Ethical axioms are found and tested not very differently from the axioms of science. Truth is what stands the test of experience. — Albert Einstein
The traditional view is that science has nothing to say about ethics and morality. Science tells us what is and morality tells us what ought to be. You can’t get ought from is. Science can help us predict the consequences of our actions, but it cannot tell us which consequences we should seek and which we should avoid. How, then, do we decide what is ethical? We must look in our hearts and souls; we must turn to religion or spirituality. We must seek ethical axioms to which all can agree (except psychopaths, sociopaths, deviants, and people who are not like us) and make sure that our actions conform to those axioms. I disagree with this view.
Why do we have ethics and morals? What is their function? Let’s define an agent as a being who has beliefs and desires, and who chooses actions based on those beliefs and desires. Different agents often have incompatible desires, which leads to conflict. The function of ethics and morality is to resolve conflict among agents; to facilitate cooperation among agents. A solitary agent (if there could be such a being — even hermits and feral children are never entirely solitary, since animals can be agents, and since actions can have consequences far beyond their local origins) would have no need or use for morality.
Human agents have many goals, and the weights and priorities they attach to their goals vary with time and circumstances. One way to resolve conflict among agents is to find a shared goal and persuade the agents to agree on an action that is consistent with that shared goal. Many moral systems take this to an extreme, by attempting to formulate a moral axiom (or set of moral axioms) to which we can all agree. Some hope that science might help us to find such a moral axiom. I agree with the critics, who argue that this is not the kind of thing that science can do.
Where do our ethics and morals come from? An ethical system is an algorithm that an agent uses for making decisions in the context of other agents, when there is the potential for conflict or cooperation with the other agents. Our ethical algorithms have biological and cultural components, which have evolved by biological and cultural evolution. Science can help us to understand the evolutionary origins of our ethics.
This view comes close to the naturalistic fallacy, a type of false argument that attempts to base an ethical system on facts of nature. But I do not claim that science can tell us what is ethically right or wrong; instead, I want to make a kind of meta-ethical claim, that science can help us to choose between competing ethical algorithms. The fact that a certain ethical system has survived countless years of biological and cultural evolution does not imply that it is a good ethical system, because evolution makes major mistakes. However, it does imply that the survivor is better than some of its competitors that became extinct (with various careful scientific caveats, about statistical sampling and the specific environment).
Ethical systems evolve in much the same way as scientific theories evolve. Scientists select theories that are fitter than competing theories, in terms of their scope, fertility, and explanatory and predictive power. Ethical agents select ethical systems that are fitter than competing ethical systems, in terms of their ability to facilitate cooperation and reduce conflict with other agents. This is not because cooperation is inherently good and conflict is inherently bad; it is simply because, whatever goals an agent has, some ethical systems will make it easier to achieve those goals than other ethical systems. Science can help us to find effective ethical systems — effective in terms of our own personal long-term goals in life.
The paradigm of the scientific approach to ethics is Axelrod’s Evolution of Cooperation. The lesson of Axelrod’s experiments is that tit for tat is an effective ethical algorithm for certain types of conflict among agents. This is not to say that tit for tat should be elevated to the status of an ethical axiom; it is only to say that tit for tat is better than many competing algorithms (for certain types of environments, with certain statistical sampling assumptions). This is a good example of what we might learn from a mature science of ethics.
Arguments about whether a certain action is ethical often lead to a stalemate, because the agents involved cannot agree on a set of ethical axioms. Science cannot tell us whether a particular axiom is right or wrong, just as we never know whether a particular scientific theory is true or false, only whether it is fitter than competing theories. But we can take the ethical argument to a meta-level and ask which algorithms the participants in the debate should adopt in order to more effectively achieve their goals; this is a problem that science can address.
I believe that a science of ethics (which is currently in its infancy) will eventually justify some of our most cherished ethical beliefs, such as The Golden Rule and the importance of diversity, perhaps with various caveats, embellishments, and qualifications. I envision tit for tat-like experiments that will show these are highly competitive ethical algorithms (not that they are right or wrong).
