Sunday, December 28, 2014

Morality and Selection - How?

Group, individual, or gene level selection: Which has the best explanatory power for our ‘moral’ emotions?

Mark Sloan 12-27-2014

Charles Darwin proposed group selection to explain how “altruistic” behaviors toward others and self-sacrifice for the group, which could easily be taken to be counterexamples to his theory of evolution, might instead be examples of its explanatory power. He observed that, all else being equal, groups that were more cooperative would out compete less cooperative groups. Selection for more cooperative groups selects for biology that better motivates cooperative behavior (Darwin, 1871) and, though Darwin did not mention it, cultural norms that do the same. Thus, we arrive at the simple and appealing notion that people’s experience of compassion and loyalty and cultural norms such as “Do to others as you would have them do to you” (or variations with the same function in increasing cooperation) are arguably near inevitable products of blind evolution of beings with sufficiently complex brains. And no supernatural deity is required to explain human, or any other species’, “goodness”.

Mark Borrello (2005) described group theory’s subsequent rise and then fall from respectability in the 1960’s (mainly due to a theoretical objection and some bad science from a chief advocate) and its rise again, starting in the 1980’s and now called multilevel selection, which is how I will refer to it here where appropriate.

Steven Pinker and twenty three other respected evolutionists provided an accessible snapshot (Pinker 2012) of the lively state of the unresolved debate on group selection’s proper role in evolutionary science. The diversity of pro and con opinions expressed makes it appear the debate about group selection is not close to over.

However, Pinker’s focus on “altruistic” and “self-sacrifice for the group” behaviors as “what must be explained” was a poor choice for resolving the relative explanatory power of group versus individual and gene level selection perspectives. First, as they are commonly understood, these behaviors can have many explanations which make them insensitive tests for differentiating between hypotheses. Also, these explanations included cultural norms and forces, which divert attention to a secondary issue, group selection’s role in cultural evolution. Finally, Pinker appeared to be taking “altruism” to mean something very close to “biological altruism” (behavior by an individual that increases the lifetime fitness of others while decreasing the lifetime fitness of the actor). If what-must-be-explained is biological altruism, his argument is almost won before he starts. Without the requirement to explain cooperation behavior, the most credible candidate is kin altruism, the obvious choice from the gene level selection perspective.

To avoid these and other problems, future group level selection discussions might be made more fruitful by choosing human cooperation as what-must-be-explained. Also, making certain stipulations will clarify the discussions and avoid other above described problems. I propose the following:

1) Multilevel selection’s explanatory power for biology will be evaluated separately from culture, requiring what-must-be-explained to be chosen to insure only biology is needed to explain it.

2) Multilevel, individual, and gene level selection are different perspectives on (or bookkeeping systems for) the same phenomena. In theory, they might all have the same explanatory power.

3) To be a useful part of an evolutionist’s toolkit, multilevel selection’s perspective must show better explanatory power in practice, meaning ability to more readily produce correct explanations of specific biology.

4) A hypothesis’ “in practice” explanatory power may be inferior to its “in theory” explanatory power.

5) Multilevel selection’s “in practice” explanatory power will ideally be evaluated for multiple phenomena, not just one. Requiring explanation of multiple phenomena increases discriminatory power between alternate hypotheses.

I expect these stipulations would be acceptable to many in the field. If they are not acceptable, perhaps discussions could clarify what will be stipulated and any dissenting opinions.

The biology underlying what makes human beings the amazingly cooperative social animals we are, and arguably the heart of what makes us ‘human’, provides two candidates for what-must-be-explained.

The first candidate is our moral emotions triggered by our moral sense that, in part, motivate unselfishness. For reasons described below, I will here take these to be compassion, loyalty, gratitude, anger, disgust, contempt, shame, guilt, and ‘elevation’. (‘Elevation’ is an emotion encompassing feelings of satisfaction and pride.) Note that the existence of the emotions triggered by our moral sense is not culturally dependent. That is, the existence of these emotions does not vary from culture to culture. Only the circumstances when they are triggered and their intensity with which they are triggered are culturally and individually dependent.

