Kindness Against The Grain

Epistemic Status: Unformed Thoughts

I’ve heard from a number of secular-ish sources (Carse, Girard, Arendt) that the essential contribution of Christianity to human thought is the concept of forgiveness.  (Ribbonfarm also has a recent post on the topic of forgiveness.)

I have never been a Christian and haven’t even read all of the New Testament, so I’ll leave it to commenters to recommend Christian sources on the topic.

What I want to explore is the notion of kindness without a smooth incentive gradient.

Most human kindness is incentivized. We do things for others, and get things in return. Contracts and favors alike are reciprocal actions.  And this makes a lot of sense, because trade is sustainable. Systems of game-theoretic agents that do some variant of tit-for-tat exchange tend to thrive, compared to agents that are freeloaders or altruists. Freeloaders can only exploit so long until they destroy the system they’re exploiting, or suffer from the retribution of tit-for-tat players; pure altruists burn themselves out quickly.

Sometimes kindness is reciprocated at the genetic rather than the personal level (see kin selection.)

Sometimes it’s reciprocated by long-term or indirect means — you can sometimes get social credit for being kind, even if the person you help can’t directly reciprocate. A reputation for generosity to allies and innocents makes you look strong and worth allying with, so you come out ahead in the long run.

And one of the ways we implement the incentives towards kindness in practice is through sympathy. When we see another’s suffering, we feel an urge to be kind to them, and a warm fuzzy reward if we help them.  That way, kindness is feasible along local emotional incentive gradients.

But, of course, sympathy itself is carefully optimized to make sure we only sympathize with those whom we’d come out ahead by helping. Sympathy is not merely a function of suffering. It is easier to sympathize with children than with adults, with the grateful than the ungrateful, with those who have experienced culturally acceptable “grounds for sympathy” (such as divorce, loss of a loved one, illness, job loss, crime victimization, car trouble, or fatigue, according to this sociological study).  We sympathize more with those whose suffering is perceived as unjust — though this may be something of a circular notion.

This leaves out certain forms of suffering.

  • The stranger, who is not part of your group, will receive less sympathy.  So will the outsider or social deviant.
  • The person with a permanent problem that can’t be easily fixed will eventually receive less sympathy, because he cannot be restored to happiness and in a position to show gratitude or return favors.
  • The overly self-reliant person will receive less sympathy; if sympathy is like a “credit account”, the person who has never opened one will be offered less credit than one who maintains a modest balance. We require vulnerability and a show of weakness before our sympathy will turn on.
  • The angry or assertive person who does not show gratitude or deference will receive less sympathy.  Appeasement displays evoke sympathy and reconciliation.
  • The person whose suffering takes an illegible form will receive less sympathy.

To be a recipient of sympathy one must be both weak and strong; weak, to show one really has received a misfortune; strong, to show one can be a useful ally someday. Children are the perfect example, because they are small and vulnerable today, but can grow to be strong as adults.  The victims of temporary and easily reversible bad luck are in a similar position: vulnerable today, but soon to be restored to strength.  Permanently disadvantaged adults, people who may be poor/disabled/nonwhite/etc and have developed the self-reliance or resentment associated with coping with long-term deprivation that isn’t going away, are less easy to sympathize with.

Some of this has been shown experimentally; subjects in an experiment who viewed other subjects appearing to receive electric shocks were more unsympathetic when they were told the shocks would continue in a subsequent session, versus when they were told the shocks had ended, or when they were told that their choices could stop the shocks. Permanent suffering is less sympathetic than temporary or fixable suffering.

Sympathy provides an immediate emotional incentive to respond to suffering with kindness, and it’s pretty well calibrated to be “good game theory” — but it’s not perfect by any means.

Cooperation Without Sympathy

Imagine a space alien — a grotesque creature, one whose appearance makes you want to vomit — offers you a deal. Let’s say this alien is, like the creatures in Octavia Butler’s Xenogenesis trilogy, a “gene trader”, one who can splice DNA with its bodily organs, and has a drive towards genetic engineering analogous to what Earth animals experience as a sex drive.  If you have “sex” with the alien and produce part-alien babies, it will give you and your children access to the vastly advanced powers in its alien genes, in exchange for gratifying its biological urge and allowing it to benefit from your genes.

