Wednesday, 18 April 2012

Harris's Moral Axiom

In The Moral Landscape, Sam Harris argues for the reality of moral truths and maintains that they can in principle be discovered by science. As such, Harris is a moral realist, however he cannot base the reality of his morality on the dictates of a God in whom he does not believe. Instead, he declares that we can choose another basis for a system of objective morality.

As Harris points out, any philosophical, scientific or mathematical pursuit must start out with some basic assumptions which we hold to be self-evident. In formal systems of logic, these are usually called axioms.

All logical systems must have foundational axioms. For the familiar Euclidean geometry, it turns out we need five axioms (other forms of geometry have been discovered by altering these axioms). Arithmetic depends on a set of axioms known as the Peano axioms, without which we cannot prove that 1 + 1 = 2.

But even informally, if we wish to engage in logic at all, we must value reason and consistency just as we must reject contradictions. For the pursuit of scientific research, we must assume the axiom that beliefs should be based on evidence and not guesswork and that the behaviour of objects in the universe is not random and unpredictable but governed by a system of laws waiting to be discovered.

For any mode of reasoning, thinking or philosophy, there must be some foundational assumptions. These axioms are not derived from anything, they are simple statements which usually appear to be intuitively obvious and upon which we can base our reasoning.

After the jump, Harris attempts to found an objective system of morality upon a simple axiom.

Harris's axiom (my wording): It is morally right to seek a world where the well-being of conscious beings is maximised.
Corollary: It is morally wrong to seek a world where the well-being of conscious beings is minimised.

Harris's axiom, according to him, is simply the axiom upon which morality is based, and so it does not need to be derived from any prior known facts about the universe.

Given this axiom, the goal of maximising well-being for conscious creatures seems to lend itself to scientific study, and in Harris's view, the study of the neurological causes of well-being and happiness in particular.

While the definition of well-being is left perhaps intentionally vague, it should be clear to us that some outcomes are preferable to others, and disagreements should be relatively minor. Science can help us make decisions in attaining favourable outcomes and so provide an objective basis for moral decisions. Note that Harris does not say that this is easy, nor that we will always get it right. He is simply making the point that we do not have to resort to religion as a basis for moral decisions, nor do our moral decisions have to be baseless. We have, in Harris's framework, a solid foundation from which to analyse data and attempt to make the right choices.

In the writing of this book, I feel that Harris largely succeeds in attacking the two most prevalent viewpoints with which he most strongly disagrees, namely religious morality and moral relativism, however I'm not sure that he succeeds in establishing an entirely objective basis for morality. In later posts I will outline the main problems I see with his point of view, namely:
  1. Objective values cannot be derived from a subjective axiom.
  2. The chosen axiom is good but not unimpeachable, and therefore it is subjective.
  3. No possible moral axiom can be objective, as subjective differences of opinion would leave it open to criticism.
  4. The subjectivity of the concept of well-being means that an objective science of morality may be impossible.
  5. The subjectivity of what is considered a conscious being is problematic.


  1. I think you have over simplified Harris' conception of morality and so his argument as well (i.e. I don't think its foundations are as subjectively arbitrary as you make it out to be). The following is a general sketch of mine of the kernel of Sam's argument. (Perhaps you can help me beat it into something more formal)

    1. There exists a continuum of conscious states (or “landscape of possible experiences”; p. 198); human consciousness represents a subset of this continuum and consists of states that can be described by words such as “misery,” “terror,” “agony,” “madness,” etc. on one end and “well-being,” “happiness,” “peace,” “bliss,” etc., on the other.

    2. Furthermore, this landscape of consciousness is the only intelligible domain of value (32). For it is the only context in which any discussion of values makes sense (41). All notions of value will bear some relationship to the actual or potential experience of conscious beings (32). (I.e., things—“attitudes, choices, and behaviors”, etc.—have value only to conscious creatures as a consequence of how they affect the psychological states of conscious creatures on the continuum of consciousness (p. 12); there is no other context or medium for values to exist within or derive meaning from.)

    3. Also, the continuum acts as a measure unto itself, and each state derives its own meaning of value as it relates to all other (possible) states. ‘Well-being’ then is the subset of all those states that hold positive—conscious life promoting— value verses the rest.

    4. Conscious states are not randomly determined. Rather, they depend upon the “states and capacities” of brains (p. 8) and the physical dynamics of such within bodies, societies, the world, and the universe; i.e., “human well-being entirely depends on events in the world and of states in the human brain” (p. 2).