Where do altruism and self-sacrifice fit in this scheme? It might be said that I am arguing for a kind of enlightened self-interest, which is incompatible with pure altruism. The way to deflate this criticism is to focus on the question, what is the self? I do not necessarily identify myself with my body or my genome; I can identify myself with a certain set of ideas, a certain set of values, or a certain group of agents. Pure altruism can arise from enlightened self-interest within this broader sense of self.
(This blog post is partially based on discussions with my son, Craig Turney, and with Peter Watts.)
Filed under: Evolution, Philosophy of Mind, Philosophy of Science, Political Science | Tagged: conflict, cooperation, ethics, facts, morality, science, values
The relationship between self and your body is interesting. What if I lose 10% of my brain and my arms and my legs? What if I lose 50% of my brain and every muscle in my body? Where is this “self”?
Axelrod’s reciprocal altruism is rendered impotent by migration.
The algorithm:
“Defect, migrate, repeat.”
Is highly adaptive so long as you don’t cycle back to a previously exploited environment before it has recovered some social capital by your absence, and those against whom you defected are long dead.
Yet many believe immigration restriction is unethical.
Axelrod’s reciprocal altruism is rendered impotent by migration.
Axelrod’s model is highly abstract and omits many important details. It is like a simple physical model using Newton’s laws of motion, in which one is asked to ignore friction. We should be very careful about how we apply the model to the real world. In particular, the model omits direct communication among agents; the only communication is the signal given by a decision to cooperate or defect. In a slightly more advanced model, the agents would send messages to each other, and the “defect, migrate, repeat” strategy would fail.
As I said, the science of ethics is in its infancy. Axelrod has pointed us in the right direction, but we have a long way to go.
Well, multiplayer prisoner’s dilemma game theory was around long before Axelrod’s two-player iteration/memory, but I won’t quibble here about priority.
You bring up the very important, particularly for human culture, topic of communication which presents its own dilemmas since signals can be forged. Evolving the capacity for communication involves a similar Dilemma. Michael Oliphant did some important work in a 1D world and found spatially organized kin selection (migration restriction) as well as iteration/memory to be routes via which communication can evolve:
http://cogprints.org/169/
Several years ago I ran a model of the evolution of communication involving 2D spatial organization, iteration/memory and climate variation. Agents could say “cooperate” or “defect” to each other based on their prior experience within their local ecology and would migrate by Gaussian diffusion:
http://laboratoryofthestates.com/climatepd.html
In that world, virtually any migration would destroy any evolution of communication, which then forces reversion to “defect, migrate, repeat”.
This has an analogy in the evolution of virulence. In the older literature, the term was ‘passaging’. Nowadays guys like Paul Ewald use the term “horizontal transmission”.
Another really huge problem with the evolution of communication is that the ability to detect defection may be limited or diverted by sophistries like “Correlation doesn’t imply causation.” Ecologies are very complex systems and validating causal structure may, itself, require extremely prejudiced restrictions on migration just so you can have experimental control groups to in/validate your causal hypotheses.
I have Google set to send me email when it finds references to the Golden Rule, which is what led me to your article.
I am fascinated by the work that’s being done on the biological and cultural drivers for co-operation and altruism. The more we unravel the underlying universal factors that give rise to our individual and societal values, the more we’ll be able to set aside our culturally based prejudices and cooperate on a global scale.
Thanks for an interesting article.
David Keating
For whatever it’s worth, it seems to me that a fundamental question in any sort of evolutionary model of ethics is — what’s the unit that’s assumed to evolve? I suspect that ethics are culturally transmitted, and that they confer fitness to the culture as a whole, not to individuals in the culture directly. Individuals might survive or even thrive using a defect-and-migrate strategy, but a society of individuals that all worked that way wouldn’t last long … it would soon disappear and be replaced with a society that transmitted more conventionally ethical values.
For whatever it’s worth, it seems to me that a fundamental question in any sort of evolutionary model of ethics is — what’s the unit that’s assumed to evolve?
