Note: This comment system was replaced with Disqus in May 2010.
Saturday, October 5, 2002 at 7:32:16 mdt
In "Induction: A Problem Solved,"
(2002), David Miller argues for a throughly anti-justificationist, i.e., Popperian, view of induction. In this way, messy if not outright dubious lists are dispensed with as unnecessary clutter to the greater problem of ratiocination.
In other words, why have "lists" in the first place if they are dispensible or else wrongly or otherwise overly restrictive? Instead they are seen as deadweight or redundant.
"[U]niversal laws do pass judgement on their own future performance, and if we classify them as true, so do we."
So--where is the validity of induction? If not, then there is no parsing of whether or not geology or history--the historical sciences--or astronomy are genuine "sciences" because there is no difficulty encountered to begin with. This makes greater sense to me.
Tuesday, October 8, 2002 at 8:41:08 mdt
Name: Robert Campbell
What comes through in the summary of Peikoff's new lectures on induction is familiar from his coverage of induction and scientific method in his old course on logic. But, as always, there has been peculiar hardening on some issues: saying that induction is concept-formation, for instance.
On your objection #2, Peikoff seems to be appealing to the idea that there is an underlying causal explanation for a generalization that isn't causal on the face of it. Maybe we don't know why zebras have black and white stripes, but if we knew enough about zebra DNA we'd be able to explain why. The idea that "lawful" generalizations involve a causal "promissory note" is in Harre and (middle period) Putnam, among others. It even figures in some developmental studies of generalizations that children are willing to make (Frank Keil, Susan Gelman).
Objection 3: I recall Peikoff using "experiment" rather loosely by the standards of my trade--for instance, he said that retitling one printing of a Victor Hugo play, to see whether that would affect sales, was an experiment, even though it would not satisfy a strict scientific definition of one. He seems to have allowed "experiment" to balloon to cover all kinds of empirical studies that use Mill's methods (which he referred to in the old course). I agree with you that there's nothing to be gained by doing this.
The qualification that Peikoff puts on Ockham's Razor does seem to be new--and I agree with him on it.
Hypothetico-deductive method. Note that in setting the bar so high for concluding that a hypothesis is true (not only must the existing evidence support your hypothesis, but you also need to be able to rule out every possible alternative hypothesis). Rand does qualify her statements, by limiting the alternatives to those that the investigator--or others in the scientific community--currently know about, but notice how Rand and Peikoff are much closer to Popper here than standard presentations would lead you to assume.
As for uses of math in accounts of how we arrive at (and test) generalizations, I agree that cardinal measurement is pretty much out. (My recent post to OWL on Pollock's use of "expected value" theory gets into some of that.) There might be other mathematical tools that are useful, however--whether they use variants of automata theory, or state and phase spaces (connectionist models incorporate these), or topologies, etc.
I hope that Peikoff's lectures get into the public record soon. If hearing them requires reciting the password at a guarded entrance to the ARI citadel, they're going to remain inaccessible to people like me!
Monday, March 3, 2003 at 10:54:55 mst
Name: Glenn Fletcher
I just reread your summary of
Peikoff's lectures on Induction.
I, too, am awaiting the release
of the tapes. I thought that you
might be interested in a webpage
that contains a summary of the
lectures. The URL is:
There is no author listed, but it
looks like a legitimate set of
Monday, April 17, 2006 at 12:16:28 mdt
I am not an intellectual of letters, as I am an animator by trade, but I write about philosophy regularly and discuss it regularly with my fiance and friends. I always enjoyed your observations, and I am glad that you are now enjoying ARI's approach to the dissemination and cultural application of Rand's ideas.
My special interest in philosophy is induction. I wanted to possibly help you amend a view that you formed after Peikoff gave his Induction and Method lecture. I realize that you were working only from notes, and that the CD's that were released much later in his "Induction in Physics And Philosophy" package include another lecture which was originally called "The Axioms of Induction." However, ignoring that, I would like to help direct you to Peikoff's actual view on the subject.
On Tuesday, October 01, 2002, you wrote:
"My second objection to Peikoff's basic account of induction concerned his claim that all generalizations (i.e. all inductions) are causal statements because they connect an entity with its modes of action. But not all inductions concern action; some concern attributes or relationships. For example, to say that all zebras have black and white stripes is a claim about the identity of zebras, not about the actions of zebras. Of course, we might say that the entity is "acting" in that those stripes are created by biological processes grounded in the DNA of zebras or whatnot. But this seems to stretch the notion of action and causality beyond usefulness because the generalization concerns the existence of the stripes, not their origin. In short, I can see no valid reason to reduce all generalizations about an entity's attributes or relations to generalizations about the actions of that entity."
Peikoff actually gives generalization THIS definition at the beginning of his Induction lectures:
"A generalization is a proposition that ascribes a characteristic to every member of an unlimited class however it is positioned in space and time."
He goes on to say that generalization is described by the statement "All S is P, where S is the subject and P is the characteristic being ascribed to it."
This definition would seem to satisfy BOTH the issue of causal statements, as well as statements of characteristics. Characteristic could be construed as a behavior (hence a chain of events related in terms of causes and effects), or it could be construed as a quality (such as with the stripe, or an attitude, such as the 'aggro level' or 'aggressiveness' of a particular breed of dogâ€" a gen that breeders have used for centuries to create and suit particular canine breeds to men's purposes).
In a major way, your zebra stripe idea is not wrong if you consider it as a statement of quality explicitly, and a statement of causality implicitly. Or, if you ratherâ€" it is a quality in one context, and if arrived at in another context, it would be causality. I don't believe I am equivocating here. It is common to consider a group of conceptualized causal chains as the content of a conceptualized quality... such as considering a "dishonest" person (a qualitative statement reached by induction) or an "aggressive" breed of dog (a qualitative statement reached by induction). However the CONTENT of these claims presupposes a causal chain. To make these claims, we see several events where a man repeatedly and knowingly chooses to steal, or lies to cover up his short-comings or failures, or cheats on his wife. We induce that such a man "dishonest" based on this information in a given context. As a result, we can make CAUSAL predictions based on this QUALITY. (Please forgive the all caps!) The same logical chain can easily be made with the â€śaggressive dog breedâ€ť example.
Qualities of entities cannot be sundered from their causal possibilities and vice-versa. This is even true and implicit in perception. However, I consider this nugget too much of a jewel to disseminate here. If you find it at all useful, please let me know!
A possible solution to your Zebra/DNA situation
All zebras, when they are conceived, carried to term and birthed, by their nature, must result in such and such a style of coloring (barring genetic alteration within a certain range). That means, a causal chain can be inferred or ruled out based on the presence of a stripe on a zebra's offspring (that of no mutation or alteration of certain genes). I don't think I need to assert explain the use of context to delimit this here. For example, multiple types of mutation may be possible to create a healthy, yet stripeless zebra progeny, but even these can be identified and distinguished contextually (diseases in utero, parental hereditary diseases etc.)--- and necessarily they would show other signs of existence apart from the stripe itself.
I hope this was clear and didnâ€™t ramble too much. I hope you weigh the idea slowly and carefully, but please, let me know if I missed the point altogether.
Thanks, and keep up the good fight.