    5. There exists a physically objective difference (on both personal and social scales) between being healthy and unhealthy, thriving and dying, and alive and dead. And because conscious states correlate with (or are identical to) physical brain states, so does its continuum map onto the various degrees of physical health.

    6. And because “…the experiences of conscious creatures are lawfully dependent upon states of the universe…it is possible for a person to be right and wrong about how to move from one state to the other” (198).

    7. Thus, from all this it follows that morality is the proper subject of conscious creatures’ psycho/physical states as they vary from, e.g., miserable to happy and unhealthy to healthy; i.e. it is the subject of a creatures psycho/physical “well-being,” where states of health and happiness are better than states of decreasing health and happiness—i.e. dying misery—as the former are the only means by which consciousness and values exist in the first place.

    8. And, therefore, because moral values only have meaning within the context of consciousness, and conscious states vary non-randomly but intimately correlate with(or are identical to) physical facts about physical brains in relation to the world, morality is a proper subject of science.

  2. I agree that "how to maximise the well-being of conscious creatures" is a proper subject of science.

    I agree that we can use science to help us answer moral questions, given this interpretation of morality.

    I even think that it would be appropriate to call this a science of morality.

    I just don't think that Harris has demonstrated why it is objectively true that any given conscious creature should give a damn about the well-being of any other conscious creature. I don't see anywhere in that argument where this case is made.

    Again, I agree with almost all of Harris's book in practice. Using science to help answer moral questions is a good idea, and a position I would also advocate.

    My main issue is that I reject moral realism and I think Harris is incorrect to claim to have come to a truly objective basis of morality.

  3. Hi Disagreeable,

    I've just come across your blog, and I've been reading some old posts. Very interesting stuff. I agree with much of what you write, especially on AI and consciousness, but I've managed to find a few things to quibble with! I hope my comments will be of some interest, although I'm responding to rather an old post. If I seem to be mostly disagreeing, that's only because I don't bother mentioning the many things I agree with.

    I haven't read Harris's book, only his online articles. But I'm quite critical of what I've read. (FYI my own meta-ethical view is moral error theory.) I'll start by addressing the topic you raise here, that of axioms.

    I don't know whether Harris himself uses the word "axiom", but I find the expression "moral axiom" very unhelpful. Labelling a belief an axiom can too easily serve as an excuse for not subjecting that belief to sufficient skeptical scrutiny. I'm afraid the sorts of views about axioms which you give here may contribute to this tendency. If "any philosophical, scientific or mathematical pursuit must start out with some basic assumptions which we hold to be self-evident", then it seems we are powerless to scrutinise these basic assumptions. Apparently we must all just accept whichever basic assumptions feel self-evident to each of us, including moral axioms.

    This view of axioms seems to be based on a traditional, foundationalist view of epistemology. I would reject that view in favour of a more naturalized epistemology. Our knowledge does not come ultimately from arguments, and therefore does not need axioms as initial premises. Instead, knowledge starts from non-verbal non-conscious processing of sensory data. If I look out of the window when it's raining, my knowledge that it's raining is usually not the result of an argument.

    Of course we do also make arguments, particularly in the formal fields that you mention. But even there, much of the belief-forming process operates at a non-conscious level, and cannot be treated (even in principle) as the result of deductive reasoning. The underdetermination of theory by evidence means that reasoning from evidence is not primarily deductive. I include philosophy here, as well as science. One of the mistakes some philosophers make is to see philosophy as more like mathematics than science.

    Even mathematics did not ultimately originate from axioms. People were able to do arithmetic long before Peano came along. I would say that arithmetic is rooted in successful practice. It probably evolved from people finding that proto-arithmetical practices with numbers were useful for predicting and explaining their observations of the world, e.g. "I have one apple, and if I find another one I'll have two apples." At some point people abstracted this to "1+1=2", but they still didn't need axioms like Peano's. That said, I will happily call arithmetic an "axiomatic system", and I'm prepared to say that in some abstract sense it had axioms even before it was formally axiomatised. But if we take "axiom" in a more literal sense, referring to a statement or other symbolic representation, then I would deny that it had axioms. (In fact you do use the word "statement", so you seem to have in mind the interpretation I'm rejecting.)


  4. [...continued]

    For the pursuit of scientific research, we must assume the axiom that beliefs should be based on evidence...

    All we need is the habit of paying attention to evidence. We don't have to argue from a premise such as "Beliefs should be based on evidence". And in the broadest sense we cannot help but have such a habit. The evidence of our senses forces itself on us. Evolution has programmed us that way, because learning from evidence helped our evolutionary ancestors. If we're tempted to ignore the evidence in a particular case, we may be able to keep ourselves on the right track by reminding ourselves of the past success of attending to evidence in general.

