NB: Some of the terminology and concepts I used here are incompatible with my more recent work because this was written before I formalized my philosophy, but is dangerously close enough to them that it may cause confusion. Caveat lector.

In the future, the world may be filled with creatures other than humans and animals. There may be ems, superintelligences, and other sorts of processes we’d recognize as alive, and it seems likely that there will be many orders of magnitude more of these living things in the future than at present. As a result there may be many orders of magnitude more future suffering. Depending on your moral stance, even if you’re an anti-realist like me and simply express a preference for the satisfaction of preferences of others, this means that the lives of these future creatures are of great concern.

Brian Tomasik and others with the Foundational Research Institute have been exploring the interaction between suffering-focused ethics and the far future for a few years now. Their work has included measuring suffering and happiness, formalizing ethical calculations, and understanding the benefits of compromise with other value systems, but of particular interest to me are Brian’s looks into issues around suffering of minimally conscious processes because his thoughts on consciousness resemble mine. Thus when I read FRI researcher Lukas Gloor’s recent piece on tranquilism I was primed to find myself wondering, rather alarmingly, is feedback suffering?

To understand why, recall that feedback is another name for the process by which a subject is phenomenologically conscious and experiences itself as object. Further, most feedback is either positive or negative, meaning it causes action towards or away from a state, respectively. As humans we experience positive and negative feedback as a desire that the world be in a state other than the one it is currently in, and tranquilism suggests that suffering arises from desire since not experiencing desire yields a state we call contentment. But if we can avoid suffering by not experiencing desire, which is to say positive and negative feedback, then it looks a lot like feedback is inherently painful and thus the source of suffering. That’s distressing because it implies most existence is pain and almost all physical processes suffer constantly, so it seems to me worthwhile to make a formal study of suffering to see if we might poke some holes in this line of reasoning.

I’ll do this from an existentialist phenomenological stance, which is to say that stuff exists, stuff is only known through experience, and experience is an inseparable, directed, intentional relationship between stuffs where we term the experiencing stuff “subject” and the experienced stuff “object”. Consequently in order to be precise and to avoid equivocation I’m going to have to write the word “experience” a lot, so my apologies for the word soup that sometimes follows.

Phenomenology of Suffering

Whether or not feedback is suffering is ultimately a question of axiology, the study of value. That is, when we ask if feedback is suffering we are asking a question about how to measure the value of feedback. Since I’m an existentialist and a phenomenologist, I’m additionally willing to say that we must calculate value for ourselves rather than find it because there is nowhere for it to come from other than our experiences, and since value is a measure we must then choose what to measure against. This choice forces an inescapable telos upon value, and in this case that telos appears rooted in emotion since we think of suffering as an emotional state with negative valence. Further, since I feel positive emotional valence (happiness) when other people feel positive emotional valence and negative emotional valence (sadness) when other people feel negative emotional valence — viz. I am empathetic and care about other people — and most other people feel the same way, our telos is additionally ethical because we want to know how our actions will affect the happiness and sadness of others. Thus, as you probably already surmised, we’re going to be valuing if feedback is ethically desirable or not.

Ethical systems, being axiological systems for the purpose of judging intent, are calculi for how to combine experience via relations that yield value. We can additionally place optimization constraints on these systems, like maximizing the value of preference-satisfying experiences or minimizing the value of preference-violating experiences, to produce specific axiologies, like hedonic axiology and negative hedonic axiology, respectively. This gives us a language in which to talk precisely about axiological and ethical systems, and in this language tranquilist axiology is the axiological system where experiences are combined to maximize the value of experiences of contentment.

To be intellectually honest to myself about what tranquilism means, I need to phenomenological deconstruct contentment. Descriptions of contentment make it sound something like an experience of anhedonia where the subject experiences indifference towards suffering and pleasure, but this seems unsatisfactory because these descriptions also assign contentment a positive rather than neutral valence. Lukas similarly agrees that contentment is a happy experience when he writes that contentment is “untroubled by any cravings for more pleasure”, “experienced as completely problem-free”, and that “conscious states completely free of cravings should thus elicit very positive associations”. So it seems when one is content one is happy, indifferent to more happiness, and does not suffer, but this appears contradictory because to be content one would have to be happy about preference violation, a definitionally sad experience because negative valence emotions are the feedback mechanism we use for experiences of preferences.

This is only a problem, though, if we limit our axiology to valuing experiences and experiences of experiences. If we include experiences of experiences of experiences in our ontology rather than compressing them into experiences of experiences, thus making a systems-level demand for ontological complexity, we can understand contentment as an experience of happiness towards all experiences of experiences and use it to direct tranquilist axiological reasoning. In this way contentment wraps happy and sad experiences in an experience of happiness, making it of a different type and avoiding apparent contradiction by adding happiness to rather than changing the original experience from pleasure or suffering.