The second “cooperation biology” candidate is the cross-cultural universal set (and therefore likely biological, but not certainly biological) of categories of circumstances which can trigger our moral sense to make one of its near instantaneous judgments (and in turn trigger our ‘moral’ emotions). The most commonly referred to version of Moral Foundations Theory identifies five cross-culturally universal categories: care/harm, fairness/cheating, loyalty/betrayal, authority/subversion, sanctity/degradation (Graham 2012).

Another candidate for what-must-be-explained is the evolution of cancers with their intrinsic complexities of simultaneous cooperation and competition at the cancer’s gene, cell, and cell strain levels in the hostile environment of the patient’s immune system. Due to the specialized knowledge required, such group selection debates would likely be most effective among experts.

However, it may be that such a debate is not needed. Researchers in the field use group selection perspectives as standard practice (Merlo, 2006). It seems unlikely that any debate among those experts would conclude using the group selection perspective is not useful in actual practice.

There are doubtless other candidate biological phenomena that could be useful choices for what-must-be-explained.

The following is an example of how the character of group selection discussions could change if what-must-be-explained was human cooperation and the above stipulations were accepted. In the interest of providing a brief example, I will only compare the “in practice” explanatory power of multilevel and gene selection perspectives for our ‘moral’ emotions, the emotions triggered by our moral sense.

Drawing on many sources, Johnathon Haidt argued moral emotions include compassion, gratitude, anger, disgust, contempt, shame, embarrassment, guilt, and ‘elevation’ (Haidt 2003). I largely agree, but add ‘loyalty’, meaning the emotion that motivates “loyal” behavior, since unselfishness motivated by loyalty is a central part of morality. I prefer the word indignation to anger because “indignation” better captures an emotion that is intrinsically about right and wrong. Also, I would omit the emotion “embarrassment” since embarrassment can be about matters having nothing to do with right and wrong, and shame and guilt adequately convey this category of moral emotions. For these reasons, I will use explanatory power for compassion, loyalty, gratitude, anger, disgust, contempt, shame, guilt, and ‘elevation’ as what-must-be-explained.

If our moral emotions are components of cooperation strategies, what necessary functions are they fulfilling?

Tit-for-tat2 is the simplest highly effective cooperation strategy known in game theory. It seems almost inevitable that, if evolution did encode cooperation strategies in our moral sense, something like tit-for-tat would be one of those strategies. Tit-for-tat has two necessary behavioral components: 1) cooperation by unselfishly risking exploitation in any first interaction and also if the other individual cooperated in the previous interaction and 2) retaliation if the other individual attempted to exploit unselfishness in the previous interaction.

However, this simple tit-for-tat strategy is not always the best, and what is better varies with circumstances. Perhaps sometimes there are ways to predict who is likely to be a good cooperator and who is likely to exploit you – for instance from gossip about other’s interactions with the individual. Perhaps ‘forgiveness’ of an instance of exploitation could avoid cooperation destroying endless cycles of retaliation.

How might evolution most efficiently encode motivation for people to develop more effective variations of tit-for-tat and even switch between them in the many different circumstances encountered over a lifetime, and sometimes even over the course of a single day – as we know people readily do? If a mutation produced a psychological reward for achieving cooperation, then that mutation could be selected for because it motivates the individual to find those strategies that are more effective variations of tit-for-tat and to switch between them according to circumstances.

So if evolution did encode cooperation strategies in our moral sense, the simplest implementation that would be effective in diverse circumstances would have three necessary components: 1) motivation for unselfish behaviors, such as helping actions, which can initiate cooperation but risk exploitation, 2) motivation for retaliation against exploiters, and 3) psychological rewards for achieving cooperation.

This is exactly what we find in our moral emotions. 1) Compassion, loyalty, and gratitude motivate unselfish behaviors that initiate or maintain cooperation. 2) Indignation, disgust, and contempt motivate behaviors that punish other’s exploitation of unselfishness; shame and guilt internally punish our own selfishness. 3) The emotional experience of ‘elevation’ rewards our own successes at cooperation, motivates imitating other’s successes, and encourages staying in cooperative groups and seeking to increase cooperation within those groups.

In contrast, consider the gene level selection perspective’s most common explanation of unselfish behavior toward non-kin (and moral behavior in general): that unselfishness toward non-kin is best explained as a product of kin altruism and cultural influences (Pinker 2012). How well does that explanation explain the existence of our moral emotions?

