In this post I list many different evolutionary purposes that suffering and pleasure serve, and also list situations in which creatures will feel suffering or pleasure as an incidental effect of how evolution plays out.

Different purposes for physical sensations

Here are several different purposes served by physical discomfort. (Many of these are from this paper by van Hooff.)

Like van Hooff, I have separated these out into acute and chronic pain and pleasure. Acute pain is the sensation you get during an unpleasant experience, chronic pain is the feeling afterwards.

Acute pain and pleasure

Activation of instinctual behavior

Pain warns when there is (the danger of) injury, and that releases behaviour averting the danger and preventing further damage by withdrawal and/or defense. (This is called “acute pain”.)

There’s an important distinction there–sometimes you get warned that you’re about to be hurt, and sometimes you are warned that you have been hurt. Also sometimes the pain warns you that you’re about to be hurt more (eg if you’ve started to cut yourself accidentally, the pain might cause you to notice and stop).

“These reflexes remain after the brain connections have been severed. So, even though we may be aware of this pain, the reactions are not dependent on such awareness. This implies that the existence of such reflexes in animals does not necessarily imply that they sense pain.”

I don’t know if there are any positive experiences associated with this kind of jerky reflex.

Facilitation of learning

Suppose you are deciding whether to eat a new type of fruit, and you don’t know if it’s worth eating or not. You might just take a bite and see how you feel about its flavor. So sometimes the only interface that we have to some of our senses involves valence.

Generation of fear

The generation of fear (readiness to flee) of things or circumstances that were associated with the infliction of damage has also been regarded as an important function. Pain then figures as the unconditional stimulus in a Pavlovian conditioning scheme. This has led to the idea that fear is the conditioned response to pain.

But pain isn’t the only thing to lead to fear:

There is a variety of flight releasing stimuli which do not derive their effectiveness from an association with pain. Fear of predators, for instance, cannot primarily be based on pain conditioning, because the first confrontation with the supposed unconditional stimulus is often the last. Other mechanisms, for instance instigation by social example, must be postulated.

I think that fear of predators is also probably instinctual as well as from social example. Chickens are afraid of snakes and hawks even if they don’t know any chickens who’ve been eaten by a snake or a hawk.

Chronic pain

Encouraging recuperation

Chronic pain causes you to have behavior appropriate for recuperation. It makes you move differently, eg limping, in ways that are good for your revovery. And it incetivises “withdrawal to a secluded and safe place and abstinence from all but certain most urgent activities. Apathy, depressivity and unresponsiveness are mirrored in dejected movements and posture.”

Motivating behavior

eg hunger, loneliness.

When you get hungry, you don’t just have a lack of pleasure, but you have an active discomfort. This is sort of like “make everything painful except eating.” The same can be true for other pleasures, including addictions: Sometimes it’s not that the drug makes you feel good – it’s that doing anything other than taking the drug makes you feel bad.

This kind of pain might increase certain kinds of alertness?

Dwelling on past pain for learning

I spend a lot of time feeling guilty (or embarrassed) about things I did many years ago. One benefit of this is that if there was something to learn from recapping social mistakes that I made when I was 15, perhaps being driven to think about them by guilt would cause me to learn those lessons more effectively.

I don’t really spend time dwelling on past successes, which is kind of surprising because I’m quite proud of many of my accomplishments.

Pain and communication

Animals often communicate that they’re in pain, which counts as a behavior released by either acute pain or chronic pain.

  • “A cry of pain can, just as a cry of fear, alarm conspecifics and thus warn them for the danger; it can also help them to learn about the characteristics of possibly dangerous situations. Even if the sender can profit no more from this, natural selection will nevertheless favour the evolution of such signals, as long as relatives of the sender profit from these; according to kin selection theory, the ‘inclusive fitness’ of the sender will be benefitted by such ‘altruistic’ signal specialization.”
    • I think that the kin selection argument is weak; a stronger argument is that evolution doesn’t really care what happens to animals as they’re dying and so they end up exhibiting the response which they have evolved for less dire situations–they feel pain and fear and so cry out because that’s what they do when they feel pain or fear.
  • An animal may get support from its group members, both during the original pain-causing event and while the animal is recovering.

Situations where suffering happens for incidental reasons

So above I’ve listed ways that suffering can be useful. But I think suffering can also happen as a side-effect, rather than a direct result, of evolutionary incentives. This incidental type of suffering seems like one way that an animal’s life might end up with a total sum of suffering and pleasure which is not close to zero in expectation.

Dying

(Brian Tomasik discussed this here)

The clearest example of this is the experience of dying. If you’re dying, nothing you do matters, and so it’s evolutionarily almost irrelevant what your experience is. This might end up with you experiencing a lot of suffering, because you often die because of something bad happening which you have an aversion to.

For example, imagine you’re a really simple organism and the only sensation you need to have is fear of fire–as you wander around sometimes you get close to fires, and you have a reward function which penalizes you for going towards hot things. If the world you’re in has a temperature range between like 20 and 30 degrees Celsius, maybe you experience a reward signal that’s something like (25 - temperature), so that you’re feeling pleasure in cool places and suffering in hot ones. If the world also has firepits that you can accidentally fall into and die, and the firepits are at 100 degrees, then as a result of the reward structure that’s supposed to keep you cool, you end up experiencing -75 reward, which is extremely metabolically expensive but that doesn’t matter because you’re dying.

So evolution pushed towards a simple linear reward function here, which causes massive extra suffering in situations where the suffering is not evolutionarily disadvantageous.

People concerned about wild animal suffering have often pointed to the pain of dying as a significant reason to think that wild animals don’t have lives worth living.

Primitive pain mechanisms

Suppose you need to have a really sharp pain when you step on something sharp, so that you stop walking and don’t step down. This might be implemented by having nerves which send strong pain signals to your brain when the skin of your foot breaks. Sadly, after your skin has broken, these nerves are still going to be firing, even if they aren’t useful anymore. If evolution is limited to simple reward mechanisms (which it seems to be), creatures might end up experiencing pain after they’ve already had an injury happen.

(Everything about pain seems to be a super weird series of hacks, eg how you feel endorphins shortly after you experience pain, which inhibits the pain.)

I think that the limitations of animal pain mechanisms are really important here.

Speculations about ems and subroutines

Less work will have valence at all–better to use simple optimization functions than to use sentient creatures for it.

Less sensation per unit of reward.

EV will be closer to zero, because at the moment it’s mostly not for reasons that won’t apply in the far future.

Speculations about moral relevance and brain size

Perhaps