From an intuitive standpoint, this is grotesque. The alien is not sexy. You cannot feel compassion for its desires to trade genes with you. It feels violating, disgusting, unacceptable. You were never evolved to want to breed with aliens.

And yet the game theory is sound. Superpowers are a grand thing to have. Even sexiness exists as a way to incentivize you to have strong children — and your alien children will undoubtedly be strong.

It’s a game-theoretic win-win but not a sympathetic win-win. Other humans will not find your alien babies sympathetic, or your choice to cooperate with the aliens a pro-social one.

It’s a sort of betrayal against your fellow humans, in that you are breaking the local game of “sex is between humans” and unilaterally gaining superpowered alien babies; but it’s a choice that any human could make as easily as you, so you aren’t leaving others permanently worse off, or depleting a valuable commons. Since all humans would be better off with alien genes, it’s not really a “defection” if you take the lead in doing something that would be beneficial if done by everyone.

Butler is really good at expressing how a “peaceful win-win” — on paper, an obviously correct choice — can feel disgusting.  Sympathy incentives can’t get you to win-win cooperation, if the thing that the other person wants is not something that you can imagine wanting.

This is an example of incentives for cooperation being present but not smooth.  It is in your interest to “gene trade”, but you only know that intellectually; you cannot be guided to it naturally through sympathy.

In the same way, helping someone “unsympathetic” but valuable is a “good investment” but doesn’t feel like it.  You often hear about this in disability contexts. “All you have to do is give me a relatively cheap accommodation and suddenly I become way more productive! How is this not a good deal for you?”  Well, it may be a good deal but not a sympathetic deal, because people’s mental accounting doesn’t match reality; if they think that the person “ought to be able to” get along without the accommodation, sympathy doesn’t provoke them to help, and if they don’t have a strong intuitive sense of people being plastic, so that they function differently in different environments, they don’t really intuitively believe that a blind person can be an expert programmer if given a screen reader, for instance.  Abstractly it’s a good deal, but concretely it’s not being guided smoothly by emotional gradients, it requires an act of detached cognition.

In practice, you can guide a situation back to sympathy, and that’s usually the best way to get the trade done. Try to play up the sympathetic qualities of the trade partner, try to analogize the requested action to things that are considered moral duties in one’s social context.  Try to set up emotional guardrails, engineer the social environment so the deal can be done without abstract thought.

But this isn’t really feasible for a single individual to do.  If you’re alone and nobody wants to help you, even if you reciprocate, because you’re not a “sympathetic character”, you can’t reshape social pressures to make yourself sympathetic all by yourself.  If we aren’t going to brutally destroy the lives of valuable people who don’t already have a posse, somebody is going to have to think, to go beyond gradient-following.

I think that to get the best results, thought is actually necessary.  By “thought” I mean the God’s-eye view, the long-view, the ability to ask “where do I want to go?” and potentially have an answer that isn’t “whichever way I’m currently going.” But what emotional or psychological or behavioral scaffolding promotes thought?  We are, after all, made of meat.  Since sometimes humans do think, there must be a way to build thought out of meat.  I’m still trying to understand how that is done.

Forgiveness and the Very Long Term

Forgiveness, on a structural level, is choosing not to call in a debt. I’m entitled to compensation, according to the rules of whatever game I’m playing, but I don’t demand it.

Forgiveness is a local loss to the forgiver. If everyone forgave everything all the time, it wouldn’t be remotely sustainable.

But a little bit of forgiveness is useful, in exactly the same way that bankruptcy is useful.  Bankruptcy means that there’s a floor to how much debt you can get in, which allows loss-averse humans to be willing to take on debt at all, which means that more high-expected-value investments get made.

Tit-for-tat with forgiveness outperforms plain tit-for-tat.

You can also think of forgiveness as a function of time. If you expect that someone will be net positive to you in the long run, you can accept them costing you in the short run, and not demand payment now. In other words, you extend them cheap credit.  As your time horizon goes to infinity (or your discount rate goes to zero), it can become possible to not demand payment at all, to forgive the loan entirely.  If it doesn’t matter whether they pay you back tomorrow, or in a hundred years, or in a thousand, but you expect them to be able to pay someday, then you don’t really need the repayment at any time, and you can drop it.

