Buddhism and The Problem of Induction, A Quick(ish) Remark

Hume questions, among many other things, whether we can ever reasonably make inferences on anything other than what we’ve already experienced, with the conclusion that causality is something we infer even if all there is is conjunction, immediate association*. Casualty, for Hume, is a “perception of the mind,” (Deleuze, 2001, p. 26). 

You cannot therefore infer that an instance which you have never experienced before will resemble an instance that you have already experienced. 

As Henderson (2018) states, you cannot infer this by affirming what is  intuitively/demonstratively certain (called demonstrative reasoning) because you can always conceive of a possible scenario in which the future state simply does not meet the regularities you’ve come to expect (i.e. a stack of wood doesn’t light up for whatever reason, maybe because unbeknownst to you it is too humid).

You also cannot infer this by matter of fact (called moral reasoning) because that implies that the future will conform to the past which is just begging the question, as in, assuming what we are trying to conclude. 

There is thus a disconnect between Past and Future, with the limit of f(x) as x approaches both infinities being the Present. And even that, like an asymptote or empty point, is ever-fleeting.

Milan Kundera in his novel “The Unbearable Lightness of Being” says something similar to Hume’s position. Man’s life is a straight line, not a circle. Thus, we can never be sure what is ultimately right or wrong because we cannot go back and test different scenarios. A decision is a decision. When you elope with someone, you will never know how it would’ve been like not to run off with them, and even if you’ve been in similar relationships before that never worked out, you still can’t be sure this one won’t be entirely different. This is the reason why we make the same mistakes over and over again. It’s not because we are ignorant, it’s because nobody knows any better.

And who is to judge us for believing? Hoping?

This might be exactly why decisions are so scary to make: they’re irrevocable. But it also might come across as relieving. You can never be sure what is ultimately right or wrong, so why worry? It also helps to avoid the mutual suffering that emerges from blaming. Your ex, like you, also could not have been sure that it wasn’t going to work, so you can’t always get mad at them for ‘not telling you sooner’. They are humans too.

On the other hand, there is Buddhism, a way of life whose main proposition is also that of focusing on the present because the answers we seek for lie solely in the here and now**. Among the 5 hindrances to enlightenment, there is Sensual Desire, Restlessness and Skeptical Doubt, all three of which involve detrimental ruminations of the past and future as do the other two, Aversion and Dullness, in more subtle ways. 

I see both a similarity and a disparity here. One the one hand, they are both referring to the same thing – that the present is where it’s at – but while Hume terms it a ‘problem’, Buddhists view it as the highest of truths***.

I send my dearest thanks to a good friend of mine, guru in all things Buddhist and much more, for pointing out the following to me. These revisions are guided by his kind feedback and add incredible substance to my entry. In fact, the revisions are double if not triple the size of the original post itself! It makes me realize that without the sections below, my concluding contrast between Hume’s philosophy and Buddhist philosophy would have been cutting the story too short, almost lying by way of omission. Close call. Phew.

*Revision 1:

Thinkers since the time of Hume have contemplated how to reconcile association and causation, without implying that the latter is simply an imagined concept.

Judea Pearl (2019), for example, has argued for association as one level in the causal hierarchy, the lowest in fact because questions at that level can be answered without any causal information. Thus, association (correlation) does not imply causation.

In this sense, Hume would agree.

The following relates to my later point in revision 2 of the importance of both past and future considerations in understanding causality.

According to Pearl (2019), the causal hierarchy has three levels: 1. Association (What is? – a view from the present, based on observable data and statistical relationships), 2. Intervention (What if? – a view into the future) & 3. Counterfactuals (Why? – a view into the past). You can answer questions at level i (i = 1,2,3) if you have information from j, where ji (Pearl, 2019). Thus, to answer 1, you need information from 1, 2 or 3. This hierarchy is directional, meaning that if you have info from 1, you can’t answer 2 nor 3, but if you have info from 3, you can answer both 2 and 1.

Say I want to know (1. Association:) what my stomach rumbling tells me about whether I’m hungry and need food. I can answer that by knowing (2. Intervention:) what NOT eating food will do to my stomach or (3. Counterfactuals:) whether it was NOT eating food that made my stomach rumble.