Kin altruism readily explains the existence of compassion and loyalty toward close kin and the reward of ‘elevation’ for cooperation with close kin. While speculative, the argument that these emotions are triggered toward non-kin due to a kind of error in kin detection is possible.

But what about the other six: gratitude, Indignation, disgust, contempt, shame and guilt? How are these explained by kin altruism? They appear to have nothing to do with kin altruism. I expect clever people can find arguments that all these emotions can be explained by kin altruism. But inventive speculations about how they might be explained as products of kin altruism fall far short on the “best explanation scale” compared to multilevel level perspective’s robust explanation.

Based on relative explanatory power, we can conclude that the majority of our moral emotions are elements of reciprocity cooperation strategies, not kin altruism.

Does this failure to explain our moral emotions mean that gene level selection cannot, in theory, explain all gene determined biology? Again, no, it simply means that the gene level perspective can, in practice, incline investigators to produce false explanations of biological adaptations.

Multilevel selection robustly explains the set of emotions that are motivated by our moral sense. Gene level selection’s common claim about unselfishness (and moral behavior in general) toward non-kin being kin altruism misapplied cannot directly explain two thirds of the emotions motivated by our moral sense. Based on its poor explanatory power, this common claim, which flows naturally from the gene level selection perspective, appears false. Gene level selection’s perspective appears, in practice, to have led Pinker and other gene-level-only advocates astray.

In practice, it appears that the multilevel selection perspective produces correct explanations of our moral emotions more reliably than gene level selection. Therefore, the multilevel selection perspective deserves an important place in the evolutionist’s tool kit.

1. Unselfishness (including “altruism” and “self-sacrifice for the group”) is here defined as accepting a cost in order to intentionally benefit others without consideration of possible future benefits.
2. Tit-for-tat is a highly effective, but extremely simple, cooperation strategy in game theory (Axelrod, 1984). An agent using this strategy will first cooperate, then subsequently replicate an opponent's previous action. If the opponent previously was cooperative, the agent is cooperative. If the opponent previously tried to exploit the agent, the agent will retaliate by attempting to exploit the opponent.

Axelrod, David, (1984). The Evolution of Cooperation.
Borrello, Mark E., (2005). The rise, fall and resurrection of group selection.
Darwin, Charles, (1871). The Descent of Man.
Grahama, Jesse, Haidt, J., et al. (2012). Moral Foundations Theory: The Pragmatic Validity of Moral Pluralism. Available at
Haidt, J. (2003). The moral emotions. In R. J. Davidson, K. R. Scherer, & H. H. Goldsmith (Eds.), Handbook of affective sciences. Oxford: Oxford University Press. (pp. 852-870).
Merlo, Lauren M.F., Pepper, John W., et al. (2006). Cancer as an evolutionary and ecological process.
Pinker, Steven, et al. (2012).The False Allure of Group Selection - An Edge Original Essay with Replies.

Mark Sloan’s personal interest is in showing how understanding morality as an evolutionary adaptation can be culturally useful. As associate editor of This View of Life, Morality section, his goal is to promote a cross-disciplinary view of the science of morality and how that science could be culturally useful. He has degrees in engineering and physics and has had a career in the aerospace industry.

Copyright©2014 by Mark Sloan – All Rights Reserved – No Reproduction or Publication for Commercial Use Without Express Written Permission


Hostirad said...

This essay does an excellent, comprehensive job delineating the evolved moral emotions that have motivated, and continue to motivate, members of our species toward cooperative behavior.

However, that is not the essay's main purpose. The essay's purpose is to argue that group selection (or multi-level selection) provides a better explanation for the evolution of moral emotions than gene-level or individual-level selection. To Mark Sloan's credit, he provides a link to what I've found to be the single best argument against group selection, Steven Pinker's Edge essay published in 2012. I was convinced by Pinker's arguments when the essay first appeared. After carefully reading Mark's objections to the Pinker essay, I still find myself agreeing with Pinker's points.