That is, what is the unit of selection? I allude to this question in my last paragraph, when I ask, what is the self? This is closely connected to the problem of understanding altruism. I tend to believe that selection takes place simultaneously at multiple levels. In addition to the various levels for biological evolution (gene, cell, individual, and group levels), there are various levels for cultural evolution.
As W. D. Hamilton concluded his paper on Price’s equations regarding levels of selection Innate Social Aptitudes of Man: “Some of the main points of this paper can be summarized as an answer to this comment: that often, in real life, there is a law, and we can see why, and that sadly we also see the protean nature of this Dilemma, which, when suppressed at one level, gathers its strength at another.”
The attempt to eliminate the Dilemma by making a unified identity of all identities may be achievable but it does not mean the resulting unified identity is viable and, indeed, if a component replicator has become adapted to migration for survival, it has also evolved virulence thereby making such unity almost certainly nonviable. Better to have some uninfected firewalled identities despite the fact that this invites the same old Dilemma. At least with the Dilemma, you might have some firewalled identities somewhere that are capable of filling the population void created by non-firewalled identities.
I am extremly suspicious of using logic to “improve” ethics because of 2 main drawbacks:
- Whatever the “best” solution would happen to be it will have to be enforced one way or another (for it will not be a natural outcome of social interactions), this raises the question of the authority which will enforce compliance and the means it will use, not to speak of the legitimacy of enforcing rules (of any kind) unto non consenting people.
- Introducing yet another set of constraints on social behaviors changes the whole “landscape” (the phase space) of social interactions and I am quite sure that the sought for “ideal solution” cannot anticipate the consequences of its own application since it is already utterly difficult to sort out what is desirable from what is not (abortion and the like…)
Ethics is a mess and (unfortunately…) very likely to stay so, “drastic”, “rational” approaches are bound to bring more harm than good, more humble and pragmatic ways are certainly safer.
Ethics is a mess and (unfortunately…) very likely to stay so, “drastic”, “rational” approaches are bound to bring more harm that good, more humble and pragmatic ways are certainly safer.
Evolution is incremental. I do not advocate drastic, radical change. I advocate tiny, incremental steps, with constant checking and backtracking as necessary.
it will not be a natural outcome of social interactions
You suggest a dichotomy:
(1) logic, rational, drastic, enforced
versus
(2) natural, social, humble, pragmatic
I think this dichotomy reflects several false assumptions. Logic is not unnatural. Rationality can be pragmatic and humble. Authority and enforcement exist in current “natural” political systems.
Here is my core point: ethics and morality should be based on experimentation and observation. In a sense, this is the basis for our current ethical systems, but the experimentation and observation has been informal and subconscious. I am arguing that we should use more formal and rigorous methods (the tools of science) to further evolve our ethical systems. Science itself is the product of evolution, so there is no dichotomy here. This is merely the evolution of evolution.
Here is my core point: ethics and morality should be based on experimentation and observation.
This is indeed the “core point”! Because of the “should”: where does this “should” come from if not an already moral judgement?
Furthermore, though “experimentation and observation” would not be such a bad idea (it is what all anthropologists actually do already), the assessment of the results requires preexisting agreement on ethical values for everyone involved (not just your own ethical prejudices like in the above paragraph). There is a hopeless circularity in this search for “rational ethics”, and the more rational you try to be the worse it will become. It is (paradoxically, for you, maybe) an utterly naïve and unenlightened position; the Romans knew that already:
fiat justitia et pereat mundus / fiat justitia ruat caelum
— Lucius Calpurnius Piso Caesoninus (d. 43 B.C.)
Because of the “should”, where does this “should” come from if not an already moral judgement?
The whole essay above is directly addressing this very question. Ethical systems are for resolving conflict and facilitating cooperation. If science can show you that ethical system A is better for resolving conflict and facilitating cooperation than ethical system B, then you should prefer system A over system B, simply because A will be a better tool for you to attain your own personal goals, in the long run, whatever those goals might be. Ultimately, the “should” comes from you, from your own personal goals. This is what is meant by enlightened self-interest.