This is sort of similar to the heuristic of “be tolerant and kind to all persons, you never know when they might be valuable.” The fairy tales and myths about being kind to strangers and old ladies, in case they’re gods in disguise. You don’t want to burn bridges with anybody, you don’t want to kick anybody wholly out of the game, if you expect that eventually (and eventually may be very long indeed, and perhaps not within your lifetime), this will pay off.

Tit-for-tat or reinforcement-learning or behaviorism — reward what you want to see, punish what you don’t — makes a lot of sense, except when you factor in time and death. If you punish someone so hard that they die before they have a chance to turn around and improve, you’ve lost them.

And, on a more abstract level: it can make sense to disincentivize the slightly worse thing in general, that’s how evolution works, but that leads to things like rare languages dying out. Yes, it’s perfectly rational to speak Spanish rather than Zapotec, and Zapotec-speakers need to make a living too, but my inner Finite and Infinite Games says “wouldn’t you like to preserve Zapotec from dying out altogether? Couldn’t it come in handy someday?”  Language preservation is an example of preserving a “loser” because, if the world went on forever, nothing would be permanently guaranteed to lose.

It’s like having a slightly noisy update mechanism. Mostly, you reinforce what works and penalize what doesn’t. But sometimes, or to a small degree, you forgive, you rescue someone or something that would ordinarily be penalized, and save it, in case you need it later. In gradient descent, a little stochasticity keeps you from getting stuck on local maxima. In economics, a little bankruptcy or the occasional jubilee keeps you from getting stuck in stagnant, monopolistic conditions. You don’t ruthlessly weed out the “bad” all the time.

Sometimes you throw some resources at someone who “doesn’t deserve them” just in case you’re wrong, or to get out of the nasty feedback loops where someone behaves badly in response to being treated badly.  If you unilaterally gave them some help, you might allow them to escape into a cooperative, reciprocal-benefit situation, which you’d actually like better!  Even if this didn’t work one particular time, doing it in general, at some frequency, might in expectation work out in your favor.

A sense of the very long term may also make sympathy easier, because in the very long term nothing is permanent and everything is eventually mutable. If permanent suffering is what makes people unsympathetic, then a sense of the very long term makes it possible to realize that under different circumstances that person might become fine, and thus their suffering is ultimately the “temporary kind” that can elicit sympathy.  “The stone that the builders rejected/ has become the cornerstone” — well, if you wait long enough, that might actually happen. Things could change; the “loser”‘s or “villain”‘s status on the bottom is not eternal; so with a long-enough-term mindset it’s not actually appropriate to treat him as definitively a “loser” or a “villain.”

Forgiveness can be a lot easier to implement than “cooperation without sympathy”, which requires you to actually ascertain where win-wins are, with your mind. You can mindlessly add a little forgiveness to a system.  Machine-learning algorithms can do it.  Which may make it a useful tool in the process of “trying to build thought out of meat.”



20 thoughts on “Kindness Against The Grain

  1. Forgiveness, on a structural level, is choosing not to call in a debt.

    This is forbearance, an agreement to collect slower than was initially specified.

    Forgiveness is a bit stronger than this. It’s publicly *writing off* the debt so that you will *never* have that enforceable claim again. It’s useful in bankruptcy when debtors collectively do better if they own little enough of an enterprise’s future earnings that it has some incentive to perform well.

  2. i wouldn’t think of slower vs never as a quantitative distinction? the point of slowness is that it’s a little bit of a lot of slowness, and a lot of slowness acts exactly like never

  3. You can permanently lose even with a very long time horizon. Even if your gamble eventually starts paying off to some degree, you might’ve fallen behind where you would’ve been otherwise and never catch up. Factoring in opportunity cost, it’s not just a matter of whether your investment is net positive, but whether it’s better than however else you could’ve used it.

  4. I feel I’m missing something with Butler’s example, in that it isn’t triggering any “this is bad” or “you’re defecting” instincts of mine. My only reservations are that I don’t want to have physical sex with the alien (but would be totally fine with sperm donation), and I don’t want to find my children viscerally disgusting if it would interfere with my ability to ensure they had a happy life. Am I wrong in thinking most of the force of this thought experiment comes from elsewhere? Where is the grotesqueness supposed to come from?

    • I feel the way you do, and I suspect it’s because we’re dudes. We’re a lot less picky where sex is concerned.

