Despite appearances, this post isn’t intended as a political post. I simply utilized the example that led to the thought that inspired the post.
We are in one of the most interesting political elections ever. We have a sitting president that is legally contesting his loss in the election, a near propaganda-like News station (Fox) that is getting attacked by said president for not ‘supporting him’ enough — even while they bring on and interview all manner of people claiming the election was a fraud. The whole situation comes across as comical to me. Yet I know many others (on the right and the left) that think this is a deadly serious situation.
So I had to ask myself a hard question that is related to the question of what induction really is. Why do I see it as comical rather than deadly serious?
My reason is quite simple: based on his past behavior, I believe Trump is a narcissistic individual that will gladly lie (or delude himself) to protect his ego.
It’s not like we don’t have a long history of Trump behaving in this way in the past and playing very fast and loose with the truth when it benefits him. Therefore, I do not take his claims of wide-scale voter fraud at all seriously unless he can produce evidence first. My default belief is that nothing is going on but that Trump is a sore loser. I’ll change my mind if he ever produces actual evidence for me to consider.
However, I have no doubt that if the claims of voter fraud came from someone else, say Obama or Bush, that I’d feel very differently. It’s not like Obama or Bush were always strictly honest. They weren’t. But certainly, they didn’t stretch the truth to its breaking point nearly as often as Trump has.
So my view is based on me generalizing from past observations about Trump. Did I just use induction?
It seems to be widely agreed upon, even among Critical Rationalists, that ‘induction’ is the process of generalizing from past observations. It’s just that Critical Rationalists will tell you that it’s a refuted process that doesn’t exist.
So did I generalize from past observations? It certainly seems to me that I did!
So if ‘induction’ is the ‘process of generalizing from past observations’ then it would appear I just used induction and that Induction does exist.
I’ve seen people insist that ‘induction’ is a claim to certainty. But I’ve made no such claim. I’m even leaving the door open to re-evaluate the situation if Trump should ever actually produce real evidence of his claims. Does me not claiming certainty make it not induction? The problem is that to most people, ‘induction’ does not require a claim to certainty. Only that repeated observations made me subjectively feel more certain that I had the right generalization. Which, in this case, is exactly true given that I’d feel differently if the voter fraud claims came from Bush or Obama.
Another objection I’ve heard is that I’m utilizing background theories. In this example, perhaps a background theory is that I’m assuming the existence of character in the first place. Does my use of a background theory make it not induction? The problem is, again, that to most people ‘induction’ does not include a claim that there is no background theories. In fact, modern induction — such as machine learning — fully realizes the importance of such background knowledge. They call such background theories ‘inductive biases.’ To someone using ‘induction’ in machine learning, the key thing that makes it induction isn’t that there is no background theories, but only that we’re generalizing from repeated observations. (Which I note, actually does work in real-life machine learning.)
The Critical Rationalist Explanation of My View of Trump
A thoughtful Critical Rationalist will have the right answer here. They’ll say, “You actually used Critical Rationalism to come to your conclusions. You’ve observed Trump’s behavior in the past and you found you couldn’t explain his behavior without conjecturing a theory about his character, namely that he’ll gladly lie (or be delusional) to protect his ego.”
And this is correct! My view of Trump is actually entirely consistent with Critical Rationalism!
We are constantly creating theories about other people and what their motives and moral limits are. These theories are what we call ‘character.’ I perceive (based on his past actions) Trump’s character as being dishonest and narcissistic whereas I didn’t, in general, perceive Obama or Bush that way. Therefore I perceive Trump’s current claims of voter fraud as being best explained (given the current lack of evidence) as merely Trump behaving badly. If Obama or Bush made such a claim and it was entirely false, that would have shocked me. That’s why I’d give their word more credence prior to them producing evidence. (Though it would always come down to the evidence in the end.)
But this brings us to an interesting point. In fact, my conclusion is both ‘inductive’ (i.e. generalizing from past observations) and Critical Rationalist (i.e. a theory about Trump’s character conjectured to explain his past actions.) So apparently the two theories aren’t mutually exclusive.
Is Induction “Refuted”?
But how can this be? Didn’t Popper refute induction?
The problem with that question is that it misunderstands the nature of refutation. Refutations, in the Popperian sense, don’t always mean that we proved the theory to be entirely without verisimilitude. To use an analogy, Induction is like Newtonian physics while Critical Rationalism is like General Relativity. There are no cases of Induction being correct that aren’t also examples of Critical Rationalism being correct.
Moreover, the Critical Rationalist explanation of why it was correct is the better explanation! Go back and look at both of the offered explanations. Merely saying I ‘generalized from past observations’ tells me very little.
Critical Rationalism Is A Deeper Explanation Than Induction
By comparison, the Critical Rationalist explanation is quite deep. It explains why past observations mattered (i.e. they were phenomena needing explanation) and moreover, CR explained how I generalized (i.e. I conjectured a theory about Trump’s character to explain his actions that I can then use to explain his current and even future actions.)
Consider the cases where Critical Rationalism is correct while Induction is incorrect. Let’s say I only once see Trump do something dishonest. I may still need to explain his behavior and might still do that by hypothesizing about his character. So there is no need for repeated observations like Induction claims, at least in some cases. However, Induction isn’t entirely wrong here either and Critical Rationalism explains why: I’m more in need of an explanation about someone’s overall character if the person does multiple dishonest things than if they do only one. Why? Because I know everyone has a bad day or makes occasional mistakes, but no one has constant bad days. This is why I subjectively feel more certain about Trump’s character the more times I see Trump behave dishonestly. It’s not that repeated observations equate to certainty, it’s that I have a harder time explaining his actions as not related to his character as the number of observations I need to explain increases.
Induction is Really Critical Rationalism but Critical Rationalism Isn’t Necessary Induction
In other words, Induction does exist and really works: it’s called Critical Rationalism. Whether you choose to call this ‘refuting’ Induction or ‘discovering the mechanism by which Induction actually works’ is just a matter of how we happen to define things and that is likely why Popper didn’t care if we called his theory “inductive” or not. (See footnote 1 of this post.)
However, if Induction is always Critical Rationalism, the reverse is not true. Critical Rationalism works in cases where Induction can’t — such as the aforementioned conjectures based on a single observation. Critical Rationalism is therefore a more general theory. That is why it can be used both for ‘generalization’ as well as finding explanations for single observations.
Has Critical Rationalism Subsumed Induction?
I conjecture that there are no cases where Induction is superior to Critical Rationalism but many cases where Critical Rationalism is superior to Induction. If that is correct, then the true sense in which Induction has been refuted is that Critical Rationalism subsumed it. Moreover, this is now a testable theory. All an Inductivist would need to do to disprove this theory is to show an example of Induction working where Critical Rationalism can’t.
So why do Critical Rationalists say Induction doesn’t exist? Typically, what they mean, is that it is impossible to generalize from observations without theory coming first and that you don’t need repeated observations to make a conjecture. And this is consistent with the above discussion. I conjecture Induction only exists in so far as Induction can be thought of as an approximation of Critical Rationalism not unlike how Newtonian Physics is only true in so far as it approximates General Relativity.