I highly recommend that you read The Evolution of Cooperation, if you haven’t read it yet. It will give you a good example of how science can be used to evaluate ethical systems.
If science can show you that ethical system A is better for resolving conflict and facilitating cooperation than ethical system B, then you should prefer system A over system B, simply because A will be a better tool for you to attain your own personal goals
Once again you are missing my point. I am just as fond as you of rationality and very much caring about “enlightened self-interest” and I knew about Iterated Prisoner’s Dilemma well before you tried to “educate” me. I will not try to reciprocate because the basis of the misunderstanding is not a lack of information or “intelligence” on your part but some sort of blind spot about how other people could have a different opinion than you, most especially about rationality. But this “blindness” is unfortunately not peculiar to you, try to sell an Iterated Prisoner’s Dilemma conciliation about Mohammed pictures to an Islamist for instance …
To get back to the “personal” word I emphasized in your sentence, what if the other party prefers system B? The other party can furthermore care not the least about “science”!
To get back to the “personal” word I emphasized in your sentence, what if the other party prefers system B?
In terms of the Iterated Prisoner’s Dilemma, this is analogous to asking, what if I am using tit for tat, but the other player always defects? Axelrod’s experiments show that this is not a problem. In the long run, tit for tat will perform better than always defecting. Similarly, there are numerous experiments that show that “evil” invaders cannot compete in a population where tit for tat is common.
A strategy that is based on reality (experimentation and observation) will win in the long run. It is not necessary to argue with the opposition until they see the light. They may never see the light. But that is their loss, not mine.
The real conflict isn’t so much between experimentation and, shall we say, theorization — that conflict pretty much went the way of Protestantism vs theocracy during the enlightenment. The real conflict is between the territorial range of ecological hypotheses. Ecological correlations are, of course, inadequate for certainty “beyond a reasonable doubt” — the kind of certainty one needs in order to ethically deprive someone of their freedom to pursue their own ecological hypotheses (there really may be no other kind of freedom in the final analysis). It is absolutely essential that experiments be carried out with territories appropriate to the ranges of the ecological hypotheses being tested.
To take a quick example from libertarian thought: Many libertarians object to heroin laws because they say that an individual’s use of self-prescribed heroin does not damage others. The non-libertarian view is essentially ecological: there are negative externalities involved with the heroin trade which justify prior restraint — not remedial tort law or any such libertarian notion. So we have two hypotheses in conflict and in order to test them against each other territory must be allocated. Clearly, there needs to be more detail to the ecological model purported by the heroin law proponents so that there may be some space somewhere on the planet where people might test the libertarian ideal of no-negative-externalies heroin use.
And now we are in the sticky wicket of economics.
Might I suggest something?
Prior to government there is this thing called animal territory which also corresponds to habitat with carrying capacity of 1 (or 4 if you want to speak of a reproductively viable household as a territorial unit). If we are to respect the individual’s choice of hypotheses, then let us put government to the single task of allocating territory to groups of individuals of like-mind — meaning sharing an ecological hypothesis. Government’s job is then to assist in not only assortative migration, but also in territorial reallocation so that there is some sort of reasonable carrying capacity for the people pursuing an ecological hypothesis.
The question arises as to whether this doesn’t simply boil down to a competition to find the ecological hypothesis/genotype combination that results in females churning out the most babies per unity time. Perhaps. Is there something better? Perhaps. This is where economics comes into the picture to hopefully save us from Malthus. One may presume within this hypothetical laboratory of the scientifically ethical States, that the total carrying capacity of the Earth is not exceeded. If one then places the remainder of carrying capacity “in play” in some sense, and requires that prior to being allowed to reproduce, a couple obtain or create sufficient carrying capacity to support a child.
OK, so we have essentially replaced a huge laundry list of vague and selectively enforced human rights with a few simple rules that, while they may be somewhat vague and somewhat selectively enforced, as still far clearer, simpler and more subject to numerical verification of various ecological hypotheses.