Having deconstructed contentment and understood tranquilism precisely, it appears my original concerns about feedback being suffering were confused because if suffering, a negative valence experience, is something we can be content with, then in this context suffering must be an experience of experience and thus feedback cannot necessarily be suffering because feedback can exist as a direct experience not just a meta-experience.

I seem not alone in needing to clear this confusion because our everyday use of the word “suffering” points to at least two different categories the same way “consciousness” does. Just as we can separate naive notions of consciousness into phenomenological consciousness (self-experience) and phenomenological sentience (experience of self-experience), we can separate suffering into phenomenological desire (intention to make the world otherwise) and phenomenological suffering (experience of negative valence over desire). Cleaved in this way we see that feedback is desire but not necessarily suffering, but suffering is often confounded with desire in sentient subjects because meta-experiences can be both desire and suffering. This unfortunately leaves open the possibility that the feedback of sentience is often suffering even if it doesn’t have to be.

Panpsychism and Suffering

If something must be sentient to suffer, it may seem we need only worry much about suffering among animals and animal-like things such as ems and AIs, and even their suffering is only ethically relevant as far as your empathy extends. But what if more things are sentient than we think?

Phenomenological consciousness implies a weak form of panpsychism — the idea that everything is at least a little bit conscious. But most people don’t care about the phenomenological consciousness of rocks and chairs, so there may be panpsychism but it’s a very boring kind of panpsychism that isn’t affecting anyone’s ethics. A strong panpsychism based on sentience would be another story since many people, including many effective altruists, say sentience is the criterion for ethical relevance.

Many simple control systems, such as thermostats, may have phenomenological sentience because their operation can be construed in terms of meta-experience of self. The leading mathematical theory of sentience, Integrated Information Theory, would seem to agree, likely scoring such systems as minimally “conscious” (IIT uses “consciousness” where I use “sentience”). Although it seems unlikely that something as simple as a thermostat experiences suffering in the sense of a negative valence experience of desire, perhaps there is some way in which meta-experiences we’d recognize as suffering can be defined that do not depend on negative valence emotions. This is important because it would allow us to identify suffering or suffering-like experiences in subjects that do not have something resembling the evolved emotional systems of animals.

I don’t (yet) have any further thoughts on such suffering-like experiences, but having a phenomenological deconstruction of suffering it may prove possible to make progress on understand suffering in terms of non-emotional experiences.


Okay, that wasn’t too long, but it was pretty dense, so in case you were wonder what the heck I was saying, here’s the short version:

  • Feedback is desire but not necessarily suffering
  • Suffering is a kind of feedback with negative emotional valence and animals seem to experience desire as suffering.
  • Contentment wraps up suffering in happiness by adding ontological complexity to create distance from the experience of negative emotional valence.
  • Many things may have some sentience and so may be able to suffer or experience something like suffering.
  • I’m going to think more about if there is some way to talk about suffering prior to emotion so we can considering the suffering of non-emotive sentient beings, especially for those who do not have the ontological complexity to learn contentment.


New Comment
8 comments, sorted by Click to highlight new comments since: Today at 1:26 AM

If the title is a question the answer is probably no.

I know this is Betteridge's law of headlines, but do you happen to know if it's accurate?

According to this study, the law appears to be inaccurate for academic articles.

  • Scholarly article

  • Title: Do scholars follow Betteridge’s Law?

  • Answer is no


Yes, I always thought that sufferings as pain are ok as long as it is just feedback about my wrong actions.

But there is a problem of maladaptive sufferings on almost infinite intensity, which I call “unbearable sufferings” - that is cancer pain, psychological pain which could make one to commit suicide etc. We need to find the ways to cure only unbearable sufferings and make ordinary sufferings more informative.

It's a particular sort of confused person who starts out trying to identify and quantify suffering far more precisely than they can define and detect pleasure/joy.

Under what conditions would an entity prefer to suffer for future purposes vs ceasing to exist? So far, we have mostly positive examples. Identifying why this is, in order to make more of it, seems more fruitful than identifying suffering to remove it.

Oh boy, really? Suffering? Wait till some neomarxist SJWs discover this and they will show you who's THE expert on suffering... especially in indentifying who could be susceptible to persuading they are victims (and why not some superintelligent virtual agents?).

Maybe someone could write a piece on SS (SocialistSuperintelligence). Possibilities are endless for superintelligent parasites, victimizators, guilt throwers, equal whateverizators, even new genders and races can be invented to have goals to fight for.

Could you please try to keep discussion on topic and avoid making everything about politics? Your comment does not contribute to the discussion in any way.

New to LessWrong?