With this example in mind, you can test the directionality of this model: if 3. I know that it was not eating food that (likely) made my stomach rumble, then 2. I know that in the future, not eating food will (likely) make my stomach rumble, meaning that 1. I also know that my stomach rumbling is (likely) a sign I haven’t eaten food.

In this sense, Hume would disagree.

To explain my use of “likely” in brackets, here is the causal hierarchy in terms of probabilities as given by Pearl (2019) in relation to my example:

  1. Association: P(y|x) = p → The probability that my stomach rumbles (event Y=y) given that I did not eat (event X = x). The least amount of information I need to understand this is basic statistical information of observable data, which is quite Bayesian from what I understand due to its appeal to evidence and new information. Based on Bayesian inference, p = ( P(x|y)*P(y) ) / P(x) = P(x and y)/P(x). That is, {the probability that I did not eat given that my stomach rumbles and that my stomach rumbles} divided by {the probability that I did not eat}.  

A quick detour: I like to mention Bayesian statistics here because it can help explain why we tend to (wrongly, as deemed by Hume) form the belief of causation from repeated correlation (although the explanation still falls short).

For Hume, we innately believe that nature will be reliably uniform, and we believe this because of a principle of human essence, a habit (De Pierris & Friedman, 2018) even if it is not rational in his opinion. So let’s say our ‘prior belief’ in uniformity is any number other than 0. The more evidence for a correlation between A and B we get, the more that prior is enhanced into a deeply engrained posterior belief in uniformity. Thanks to Bayesian statistics, however, we know that these beliefs are not just plain habit but a habit that is sustained by probability and evidence. We think the world will be sufficiently uniform because it is mostly uniform.

Sure, Hume, everything is possible, even that the future will not conform to the present. But it’s a matter of how probable, and how probable is the guiding principle of the universe. At the quantum level, there is a chance that electrons appear on the other side of a wall, but the probability is too minute to make a difference. Thus, acting on probabilities is quite rational.

The hypothesis, then, is that once our trust in uniformity is maximized, the bridge is built for us to cross over into the territory of causation. According to Pearl (2019), probability and causation are connected (by something called d-separation, fun fact) but how? Maybe the only way to explain uniformity is that one thing has to cause the other, although I’m not sure where this propensity comes from. There is a bridge, yes, but the final plank is missing. Bayesian statistics and Hume are quite silent in this regard. Admittedly, that silence is justifiable; as I said, Pearl’s level 1 of association is pretty much Bayesian probability in action, and association does not require casual information, which means that you cannot conclude causation from it. And this, too, was Hume’s original conundrum: how do we make the leap from mere correlation to causation?

Overall, it doesn’t seem like causation is an imperative conclusion to derive from above-chance uniformity. We might as well have believed in harmony and reciprocity. But maybe the tendency towards hierarchy and categorization is too strong for us to dispel (#shade). Dennett (2017) suggests there is contemporary research showing that causation is actually a reflex but he stops there (a great read btw, especially if you’re interested in how Hume’s claims around causation help elucidate our understanding of what consciousness is).  

But what is nice about Bayesian statistics is that, despite all the evidence for the correlation that is undoubtedly obvious to Hume as well, it helps to explain why Hume might still not have eventually come to concede the existence of causation like the rest of us. If his initial prior belief of the probability of uniformity was 0 or close to it (in the sense that, although he saw the future conform closely to the past, he didn’t take that as a ‘must-be’), then even a lot of evidence will churn out a very low probability for uniformity, and eventually a posterior belief against causation.

This appeal to the connection between priors and posteriors might help explain a whole load of other phenomena in life: why we might feel lonely despite being surrounded by endless company, unconvinced despite being shown so much evidence, unloved despite being given so much compassion. No wonder predictive processing fits nicely with the Bayesian doctrine! It’s a matter of how much we weigh our models vs data. 

I like to think that the weighing is something we can (mostly) control, if nothing else.

But, anyways, back to paved roads. Finishing off Pearl’s (2019) causal hierarchy in terms of probabilities:

  1. Intervention: P(y|do(x), z) → The probability that my stomach rumbles (event Y=y) given that I make myself not eat (this is an intervention, purposely setting the value of event X to x instead of having it occur naturally) and subsequently observe event Z = z.
  2. Counterfactuals: P(yx|x’, y’) → The probability that my stomach would have rumbled had I not eaten (yx) given that I actually did eat (X = x’) and my stomach did not rumble (Y = y’). 