    The is essentially the "problem of induction", where a mistake often made is to conflate induction in the broadest sense with induction in a narrower sense. We cannot justify induction in toto, because all our justifications depend on induction. But we also have no need to do so, since no one can possibly give up induction in toto. However, we can justify more specific inductions or classes of induction by invoking our broader tendency to think inductively. We can justify scientific induction with the argument, "It's worked in the past", because then we are appealing to the listener's broader (extra-scientific) tendency to induce.

    That said, I doubt that many people have been influenced by such an argument into thinking more scientifically. On the whole we don't think scientifically because we've been argued into doing so. We do so because our experience has caused us to develop more scientific cognitive habits (non-conscious as well as conscious). Good epistemic judgement is much more about developing good non-conscious cognitive habits than is generally appreciated. We overestimate the role that verbal reasoning plays in good judgement.

    1. Hi Richard,

      Thanks for the kind words, and I appreciate the feedback.

      All good points. However I still think it is useful to think in terms of axioms.

      Firstly, just to clarify that Harris doesn't himself use the word axiom (as far as I recall). I use it to describe the foundation upon which he builds his moral framework because it seems to me to be analogous to a mathematical axiom. It's a principle he regards as self-evident from which he derives all of his morality.

      Of course we can be very productive in reasoning and acting without being aware of any axioms. However, I think the axioms are still there, even if we haven't really identified them. If you look outside and see the rain, your conclusion that it is raining is only valid if we take as self-evident the unconscious assumptions that you can trust your senses and indeed your ability to reason, and these are axioms of a sort.

      Far from seeing the identification of axioms as a barrier to scrutiny, I see it as absolutely necessary for scrutiny. As long as we are reasoning and acting without considering our assumptions, those assumptions will not be challenged. By identifying our axioms - those assumptions we regard as so self-evident that we struggle to recognise that we are making them at all - we can then start to think about what might be true if these axioms were false and whether our confidence in them is justified.

      So, while Euclid's axioms might be considered a post hoc formalisation of geometrical knowledge that was already well established, the clear identification of these axioms opens up new possibilites as we consider what might follow if they were false, giving us various non-Euclidean geometries.

      That said, we might not disagree that strongly, as you acknowledge that axioms may exist abstractly even if they have not been formalised. My position is that this is so and it is useful to formally identify these axioms so as to open them to scrutiny.

  5. Hi Disagreeable,

    Thanks for your thoughtful reply. The point I'm making goes much further than just the question of axioms. It's quite a radically different view of epistemology, and leads towards similarly radical views on language and meaning, with significant implications for much of philosophy and AI. I don't expect to change your views so radically. But perhaps I can start the ball rolling by pointing out some flaws in the conventional view, and giving you a hint of another way of looking at things.

    Of course we can be very productive in reasoning and acting without being aware of any axioms. However, I think the axioms are still there, even if we haven't really identified them. If you look outside and see the rain, your conclusion that it is raining is only valid if we take as self-evident the unconscious assumptions that you can trust your senses and indeed your ability to reason, and these are axioms of a sort.

    This is similar to an earlier point. You said, "we must assume the axiom that beliefs should be based on evidence...". I responded that we don't need any such assumption. All we need is the habit of paying attention to the evidence.

    Similarly here, I don't need to assume my faculties are working properly. My faculties just get on and do their job, acquiring information, processing it, possibly bringing the results to conscious attention, perhaps even forming a verbal thought like "It's raining". It's simply a brute fact that, if my faculties weren't working well enough, I wouldn't form the correct belief (or I'd only form it by chance). Assumption doesn't come into it.

    You seem to be interpreting my belief (that it's raining) as the conclusion of an argument, so you're attributing to me assumptions to act as the premises of that argument. But why think I made an argument? By hypothesis I didn't make an argument consciously. And I for one would deny that we make unconscious arguments.

    What about an animal? When a dog notices that it's raining, has it made the unconscious assumption that it can trust its senses? It seems harder to say this of a dog, because dogs don't have language and so don't make arguments. Or consider an even simpler system, like a digital thermometer. Are its readings only valid if it made the unconscious assumption that it can trust its sensors and circuits?

