said, it's the tendency for individuals sharing a common ancestor to interact and this happens in populations where, let's say, were structured into groups or at least the result of limited migration between groups. So if you're born in a place and then among your neighbors you will find people who were also born in the same place or whose ancestor or... Yes, sorry. And so this corresponds exactly to our evolutionary past, so there is evidence that for, let's say, the last couple of million years or so, our ancestors lived in small groups of between 5-150 grownups. They extended beyond the nuclear family and that clearly there was a limited migration between these groups. There was some migration but not fully. The population was not fully re-mixing in each time period. So this kind of population structure is part of the environment of evolutionary adaptiveness of the human lineage. So, I think it's safe to say that this 𝑟 ≠ 0 for most of our past. So we should expect, according to this theory, that behavior should have been aligned with this Homo moralis with a certain value of 𝑟.

(40:55) Now, what are some implications of these preferences? So I'm going to show you a couple of simple interactions just to highlight some key differences between altruism on the one hand and Homo moralis preferences on the other. So here I've got prisoner's dilemma and we start by Homo oeconomicus. We know that (𝐷, 𝐷) is the unique Nash equilibrium, if we take now altruists, then an altruist can take into account the negative impact or positive impact on the other and so, for a degree of altruism high enough, then the unique equilibrium will be (𝐶, 𝐶). The same is true for individuals who are Homo moralis. If their degree of morality is high enough, then the unique equilibrium is also (𝐶, 𝐶). Here, there's no distinction. I mean, both altruism and morality give rise to pro-social behavior.

(42:00) Now let's look at coordination game. Coordination games are everywhere in life. And so here the 𝐺 and 𝐵 here stands for the good and the bad equilibrium. So with Homo oeconomicus, both (𝐺, 𝐺) and (𝐵, 𝐵) are Nash equilibrium. Let's look at altruists. The same is true for altruists, both (𝐺, 𝐺) and (𝐵, 𝐵) are equilibrium. If I'm very altruistic and I expect my opponent to play 𝐵, then I will also play 𝐵. I would even have a stronger incentive to play 𝐵 than Homo oeconomicus because I would care about the negative externality that I would exert on my opponent if I were to switch to 𝐺. Here is where the difference is with Homo moralis because Homo moralis is kind of selfish, does not care about the negatives and... instead evaluates the strategies according to what material payoff would be if behavior strategy was universalized, so for 𝜅 high enough the bad equilibrium is no longer in equilibrium, so (𝐺, 𝐺) is the unique Nash equilibrium for 𝜅 high enough. So here, Homo moralis acts like an equilibrium selection mechanism.

(43:27) If we think about public good settings, where each individual's real impact is negligible, I just think of climate change and pollution. Think voting, then Homo oeconomicus will not contribute or vote, if it's costly. Altruists neither. Altruists, even if they care a lot about others, if their impact descent is essentially nil, they will do nothing either. Homo moralis would be very different because Homo moralis evaluates what the outcome would be if the behavior was universalized, so Homo moralis acts as if he or she had weight in the population, in the outcome, although this is not the case. He or she will be perfectly conscious about the fact that there is no real impact but would nevertheless act as if it was there. So this preference class is novel to economics. So recently there have been quite a few papers on this, however, I think we still need to understand better what these Homo moralis preferences actually imply for standard models.

(44:49) So I think I do have time. I'm going to Insight #3, which will actually... So Homo moralis result is good. That gives us hope for human nature, that we can... Maybe we have this latent 𝜅, maybe latently high 𝜅 somewhere if we can awaken it. However, this third insight will set us back a little bit. Why? Because we haven't been able to work out an even more detailed model, which leads us to the following prediction that evolution by natural selection does indeed favor the Kantian moral concern at the level of reproductive success, but at the level of material payoffs, which is what social scientists mostly observe, then we should be expecting then to see a mix of Kantian moral concern and some other-regarding concern, which could be altruism or spite.

(46:10) So, what this work does is it... So far I have only talked about reproductive success, I've not really talked about material payoffs, trivial material payoffs that we see in everyday life. So in this work together with Jörgen Weibull again, but also with Laurent Lehmann, who is an evolutionary biologist in Lausanne, we have worked on disentangling these two, so we still have the reproductive success but now we also have explicitly the material payoff, trivial material payoffs of everyday life.

(46:47) And so here is... So it's more complex model. Don't worry about the math here. What this theorem says essentially, is that we confirm what we found before, namely that at the level of reflective success it is still Homo moralis that rules. I actually have this term again, relatedness times reproductive success if own behavior was universalized. So do not kill your neighbor, that's perfectly easy to explain with this kind of utility function.

(47:27) However, if we now look at trivial material payoffs, so what biologists call weak selection, where material payoffs affect reproductive success only marginally, then we have the following result, which says that on the weak election, the following... this utility function is uninvadable or evolutionarily unstable and here 𝜋 stands for these material payoffs. Now we have a more complex function. It's not so important that we have these four terms but what is important is to realize that there is now a novel term compared to what we have when we look at preferences at the level of reproductive success. Here we now see the 𝜋(𝘹ⱼ, 𝘹ᵢ) appear, which is what? That is the material payoff of the other individual. So, I still have... So I don't care about all material payoffs, I care about what my material payoff would be if my behavior was universalized but I would now also care about the opponent's actual material payoff, and depending on the sign of this variable 𝜆, either I would exhibit spite or I would be willing to incur cost to lower the other's material payoff. Or 𝜆 could be negative, in which case, I would be also on top of being Kantian moral, I would also be altruistic. And this comes from local competition or rather the gains in reproductive success that can be obtained by lowering or increasing the material payoffs of my neighbors.

(49:42) So, coming to the concluding remarks here. So theory can help us understand how evolutionary forces may have shaped our preferences and we can also... This theory can help us understand how the environment has affected our preferences, and in particular, it can help us understand the variations of preferences across the world, potentially, and it can also lead us to discover novel preference classes. To come back to the introduction, we'll be thinking again about, so what are the implications for policy recommendations? Well, hopefully they can be a complement to behavioral economics and to insights from other disciplines as well about human nature and to the empirical work that we have from there. It can be a valuable complement because it allows us to enlarge the set of potential motivations and it can also help us explain the testable predictions on how we should expect potential cultures or preferences that vary across the different parts of the world. But of course, to formulate desirable and effective policy recommendations, we need also to do work now to assess the theoretical implications of these novel preferences and also to assess the empirical relevance to try to get a sense of what the distribution of preferences are. So there is some initial experimental work on this, but we're far from understanding the full implications of these results. Just to mention a few recent surveys if you're interested in this work, I would also like to mention John Newton's recent survey on evolutionary game theory, as well as very interesting work by Arthur Robson on preference evolution in decision problems that do not involve strategic interactions.

(52:08) So thanks to my co-authors on these papers without whom this certainly never would have occurred and I should mention that in the one month with the biologists, we really got to interact with them, thanks to the the IES team, which is another home for interdisciplinary dialogue. And yes, thanks to of course, funders and inspiration from the 19ᵗʰ century. Thank you.