      On an intellectual level, I’d have reservations. The alien is motivated by a biological compulsion, which probably doesn’t perfectly align with the offspring’s best interest. Honestly, the alien probably has very little idea of what traits are adaptive for humans. Also, when the gratification of basic biological urges is at stake, aliens probably aren’t much more honest than people are.

      On the other hand, if it’s late and I’m drunk, I might go home with it. It wouldn’t be the first time I’ve woken up in regrettable company. Give it a fake phone number and get the hell out.

      • Okay, I think part of this is that sex and “creating children” are orthogonal to me more than most other people have them. I parse the alien’s request as “creating children”, and don’t apply any of my internalized thinking about sex to it, as it isn’t about sex.

      • Eric – I was overstating things. My point was that, for evo-psych reasons, the set of emotions surrounding sex, reproduction, and parenthood are very different between women and men. This particular fictional scenario seems to trigger quite a bit of revulsion in women. For men, it triggers a lot less disgust, which means we may be more open to the idea. Of course, we’re also more likely to up and leave a few years later if it doesn’t seem to be working out.

  5. Be kind even when it isn’t obvious that it will pay off, because bad prospects could turn into good. People change!

    Be pushy, jealous, grasping, and paranoid, even when it isn’t obvious that it will pay off, because good prospects could turn bad. People change! You can mindlessly add a bit of brutality to a system.

    Well, first of all – to an extent they are both self-fulfilling prophesies, and which one would you rather fulfill? I think the answer seems more obvious to you than it actually is, because you don’t take into account the costs of friendship and positive entanglements which still entangle. You also don’t take into account the rewards of being a feisty fighty sort, and rather fixate on the cost. This might be an unbridgeable gender-driven perception gap. Winning as a woman looks like having a strong social network and not too many enemies, taking care of children, taking care of people generally. A man who takes this sort of passive friends-with-everybody approach is likely to become a nonentity with poor prospects. Winning as a man looks like getting into fights, whether serious or playful, and giving a good showing. I’m not saying you have to be Hannibal, though those types do well. I’m saying whatever the arena you have to push to make a space for yourself, even when it is just a man approaching a woman. A sort of aggression is mandatory.

    Building on this, you can be rewarded for preventative violence – meaning keeping people down. If you’re on top, the world is full of things that are fermenting, spidering around, gaining size and strength and numbers to unseat you. With a constant rain of blows you nip these problems in the bud. You discipline your dog if you want it to stay well-behaved. You can recoil from this in horror as a torturous dysfunctional dynamic, or you can look at it as essentially maintenance of the hierarchy within which at least some humans may flourish. Sharpening a blade, cleaning a gun, laying traps for cockroaches. Maintaining structure.

    Of course it is torturous for some. Is that avoidable? Can there be cooperate-cooperate equilibrium in an incredibly broad sense? I think a cooperate-cooperate equilibrium between cats and mice is self-evidently impossible. I’m not sure that a deep delve into human social dynamics wouldn’t find a world full of doves equally impossible.

    • I thought literal animal training worked best with positive reinforcement? At least that’s what Karen Pryor thinks.

      Men doing better when more aggressive is plausible to me. Though note that a good deal of my thinking on this issue is a gloss on/interpretation of men’s thoughts, both writers (James Carse) and personal friends.

      • I thought literal animal training worked best with positive reinforcement?

        This was a pet interest of mine for a while. Both animal training and positive vs negative reinforcement generally.

        I’m impressed with what dog trainers are able to do with positive reinforcement alone. This channel: is a good example. She trains her dogs with clicker sounds and rewards and seems to do pretty well.

        She’s not dealing with a pack of sled dogs or hyenas. People who deal with working dogs and people who deal with wild animals emphasize the importance of establishing and maintaining dominance. She claims this approach is wrongheaded. Maybe she’s right for housepets and wrong for animals in more high-intensity, high-stakes and limited-time situations. I’d like to see her manage a pack of sled dogs, who get into scuffles with each other constantly and who you need to extract labor from (getting from point a to point b) in order to put food on the table. It’s an interesting test, and I am by no means certain she would fail! But I haven’t seen any of these love-and-kindness types in the really rough situations.

        Also, look at the dogs themselves. They don’t deal with other dogs exclusively with positive reinforcement. As the stakes get higher, they shift further and further towards negative reinforcement. This should be a hint.