**Revision 2:

My original post is not a complete story in itself. I realize now that for Buddhism, the answers do lie in the here and now but in more of an indirect sense than I originally conveyed. 

Firstly, you wouldn’t be able to gain insight from focusing on the present without paying attention to the obstacles that hinder your ability to stay centered in the sensations and states of the now in the first place.

When you begin meditating, you realize that the practice of meditation is less about suppressing the past and the future for the sake of the present than it is about becoming aware of what impedes you from staying present. This is a common misunderstanding, and a potentially very detrimental one. Meditation is about achieving awareness of your distractions, getting faster and more efficient at it each time, followed by acceptance and non-identification. It is this non-identifying that encompasses the act of letting go and re-focusing on the present, but you have to notice that it is only the very last step in the training process, and it has nothing to do with denial. In fact, to achieve this awareness, you have to look at the past and future and ruminations thereof straight in the face (with due moderation, for the sake of avoiding even greater trauma than you started with).

If you ignore the past and future, life looks like a dot wandering idly through space and you don’t have the faculty to notice the sinusoidal curve that it draws out in time. It is with a focus on these ups and downs of life, with this existential curve including all of its causes and effects as a backdrop, that being present acquires any inherent value in edging you closer to enlightenment.  

In this regard, my aforementioned friend sent me a story about the Buddha speaking to his son Rahula, Ambalatthika-rahulovada Sutta: Instructions to Rahula at Mango Stone, with the following passage that is so educative in its very repetitiveness:

Whenever you want to do a physical/verbal/mental action, you should reflect on it: ‘This mental action I want to do — would it lead to self-affliction, to the affliction of others, or to both? Would it be an unskillful mental action, with painful consequences, painful results?’ If, on reflection, you know that it would lead to self-affliction, to the affliction of others, or to both; it would be an unskillful mental action with painful consequences, painful results, then any mental action of that sort is absolutely unfit for you to do. But if on reflection you know that it would not cause affliction… it would be a skillful mental action with pleasant consequences, pleasant results, then any mental action of that sort is fit for you to do.

While you are doing a physical/verbal/mental action, you should reflect on it: ‘This mental action I am doing — is it leading to self-affliction, to the affliction of others, or to both? Is it an unskillful mental action, with painful consequences, painful results?’ If, on reflection, you know that it is leading to self-affliction, to the affliction of others, or to both… you should give it up. But if on reflection you know that it is not… you may continue with it.

Having done a physical/verbal/mental action, you should reflect on it: ‘This mental action I have done — did it lead to self-affliction, to the affliction of others, or to both? Was it an unskillful mental action, with painful consequences, painful results?’ If, on reflection, you know that it led to self-affliction, to the affliction of others, or to both; it was an unskillful mental action with painful consequences, painful results, then you should feel distressed, ashamed, & disgusted with it. Feeling distressed, ashamed, & disgusted with it, you should exercise restraint in the future. But if on reflection you know that it did not lead to affliction… it was a skillful mental action with pleasant consequences, pleasant results, then you should stay mentally refreshed & joyful, training day & night in skillful mental qualities.

Rahula, all those brahmans & contemplatives in the course of the past who purified their bodily actions, verbal actions, & mental actions, did it through repeated reflection on their bodily actions, verbal actions, & mental actions in just this way.

All those brahmans & contemplatives in the course of the future who will purify their bodily actions, verbal actions, & mental actions, will do it through repeated reflection on their bodily actions, verbal actions, & mental actions in just this way.

All those brahmans & contemplatives at present who purify their bodily actions, verbal actions, & mental actions, do it through repeated reflection on their bodily actions, verbal actions, & mental actions in just this way.”

Furthermore, unlike Hume, Buddhism leans heavily on the notion of causation. Life, for Buddhists, is a lot more circular than straight; actions have intentions and consequences (karma), which in turn has implications on our cycle of rebirth (samsara). X causes Y, i.e. suffering/unsatisfactoriness, and thus X should be ‘cessated’.