    My point is that we don't need arguments and assumptions to achieve valid informational states. My belief that it's raining is valid just by virtue of the fact that it resulted from reliable (truth-conducive) causal processes. The reading on a thermometer is also valid by virtue of the fact that it results from reliable (truth-conducive) causal processes, albeit vastly simpler ones. In my case those truth-conducive processes may include the making of arguments, but they don't have to.

    If you take the view (as you seem to) that a belief can't be valid unless it's derived from another valid proposition, you run into the problem of infinite regress (and the problem of induction). Each valid proposition must be derived from another valid proposition, ad infinitum. But it's clear we don't have such an infinite chain of reasoning. One of the virtues of a more naturalized approach is that it dissolves the problems of induction and infinite regress.

  6. Hi Richard,

    That's good stuff, and I apologise if I glossed over your point about induction being natural, effective and not based on argument.

    I think we're talking about the same thing from two different points of view. You seem to be discussing how people (and animals) come to hold certain beliefs in practice. I agree with you that these beliefs often arise intuitively without argument or any explicit consideration of axioms.

    The problem with intuitive reasoning is that it's pretty fallible. If somebody expresses some opinion to you but has no argument to back it up, that's not very persuasive.

    So what I'm discussing is something else: how we analyse and justify beliefs, including assessing their truth. This is an activity that is not carried out by animals, or thermometers, or even very often by certain people. It is carried out when (rational) people enter into an argument or when a thinker deliberates on a particular proposition. Understanding the underlying unconscious assumptions that give rise to beliefs is an invaluable tool for this task.

    Because whether a viewpoint was arrived at intuitively or not, I regard it as being based on something equivalent to an argument (flawed or otherwise). I would say this is true even of a thermometer, which might give readings based on the "axiom" that mercury expands as it is heated. The thermometer is not making an argument, but if we want to assess whether it is capable of detecting the temperature, then we had better understand the principles on which it works, and these can be expressed and analysed as axioms and argument.

    So, to represent my view...

    The word "valid" as applied to belief is perhaps a tad ambiguous. One can believe true facts for bad reasons. Is this valid? I don't know.

    I prefer to think in terms of "justified" and "true" beliefs, according to Plato's definition of knowledge. So my view is that a belief cannot be justified unless it's derived from other justified propositions. Ultimately there is a regress, as you point out, but you can solve that with a few caveats.

    "I believe it is raining, because I can see that it is raining (assuming I can trust my eyes and basic intuitions about what I see)"

    This quickly becomes tedious, so we tend to drop the most common caveats from ordinary discourse, but I maintain that we should keep in the back of our minds that they are always implied.

    If we want to assess whether a particular proposition is correct or not, such as Sam Harris's statements about morality, it is useful to break it down and make these assumptions explicit - and in fairness to him he does this himself briefly before moving on and taking it as a given.

    So what I have essentially done is to again add the missing caveat to Harris's statements, and to point out that his basic assumptions are not as unimpeachable as he thinks they are.

  7. Hi Disagreeable,

    I think we're talking at cross-purposes. I'm talking about how the brain works, saying that our non-conscious cognitive processes can form rational beliefs without making any arguments. You're talking about the arguments we use to justify our beliefs.

    I think you may not have noticed that there are two senses of "justified". In its usual sense "justified" is an adjective, meaning something like "warranted". In this sense it can reasonably be interchanged with "rational". In another sense "justified" is the past participle of the verb "to justify". To justify a belief normally means to make a valid argument in support of that belief. So a belief can be justified (first sense) without having been justified (second sense), i.e. a belief can be warranted/rational without having been given a justificatory argument.

  8. Hi Richard,

    Ok, so we're agreed that beliefs are not usually arrived at by making an argument.

    I claim that such beliefs are only *warranted* if a convincing argument can be made to support them. Whether or not an argument was involved in the genesis of the belief, it should be possible to *justify* it with argument.

    Otherwise we're just intuiting, and that's highly unreliable.

    So, we all have an intuition about what's right or wrong. Is this belief warranted? I don't think so, because neither Sam Harris nor anyone else has managed to come up with a convincing justification. To make this point, it is necessary to critique his justification, which is why we need to think in terms of axioms.

  9. Hi Disagreeable,

    I was in the process of writing a P.S. before I saw your latest comment, so I'll post that now.

    P.S. I didn't fully address your comment. So let me add something more. You seem to be saying that, regardless of whether we need axioms to acquire rational beliefs, at least we need them when we come to skeptically scrutinise our beliefs.