The real conflict isn’t so much between experimentation and, shall we say, theorization
I wasn’t thinking of experiment as opposed to theory, but rather science (that is, experiments and theories based on experiments) as opposed to non-science. By “non-science”, I mean something like superstition or mythology: beliefs that perhaps seem plausible, but have not been rigorously tested. In contrast, with science, the testing never ends; everything is open to revision.
let us put government to the single task of allocating territory to groups of individuals of like-mind — meaning sharing an ecological hypothesis.
This is an interesting idea for a science fiction story (didn’t Heinlein write a story or two like this?), and there are historical cases where something like this has happened (remote colonies in sparsely settled land), but it seems to me that this proposal is not a stable scenario (it is not viable or sustainable): if you were able to create a situation like this, it would very soon change into something different (for example, civil war, rebellion, revolution, or dictatorship).
Experimentation is difficult in politics and economics, but there are ways around this problem (methods that are less disruptive than “territorial reallocation”). One way is to use passive observation instead of active experimentation. Scientists are developing increasingly sophisticated mathematical and statistical tools for inferring causal models from observational data. Another way is to use computational simulations, like Axelrod’s experiments. Yet another way is to involve human subjects in a kind of game-based experiment. This latter type of experiment has revealed interesting things about altruism.
Scientists are developing increasingly sophisticated mathematical and statistical tools for inferring causal models from observational data.
The philosophy of causality, even in such a hard science as physics, is so fraught with controversy and ambiguity that it is really hard to believe there is anything approaching a scientific consensus of inferring causal models from observations of uncontrolled ecologies.
In any case, this is beside the point since there is a more fundamental problem with consent. If we aren’t to start with consent as axiomatic, then how are we to derive it from science or, alternatively, to explain scientifically how it is ethical to impose treatments on subjects without their consent?
The philosophy of causality, even in such a hard science as physics, is so fraught with controversy and ambiguity that it is really hard to believe there is anything approaching a scientific consensus of inferring causal models from observations of uncontrolled ecologies.
I think there was a breakthrough in our understanding of causal models with the publication of Probabilistic Reasoning in Intelligent Systems (Pearl, 1988) and Discovering Causal Structure (Glymour et al., 1987), and more has been learned since then. Social policies are necessarily based on some kind of causal model, either scientific or non-scientific. Correlation is insufficient for active intervention. If you reject observational data, we still have computational simulations and game-based experiments.
In any case, this is beside the point since there is a more fundamental problem with consent. If we aren’t to start with consent as axiomatic, then how are we to derive it from science or, alternatively, to explain scientifically how it is ethical to impose treatments on subjects without their consent?
In the kind of disruptive experimentation that you describe (“territorial reallocation”), I can see that getting the consent of the experimental subjects is a big issue. But I don’t see it as much of a problem in the three alternative methods that I mention (passive observation, computational simulation, game-based experiments). In general, consent is a kind of cooperation, so what I say above about cooperation and conflict applies to consent. Agents can cooperate even when they don’t share any ethical axioms. Sharing ethical axioms may make cooperation easier, but it is not necessary.
I think I understand your perspective better now. You aren’t so much concerned about a process by which a set of shared ethics are established as you are with the pragmatics of how an individual relates to a world in which there may be no shared ethics at all.
Why don’t you use the word “pragmatics” rather than the word “ethics”?
the pragmatics of how an individual relates to a world in which there may be no shared ethics at all
Why do you assume that ethics must be shared? I know that this is a traditional philosophical assumption, but it doesn’t seem necessary to me. I think it comes from the axiomatic approach to philosophy. Instead of “ethical axioms”, I prefer to think of “ethical algorithms”. Although there is a close connection between computation and logic, I find “algorithm” more appropriate than “axiom”. In the Iterated Prisoner’s Dilemma, tit for tat can be easily expressed as an axiom, but most of the systems that have been proposed are better described as algorithms. Now, if you ask what two algorithms share, it might not be anything that could be readily expressed as an axiom.