Here’s where I’m confronted with a gap, however:

X → Y. (If X then Y)

¬ Y. (Not Y)

¬ X.  (Therefore, not X)

This is simple modus tollens. 

Nonetheless, ¬ X does not in itself entail ¬ Y. Let’s say that X is the act of craving and you stop X. It doesn’t necessarily follow that you stop suffering…

Reading Coseru (2017), however, it seems rather as if the equation should be reversed: 

Y → X.

Unsatisfactoriness, Y, leads to X: aversion (of unwanted states) and grasping (of wanted states). And what actually causes Y is ignorance (avidyā), call it Z. Ignorance is “understood not simply as lacking knowledge about particular states of affairs, but rather as a basic misunderstanding about how things truly are,” (Coseru, 2017).

Thus Z → Y ∧ Y → X

¬ X (no more grasping/craving and no more aversion) entails ¬ Y (no more suffering) and if you no longer suffer, then you are no longer ignorant, ¬ Z. 

Looked at in this way, it makes more sense why ¬ Z itself does not entail ¬ Y. Even if you are knowledgeable about how things are in the world, you might still engage in behavior that causes suffering. So you have to stop that behavior first if you wish to stop suffering; it then follows that you are also no longer ignorant.  

In addition, causation is important in the Buddhist Principle of Dependent Arising (pratītyasamutpāda). Here is a quick overview paraphrasing from Coseru (2017):

What is the Buddhist explanation for how thoughts, even metathoughts, are connected with actions? The answer is that everything, even cognitive events, emerge from a dependence between a myriad of causes (nidana) and conditions. This is neither a top-down nor bottom-up approach to cognition, i.e. thoughts don’t come before actions, and therefore don’t necessarily cause them, but thoughts (mental events) also don’t lack causal force in the sense that they aren’t simply caused by actions (physical events) without being able themselves to cause actions too (that would be epiphenomenalism). Thus everything is ‘transient’ (impermanent) and interdependent. 

I like to think of this as the causal loops being bi-directional.

***Interesting side-note:

Hume had some ideas that were quite similar to those of Buddhist philosophy however, especially with regards to personal identity. Hume said we are “bundles of perception”, i.e. just a conglomeration of ever changing mental states or events, (Olson, 2019) which is very much an aggregated view in line with the Buddhist not-self doctrine and Five Aggregates. Though while the Buddha might say those bundles are not “ours”, I wonder if Hume would agree.


References

Coseru, C. (2017). Mind in Indian Buddhist philosophy. In Edward N. Zalta

(ed.), The Stanford Encyclopedia of Philosophy (Spring 2017

Edition). Metaphysics Research Lab, Stanford University.

https://plato.stanford.edu/archives/spr2017/entries/mind-

indian-buddhism/.

Deleuze, G. (2001). Empiricism and subjectivity: A series in social thought &

cultural criticism (revised edition). Columbia University

Press.

Dennett, C. D. (2017). From bacteria to Bach and back: The evolution of

minds (p.147). W. W. Norton & Company.

De Pierris, G., & Friedman, M. (2018). Kant and Hume on causality. In

Edward N. Zalta (ed.), The Stanford Encyclopedia of

Philosophy (Winter 2018 Edition). Metaphysics Research Lab.

https://plato.stanford.edu/archives/win2018/entries/kant-

hume-causality/.

Henderson, L. (2018). The problem of induction. In Edward N. Zalta

(ed.), The Stanford Encyclopedia of Philosophy (Spring 2020

Edition). Metaphysics Research Lab, Stanford University.

https://plato.stanford.edu/archives/spr2020/entries/induction

-problem/

Olson, E. T. (2019). Personal identity. In Edward N. Zalta (ed.), The Stanford

Encyclopedia of Philosophy (Fall 2019 Edition). Metaphysics

Research Lab, Stanford University.

https://plato.stanford.edu/archives/fall2019/entries/identity-

personal/.

Pearl, J. (2019). The seven tools of causal inference, with reflections on

machine learning. Communications of the ACM, 62(3), p. 54–

60. ACM New York. https://doi.org/10.1145/3241036

 

Leave a Reply

Your email address will not be published. Required fields are marked *