    But people have been scrutinising their arithmetic results for centuries without Peano's axioms. And, if you doubt that 1+1=2, Peano's axioms won't help you. Why trust Peano's axioms any more than your general knowledge of and ability with arithmetic? It's our successful practice of arithmetic that gives us confidence in it, not Peano's axioms.

    "I believe it is raining, because I can see that it is raining (assuming I can trust my eyes and basic intuitions about what I see)"

    This quickly becomes tedious, so we tend to drop the most common caveats from ordinary discourse, but I maintain that we should keep in the back of our minds that they are always implied.

    Your point seems to be that justificatory arguments have a theoretical complete form, but in practice we omit part of the complete argument for the sake of brevity. I deny that there is any such complete form, particularly in the case of evidentiary arguments.

    You can't completely trust the evidence of your eyes and basic intuitions. So the premise should be, "I can mostly trust my eyes and basic intuitions about what I see". But then it's clearly not a valid deductive argument. Evidentiary arguments are not fully deductive, and therefore not complete. We must still rely on our intuition (non-conscious cognitive processes) to get from an incomplete set of premises to a conclusion. It's not the case that we omit premises from our arguments because it's too tedious to give them all. There is no complete set of premises to give. And given the problem of infinite regress, we have to leave premises unjustified (not supported by further argument). There can be no complete chain of premises either.

    Evidentiary arguments are better seen as matters of inference to the best explanation than as primarily a matter of deduction.

  10. Hi Disagreeable,

    I'm afraid I'm going to drop out of this discussion, as it's taking up rather a lot of time. My goal was only to give you some food for thought, which might take you in a new direction, not to debate the whole of epistemology from the ground up, which is what we seem to be doing now. If you make another comment, I'll be sure to read it, but I probably won't reply again.

    Thanks very much for your time and effort. Good luck with the blog. I'll continue reading it, and may comment again when I have time.

    1. Thanks Richard, you've certainly given me plenty to think about.

  11. Hi again,

    I've just been reading the following article on "Internalism and Externalism in Epistemology":

    I wasn't familiar with these terms before, but it seems you are an internalist with regard to epistemology while I am an externalist. That said, there are a wide variety of views within those categories, and I disagree with some of the specific views attributed to externalists in that article.

    FWIW I would add that I have a Wittgensteinian view of language, and this plays a significant part in much of my philosophy, including here. If you have time I recommend attending to the views of Wittgenstein. Unfortunately I don't have a particular book to recommend. The relevant book by Wittgenstein is "Philosophical Investigations", but it's very hard to understand. I doubt I would have understood it if I hadn't already been predisposed towards such views, and there was a still quite a bit that I didn't understand. There are lots of books about Wittgenstein's views, but I don't think one can be confident that the authors have understood him correctly!

    I'm hoping to read this book, if I can get if from my local library:

    I hope that's of some interest.

    1. Hi Richard,

      Thanks for that.

      But whether you're an internalist or an externalist, in this article I'm just saying that if you want to assess the truth of a proposition coming from an interlocutor, the most reliable way to do so is by analysing the argument. The epistemological questions we have gotten ourselves into are perhaps a bit beside the point.

      I'm not actually so sure I fit in either the internalist or externalist camp -- I'm perhaps a bit skeptical of the claims of either. I may have referred to "justified true belief" as a shortcut but I didn't mean to say that I particularly endorse this definition of knowledge.

      Crudely put, I don't think you can really have internally justified knowledge because your reason could be faulty, and I don't think you can have externally justified knowledge because your senses could be faulty.

      Fundamentally, I'm not sure I think that justified knowledge is really a thing at all. Perhaps there is only stuff that I think I justifiably know, and you have stuff that you think you justifiably know, and we can assess how justified we find each other's knowledge by discussing and thinking about our respective arguments.

      As far as I understand his views, I think I'm largely on board with Wittgenstein. There are a lot of philosophical arguments that only come down to subtle differences in interpretations of terms. Whether mathematical Platonism is true, for example, depends largely on how you interpret "existence" and so whether mathematical objects can be said to exist. Without having really considered it properly, the internalism/externalism debate seems to me to have something of this flavour. Neither camp seems to be making factual claims so much as simply disagreeing over what definitions of "knowledge" and "justification" to use.

      But, to tie this in to the original article, no matter what view you take on personal knowledge, I think it is true that formal disciplines such as science and mathematics take certain assumptions as givens, even if they do not do so explicitly. In mathematics we have axioms, and in science we assume that scientific laws are universal and unchanging. This article discusses the analogous assumptions being made by Harris, and I think that's a perfectly reasonable thing to do no matter whether you're an internalist or an externalist.