I think the hope is that we will be able to find an ethical axiom (or set of axioms) which will win universal (or very wide) acceptance, resulting in greatly increased harmony in the world. I doubt very much that this will ever happen. What I see happening instead is the evolution of increasingly complex institutions, social structures, legal systems, and economic systems, which are best seen as complex algorithms, rather than simple axioms. When two algorithms interact, they do not need to first identify shared axioms; rather, they only need to agree on a protocol for communication and cooperation. Consider all the people you cooperate with in the course of a day. Do you need to find shared axioms before you cooperate with them?
Do you need to find shared axioms before you cooperate with them?
“Need” is a pretty loaded word. The reality is that the most rigorous research yet conducted on this topic has observed that the less shared within your human ecology, the lower the social capital: trust, cooperation, etc. The finding was so controversial that the principle investigator had to hold off on publishing his work for a decade because the idea that “diversity” (really, novel diversity within a human ecology) is socially beneficial is so “axiomatic” that careers are terminated for questioning it! Now, clearly, made sense, from the researcher’s desire for “the good life”, for him to put off delaying the results of his research so that he could compose a closing chapter describing how it really isn’t all that bad if people choose the right “ethics” — but was it really ethical for him to deprive the public of his findings just because he happened to protect himself from a backlash from an academic community more interested in enforcing “axioms” of social policy than in finding the truth?
The reality is that the most rigorous research yet conducted on this topic has observed that the less shared within your human ecology, the lower the social capital: trust, cooperation, etc.
According to the link you give, the paper only shows correlation, not causation. Furthermore, it seems that the paper has nothing to do with ethical axioms or principles.
the paper only shows correlation, not causation
And so here we are… with some evidence against an hypothesis of how cooperation is increased through “diversity” — an hypothesis that is currently enforced by governments on subjects without their consent — and the scientific burden of evidence is reversed: placed on those whose consent is being violated with the support of the “correlation doesn’t imply causation” argument.
it seems that the paper has nothing to do with ethical axioms or principles.
Your terminology of “algorithms” is just fine to use here: Ethnicity is an ecosystem of algorithms, both learned and inherited — with varying degrees of being deliberately engineered vs naturally selected. Your sense of the word “ethics” then seems to differ mainly in the selection criteria being for “the good life” rather than evolutionary fitness. However — and here is the rub — there must be very large overlap between the two. Hence it is reasonable to claim that the Putnam paper has much to do with ethical algorithms.
“correlation doesn’t imply causation”
If A and B are correlated, then either (1) A causes B, (2) B causes A, (3) C causes both A and B, (4) the correlation between A and B is due to random noise and will go away when more data are collected, or (5) A and B are part of a system with feedback loops, and it is not meaningful to ask whether A causes B or B causes A — they cause each other.
It is a fact that correlation does not imply causation. However, there are techniques for inference of causal models from observational data. That is, there are ways for determining which of the above five cases hold. Unfortunately, the research that you cite does not use any of these ways.
Hence it is reasonable to claim that the Putnam paper has much to do with ethical algorithms.
This is a hypothesis. It needs testing.
Peter: The function of ethics and morality is to resolve conflict among agents; to facilitate cooperation among agents. A solitary agent … would have no need or use for morality.” …
Ethical agents select ethical systems that are fitter than competing ethical systems, in terms of their ability to facilitate cooperation and reduce conflict with other agents.
I think that the definition of ethics that you’ve given is too narrow. What you have captured is “contractualist” ethics, a la Rawls. What about utilitarian ethics? What about consequentialist theories?
Anyway, it’s nice to see someone else trying to make some kind of rational sense of ethics. I am trying to do something similar on my blog, and I’m taking a somewhat different route.
… and I might just add: you disagreed with the traditional view that
“Science can help us predict the consequences of our actions, but it cannot tell us which consequences we should seek and which we should avoid.”
- I dislike this view too, and am trying to find a way around it. But if you say:
“The function of ethics and morality is to resolve conflict among agents; to facilitate cooperation among agents.” (implicitly: agents who have different goals)
then you have not given any one agent a scientific way of deciding which consequences we should seek and which we should avoid.