One task that Pinker took very seriously was to define what might be mean by "group selection." He notes that if we regard group selection as strictly analogous to the selection of genes, it would mean that (1) a group of persons would be the basic unit of selection; (2) groups would have to reproduce, creating copies of themselves; (3) natural selection would act upon the copies produced by different groups such that copies produced by certain groups survive and reproduce more successfully. But I think even those who are inclined to argue for the existence of group selection would agree that groups do not replicate, creating copies of themselves the way genes do, with the replicants showing differential survival and reproduction due to natural selection.

No, it seems that people arguing for "group selection" mean something else that is not strictly analogous to selection at the level of genes. But what do they mean? Pinker goes through some possibilities, and it is not clear which of these definitions of group selection Mark endorses. Not that it matters. The strength of Pinker's argument is not whether kin selection or reciprocal altruism underlies the evolution of moral emotions; it's strength is in pointing out that what people probably mean by "group selection" is actually just ordinary selection occurring in the context of groups. Pinker points out that the behaviors motivated by the moral emotions are good strategies for individual survival and reproduction within group living.

Clearly there is still work to be done in elucidating how the various behaviors motivated by the entire range of moral emotions have helped individuals to survive and reproduce. What I have yet to see the need to invoke some sort of "group selection" that does not reduce to individual selection. (E.g., is there such a thing as a "cooperative group" that means more than a group of cooperative individuals?) For now, it seems to me that moral emotions evolved just like all of the non-moral emotions: because they motivated our ancestors to engage in adaptive behaviors.

Robert Kadar said...

Here are two arguments that I think are important for our discussion especially in relation to Pinker's argument:

1) I believe people are often confused about multilevel selection because of the concepts "gene-selection", "individual selection" "group selection" and there definitions. The simple matter is that a trait is either good for the individual (at a cost to the group, most of the time) or good for the group (at a cost to other groups, most of the time). People should start looking at evolution from a "trait's eye perspective" to determine at which level the trait evolves. So we should start with the trait and then ask if it increases in frequency by individual selection (it increases relative fitness over other individuals) or group selection (it increases the relative fitness of the group over other groups).

2) We should get rid of "gene selection" because it confuses the ideas of a what a trait selects on --ultimately the genes-- with what level of selection a trait evolves (individual or group). This conflation along with the fact that gene-selection and individual selection are the same thing, doesn't merit any reason to use "gene selection".

Then kin-altruism, tit-for-tat, reciprocal altruism all are selectively disadvantageous when competing against a selfish trait. The only way they can evolve is by competition between groups. So to repeat, cooperation in the reciprocal sense or in the purely altruistic sense are both disadvantageous when faced against selfishness within groups. The only way for these traits to evolve is by selection between groups. Pinker doesn't get this because he still believes in things that were miscommunicated decades ago by Dawkins and other evolutionists. Kin-selection is a type of group selection that relies on high genetic related individuals forming groups to compete against other groups.

With this understanding, we can then compare the traits of universal human emotions like the ones Mark pointed out and ask why and how these traits were selected. I hope this helps.

Anonymous said...

Robert, thanks for commenting. Yes, what you have added to the conversation does help.

I like “People should start looking at evolution from a "trait's eye perspective" to determine at which level the trait evolves.”

And yes, after considering it for a while, we should stop using the phrase gene level selection because it leads to errors as Pinker has illustrated. As I was writing this essay, I was thinking “that is not really right”.

How about referring to gene level encoding, individual level selection, and group level selection?

However, I want to question: “Then kin-altruism, tit-for-tat, reciprocal altruism all are selectively disadvantageous when competing against a selfish trait. The only way they can evolve is by competition between groups.”

Kin altruism, such as a mother toward her offspring, works just fine when offspring are selfish regarding the mother. I don’t see what selfish trait they are losing out to.

Now let’s do a thought experiment to examine how tit-for-tat could evolve.

Assume some animal has already evolved something functionally equivalent to the emotion indignation (motivation for retaliation when another animal harms it in some way) and can recognize and remember whoever caused that harm, but has not yet evolved any biological motivation to act unselfishly. Suppose a genetic variation occurs that motivates unselfishness (and thus initiates cooperation). Since the animal already has the emotion indignation and can recognize and remember whoever caused that harm, the mutation that motivated unselfishness completes the encoding for something like a tit-for-tat cooperation strategy in the animal’s biology.

