The Inference to the Best Explanation | Gilbert Harman
The Philosophical Review, Vol. 74, No. 1 1965
Inference to the best explanation (IBE): infer the truth of a hypothesis because it provides a "better" (whatever that means) explanation for the evidence than other hypotheses.
Main claim: IBE is the basic form of nondeductive inference, not enumerative induction.
1) If enumerative induction is warranted : you can also call it "inference to the best explanation" 2) If you cannot, it is probably not warranted and thus not an inference at all.
So Harman basically claims that only warranted inductive claims are special cases of IBE.
If we have enumerative induction, we still need IBE. But once we have IBE, we can see enumerative induction as a special case of IBE.
Cases where induction falls short where IBE does not
P.89-90: a detective collects pieces of evidence and concludes: the butler did it. This is IBE and cannot easily be accounted for by enumerative induction.
Note that Harman here does not proof these examples cannot be accounted for using enumerative induction. He leaves open the option that you may come up with a very complex construct where it may be constructed as a case of enumerative induction. And yet, he is certain that these attempts would fail.
Another example of IBE is testimony: "our confidence in his testimony is based on our conclusion about the most plausible explanation for that testimony" (89)
Q: Is Harman here specifically interpreting IBE as inferring the best hypothesis, or does he include the discovery of hypotheses as well? (Abduction "proper" if you will).
Inductive logic does not indicate when induction is warranted
From a logical point of view, it is unclear when inductive inference is warranted. If you look to the logic of induction, you keep the problem that the next instance may always contradict the inferred generality.
IBE can however indicate when inductive inference is warranted
In IBE, it is clear when we are warranted, even in such a similar scenario. At each moment, we can compare and contrast two hypothesis: is the hypothesis that the next instance will be the same more plausible than the one who says it will not be?
The answer is that one is warranted in making this inference whenever the hypothesis that all A's are B's is (in light of all the evidence) a better, simpler, more plausible (and so forth) hypothesis than is the hypothesis, say, that someone is biasing the observed sample in order to make us think that all A's are B's. (91)
Note that Harman treats the actual comparison of hypotheses as a black box here.
Remember Harman's claim:
either (a) enumerative induction is not always warranted or (b) enumerative induction is always warranted but is an uninteresting special case of the more general inference to the best explanation. (88)
At p.90 Harman asks when is an inductive inference warranted ?
For example (not in paper), the typical induction on white swans is not warranted because I know it's unlikely that all animals have the exact same colour. The induction that the sun will rise tomorrow is however warranted (weak enumeration with a lot of total evidence). To say that is also will rise, not (strong enumeration, still does not weigh up against evidence).
Now the main point is that "either a claim of enumerative induction is unwarranted, or it is a special case of IBE".
Enumerative induction is unwarranted "if there is no reason to rule out a change" (91).
When it is warranted, we can represent the enumerative induction from above as IBE. Either "for all A -> B" is a better hypothesis than the alternative, or we discuss the hypothesis "the next A is of the same type".
If it is not warranted, IBE can express why it is not warranted.
Truth-inference
Let's zoom in on this sentence:
In making this inference one infers, from the fact that a certain hypothesis would explain the evidence, to the truth of that hypothesis (TODO welk pagina nummer?)
From TFCA-2-Abduction.md , Johan Kwisthout gave the following example:
We see a picture of woman having a rose at the ear:
1) Hypothesis 1: she is listening to a rose * Explains picture perfectly! But a priori very unlikely 2) Hypothesis 2: she is part of experiment (my own) * More likely but does not explain this specific picture * I can't remember Kwisthout's own suggestion.
But the point is clear: an explanation can explain the evidence perfectly and yet be complete shit! You can come up with all kinds of 'magical' reasons. Another way of saying this, is that Harman does not take the prior of the hypothesis into account.
Part 2: Knowledge and IBE
Canonical definition: "knowledge is justified true belief".
Main claim: IBE exposes used lemma's, which are relevant in the analysis and mainly justification of knowledge.
Important insight: trueness is not enough for knowledge. Knowledge must also be justified.
Authority arguments can be ways of justification and included in IBE. E.g. the square of 2 is irrational because Euclid has already proven it in 300 BC and it has not been contested since. This is a form of IBE, because it being true is the best explanation ('warrant') of it not being contested.
Lemma's play a role in IBE.
E.g. the inference of truthful testimony contains the lemma that the utterance is believed and is not a slip of the tongue.
E.g. someone pulling away a hand after touching a hot stove included the lemma: the hand hurts (inferred mental experience).
These lemmas are used in IBE. In the last example, the best explanation is that the hand is withdrawn because it hurts.
These intermediate lemma's are all the result of inference and aid in studying our inference-based knowledge. This is an additional argument why IBE is fruitful.