For the past year or so I've been trying to digest some ideas on the intersection between morality, ethics, responsibility, and the interaction between an individual and society at large. The most noticeable result of that is that I have been using the word "karma" a lot in my internal monologue, and increasingly, also when talking to other people. This is worrying me insofar as might not always obvious to everyone that I use that term in a very overloaded way, and the general conception of karma is that of a supernatural force that punishes people who do bad things and rewards them when they do good things. I don't personally believe in such a force and would like to avoid this association. But I didn't really pick this word deliberately and in any case aside from supernaturality it actually has a lot of correct and useful associations, or at least is the "nearest mapping" from my mentalese to English that I could find for a certain vague concept I have. So this blog post exists so that when someone calls me out on my idiosyncratic usage of karma, after I say something like "the question of whether we should act ethically towards language models doesn't really hinge on whether they have internal experience, but on whether they have karmic grounding, or a connection with something that does", I can point them here and clear up the confusion. (I do like the naive reading of the quote though, makes me sound like a cyberbuddhist from some pseudospiritual sci-fi like The Matrix.) The added benefit is that this gives me an opportunity to stop and clarify to myself what I actually mean by this myself.

So to explain what I actually mean by karma, lets start with some of the more traditional meanings of the word karma that I actually like. In Buddhism karma is often framed as habbituality, a set of conditioned responses to stimuli. Take this quote from an article by a Buddhist Lama Ngak’chang Rinpoche:

The ‘law of karma’ is different from externally enforced societal law – because ‘karmic law’ is directly consequential and self-implementing. We perceive the world in a certain way – and react to it in accordance with that style of perception. That is what is meant by karma. […] The responses we make to our environment will remain the same and we will attract the kind of circumstances which will match our perception. If we feel impoverished, we experience the objects of our perception as confirming our impoverishment. We tend toward aspects of life that show us what we want to see. We continuously recondition ourselves.

Admittedly this is not a precise framing, but from this description karma sounds more like a self-fulfilling prophecy, or a vicious circle.

Here is another quote from the article, this one much more to the point (parts in bold highlighted by me):

We may well need to take a look at ‘law’, in order to understand [karma's] function. The existence of laws within a society means that there is little or no awareness in that society. It means that such a society has no confidence in awareness or personal responsibility. The need of law signifies that we cannot trust or rely on awareness – because law is instituted as a substitute for awareness and personal responsibility. Where there is awareness and personal responsibility, there is no need for law. Where there is awareness and personal responsibility, there is no need for rules. Where there is awareness and personal responsibility, there is no need for moral or ethical values. Where there is a lack of awareness and personal responsibility – we rely on laws, rules, ethics, and morality.

This might sound somewhat controversial at first glance, but makes perfect sense when you think about it. All he really says here is that explicit law is largely necessary for social coordination in presence of imperfect information ("awareness") and dealing with hostile/unaligned actors ("personal responsibility"). If you had perfect awareness and personal responsibility in some group of people, such a group would probably not need laws. (I would add that most people tend not to give themselves explicit laws to follow either, unless they expect themselves to be lacking on one of those two fronts.)

Where there is no awareness and personal responsibility law serves a function – but law undermines personal responsibility and obscures awareness.

So, we have our societies where laws and rules are implemented or enforced, either for the benefit of people or to their detriment. This is inevitable with any fixed measure because no law, rule, or moral standard can possibly apply in all circumstances. Laws, ethics, and morality are always expedient, and their implementation approximates to awareness and personal responsibility. […] Morality is a means, not an end. Seeking the ideal and infinitely applicable moral philosophy is a fruitless quest. Morality is an expedient device which we employ skillfully until we find awareness.

So, lets put the two parts together. Karma is conditioned response that is meant to approximate a deeper and more fluid mode of ethical reasoning that for reasons of either lack of trust or information scarcity cannot be used directly. I like this definition and I think it broadly agrees with what I mean by this term. But it doesn't capture any of the aspects that I actually like or find useful about karma.

People usually think about karma in context of an action of an individual, whether doing something gives you "good karma" or "bad karma". This is legible under the above definition, but not very useful. If your moral approximation of what you would have done with perfect information and perfectly good intentions is decent, you got "good karma", if you messed up or just didn't care at all it is "bad karma". Perhaps the cultivation aspect comes from the idea that doing more of the good approximation makes you better at doing the approximation which is probably true often enough (though obviously not always). But this misses one aspect often associated with karma: aren't you, at least under most interpretations, supposed to "get" something in return for your hard work? "Karma is a bitch", and so on.

Here is where I find karma starts to get really interesting and useful as a concept: consider the effects of habituation not on a single individual, but on an entire society. A dynamic, self-arranging system of people trying to figure out better ways of acting morally, while weeding out defectors (or at least changing the incentive structure in such a way that they have no choice but to act in accordance with the karmic approximation of morality given to them). Yes, laws and centralized governance are a part of it, sure, but the process is much, much, much more fine-grained than this. It is a distributed system of billions of agents who are trying to organize themselves around constraints, while having heavily imperfect information about themselves and each other, and which are constantly defecting and un-defecting from the process itself. And all of that is before we Parifett it and realize that each one of these agents is a temporal sequence of snapshot sub-actors which also have to consider past and present versions of themselves under those same rules. All of this happening either unconsciously or hiding in plain sight under such concepts like "governance", "justice", "responsibility", "civil rights", "research in moral philosophy", and so on.

I think what people mean as "naive cosequentialism" is largely just "karmically unaware consequentialism". Why isn't doing the better thing always right? Well, because there is this giant machine chugging along in the planet's biosphere that you are fucking up by not making yourself legible to it.

I don't want to elaborate on this too much because critiques of naive cosequentialism are a well-trodden topic, but I think its important to stress that this high-level overview of karma in no way gives credit to its incredible complexity. I think one could write multiple beautiful interdisciplinary works of writing just sketching out the implications of this framing of karma, and drawing connections to existing frames in moral philosophy, psychology, computer science, anthropology, and probably a billion other fields. As an anecdote, I have tried writing a closing section to this article three times, each time trying to explore a different "one small concept" in my view of karma, before giving up realizing there was actually way too much that one can write on that topic.

David Chapman once said that the answer to the question "if not Bayesianism, then what?" was "all [the rest] of human intellectual effort". I think you can make the argument that even if one rejects naive consequentialism in favor of karma-aware one, "all of human intellectual effort" is still not enough to make sense of the structure of the karma machine at all. Though I do think it is a good start.