  6. Love has no need for forgiveness as it never judged in the first place & true kindness is a gift to ones Self not other. In both cases, there is no reward other than the ever deepening ever falling in Love with Self. This is the eternal cycle of Gratitude for *This*. Nothing to lose nothing to gain — it IS. Ultimately, Love of Self becomes indistinguishable from Love for other … the seeming two merge to be recognized as One.

  7. > I think the answer seems more obvious to you than it actually is, because you don’t take into account the costs of friendship and positive entanglements which still entangle.

    Most of the game theoretic simulations consider the regime in which cooperation is net positive, sufficiently so that it outweighs the costs of agents occasionally getting burned, and that groups which consistently cooperate easily outpace those that are mired in cycles of defection and retribution. Of course it is possible to stipulate a reward matrix or rule set for which this is not the case, but these may call into question the value of cooperative interactions more generally — a “game” in which the optimal state is that everyone lives by him or herself is not very interesting. And doesn’t it match up with reality too? Wouldn’t most people agree, for example, that the lifelong “reward” of a happy family life outweighs, say, the pain of a bad break-up while dating?

    > I’m not saying you have to be Hannibal, though those types do well.

    Hannibal will do well only if A) there is a dearth of information available about other agents in the game, i.e. there is no role for reputation, even for established bad actors and B) Hannibal does not pathologize the society around him, such that he has an essentially unlimited supply of willing victims and a group harboring one or more “Hannibals” does not change its behavior to a less cooperative (and less mutually beneficial) state overall. Both of these suppositions might obtain in some cases, but in general seem–at least to me–rather counterfactual.

    Of course all of these topics would be better addressed with a more mathematical formalism, e.g.

    Final note: I would like to hear the full-length explanation of why gender plays a role in the strategy selection. Of course in most game theory models sex/gender is not a relevant variable. I’m not saying it shouldn’t be–though I would like to see the case for it elucidated a bit more.

    • Both of these suppositions might obtain in some cases, but in general seem–at least to me–rather counterfactual.

      Two words: President Trump. With apologies to Abraham Lincoln, apparently you can fool enough of the people enough of the time.

      I would like to hear the full-length explanation of why gender plays a role in the strategy selection.

      The short version is that reproduction takes about 10 minutes for a man and 9 months for a woman. Since the opportunity costs are very different, the optimal strategies are very different.

      • Well, I certainly think our President is doing some damage to our society and hope that he has finally run out of victims. Time will tell.

        As for gender, we weren’t really talking about mating specifically, but rather the much larger sphere of all human interactions for which cooperation is an option. Yes, I chose mating for my example, but I set it up in a way that envisioned the mating partner as a fellow agent in a set of iterated dilemmas rather than as a resource to be fought over. The latter is more appropriate for some species, but I’m not sure that it’s the case for humans. Someone who takes a habitually brutal approach to interpersonal interactions is going to have trouble staying with someone (Trump?)– even tit-for-tat is not nearly generous enough a strategy to sustain a marriage.

        I think the claim that human reproduction takes minutes for a man (or 9 months for a woman) is false–in reality, it requires years for both parties.

  8. There’s an internally, psychologically adaptive function to forgiveness as well, and it’s the dirty little secret of this otherwise proud aspect of being human. When you say “I forgive you” someone who’s wronged you – perhaps in a way that harmed you irreparably – you’re at least partly saying, “What you’ve done is insignificant enough that I can put it out of my mind.” You’re declaring dominance over the other person in a way, and this is why once accomplished, forgiveness makes you feel better. The strongly negative reaction I often receive when telling people this, is very similar to the reaction you get when you’re pointing at other status-signalling behaviors that everyone would rather not think about as such.

    • You’re declaring dominance over the other person in a way, and this is why once accomplished, forgiveness makes you feel better.

      I think it’s slightly different – you’re not so much declaring dominance over the other person, but declaring yourself free from their dominance, and that’s what feels good.

      For as long as you keep being angry at someone, they have some power over you: and this is probably an indication of you needing something from them, or you wouldn’t have a reason to stay angry. Forgiving them is indeed an indication of their deed being insignificant enough that you can put it out of your mind, but saying “I no longer have a need to see further compensation from you” isn’t declaring dominance, it’s declaring a return to neutral relations.

Leave a Reply

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

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

Twitter picture

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

Facebook photo

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

Google+ photo

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

Connecting to %s