However, consider if only a single animal is programmed with the tit-for-tat strategy amongst always selfish animals. If it acts unselfishly toward every animal on their first meeting (the first step in tit-for-tat) it will lose fitness at each first interaction and yes, it loses to selfishness within the group.

But if it meets another animal in the group who has also been programmed with tit-for-tat (perhaps a sibling) they can form a cooperative sub-group with increased relative fitness and the ability to defeat free-riders (free-riders meaning always selfish individuals in the larger group). Any attempt at free-riding will immediately trigger indignation and the free-riders will get nothing after the first interaction. For real animals, rather than tit-for-tat idealizations, the free-riders may be lucky to escape without injury.

A claim that cooperators will always lose to free-riders within a group is false. So cooperation can evolve within a group as I see it.

Of course we could take the perspective that the two have formed a cooperative subgroup which is then in competition with the always selfish rest of the group, and that becomes group competition. Is that what is meant?

Anonymous said...

Hostirad, thanks for commenting. I will try to clarify a few points.

I am not aware of any advocates of "group selection" claiming it is directly analogous to encoding for behaviors at the level of genes. What “level” refers to in group level selection is the level at which the selection force acts not the level at which the biology is encoded. (“The level at which the selection force acts” refers to the level at which the phenotype is selected for. For groups that often means the level at which cooperative behavior is selected for.) It almost sounds like Pinker thinks group level selection refers to the level at which the biology is encoded which makes no sense, but does make group level selection easy to criticize.

Since essentially no one (perhaps no one) advocates for this, Pinker’s focus on this as a “group selection” position has the appearance of a straw man argument – though I assume Pinker was quite sincere and would not consciously produce a straw man argument.

What I said about this as stipulation 2) was that

2) Multilevel, individual, and gene level selection are different perspectives on (or bookkeeping systems for) the same phenomena. In theory, they might all have the same explanatory power.

That said, I agree with Robert Kadar’s post below in which he says: “People should start looking at evolution from a "trait's eye perspective" to determine at which level the trait evolves. So we should start with the trait and then ask if it increases in frequency by individual selection (it increases relative fitness over other individuals) or group selection (it increases the relative fitness of the group over other groups).”

As Darwin and evolutionists since him have understood, evolution selects for traits, not genes. Genes are merely the substrate that encodes these traits (makes them heritable) for future generations.

After some thought, I also concur with Robert that “gene level selection” is a term that we should get rid of because of the confusion it produces. I had to talk about it because other’s were, but all that really happens at the gene level is encoding, not selection. If you want to understand how a behavior was encoded, look at the gene level. If you want to see how a trait was selected for, look at the individual or group level.

But the big nail in the coffin is Pinker’s claim is simply wrong that unselfishness is best explained by kin selection (which deceptively easily flows from a gene level perspective) and culture. Kin selection can explain neither most of our moral emotions motivated by our moral sense nor half the cross cultural universal categories of circumstances when our moral sense is triggered (which I did not get into due to length and because I thought it was overkill). Also, note that culture has nothing to do with whether a specific emotional experience exists, and Pinker cannot claim that culture is responsible for those unexplained emotions and circumstances.

The “need to invoke some sort of ‘group selection’” is that, in practice, multilevel selection (the modern term) more readily leads to correct explanations of biology, rather than incorrect explanations such as Pinker’s. Our moral biology (centered in our moral sense) is beautifully explained as the product of selection for cooperation (a natural conclusion from group selection) and is not explained by kin altruism, which ‘sounds’ right from the gene level perspective.

Pinker’s favored gene level perspective too readily leads investigators to wrong answers. That is what is wrong with it. Multilevel selection more often leads investigators to correct answers. That is why it is an important tool in the evolutionist’s toolkit.

Hostirad said...

My response to Mark's comments on my first response has been entered on the Evolutionist Facebook page.

Anonymous said...

The things you have talked about in this post which should be exceptionally useful for us. In view of these sublime data in this post the website can be seen over and over.
Journal Selection

E♦B said...

Christopher X J. Jensen replies to this article on his blog, here:

E♦B said...

Mark Sloan revised his essay that originally appeared in ASEBL and posted it here, for those interested: