|
Author
|
Topic: Deduction and induction in intelligence: which is primary?
|
Leonid Andreev
Member
Member # 282
|
posted 08. June 2002 04:19
Knowledge is a self-sufficient system whose immanent part is intelligence, i.e. the ability to apply knowledge. Knowing how to apply knowledge is not only knowing the rules and techniques that a human being discovers through personal experience, takes granted from the human civilization experience, and reviews and consolidates throughout a life. Along with deduction (of effects following from basic principles), the ability to apply knowledge implies induction (of laws and principles following from observations). Induction abilities start developing in a human mind earlier than the deduction skills and are perfected along with those and are to a great extent determined by genetic peculiarities of a given individual, hence are inherited. Even at a minimum level of genetically imbedded “knowledge”, if we descend to primary forms of life, such as bacteria and protozoa, we observe the abilities for differentiation and integration. They are manifested in various types of taxis (chemotaxis, phototaxis, and thermotaxis) when a cell chooses a direction of its gradient motion that is best for survival. No doubt that an inference of inductive logic in unicellular organisms would be an egregious absurdity. However, believing that knowledge comes with axioms and theorems, and not with the mechanisms for obtaining knowledge when there is no prior basic knowledge, is naivety, to say the least. Therefore, it should be logical to suggest that to create the AI capable for deductive thinking, i.e. for evaluation of choices based on the matrix of knowledge accumulated by the mankind, we need to thoroughly investigate the mechanism of induction and to design such an implementation of AI that would display intelligence of its own.
Would you agree with my point or do you think otherwise?
IP: Logged
|
|
James A. Barham
Member
Member # 50
|
posted 08. June 2002 08:55
Leonid:
Yes, I think that most naturalistically inclined philosophers, not to mention cognitive scientists, would certainly agree with you. There is a widespread appreciation of the fact that "induction" in some broad sense is biologically deeper than, and indeed logically prior to, deduction.
I am not real familiar with this literature (it is enormous), but as I understand it, there are two basic approaches to modeling "induction." One is via probability theory (especially based on Bayes's Theorem) and the other is via nonlinear dynamics (physical constraint satisfaction). The former is induction properly speaking, while the latter is widely viewed as a physical correlate of the notion of "abduction" first introduced by Charles Peirce. Abduction is nowadays usually referred to as "Inference to the Best Explanation" (IBE) (see Paul Thagard's book, "Coherence in Thought and Action" for a discussion of these issues).
I not only don't think it's an "absurdity" to view single cells as carrying out IBE in a dynamical sense, I think it's an absolute necessity to do so if we are to successfully naturalize human knowledge itself.
In my view, the fundamental mistake is to view single cells as "machines." If we do this, then (a) it becomes impossible to explain how such "machines" arose in the first place, and (b) it becomes impossible to explain the transition from mere machines to us (unless, of course, you are prepared to admit that we ourselves are nothing but "machines").
In order to naturalize our own cognitive powers (which are themselves overwhelmingly IBE---deductive logic capability emergent with language being the froth on the wave of the ocean of natural knowlege), it is necessary to naturalize the cognitive powers of living matter generally. This cannot be done within a purely mechanistic framework (or so I am convinced).
What else is required? Something that gives living matter an internal impetus (conation, striving), which establishes normative goal states for the system as a whole. (This is why I believe that QFT has to play a role.) Cognition, then, is the adjustment of means to ends to achieve these goal states, and IBE is a characterization of this cognitive ability which all life possesses.
I guess that was a long-winded way of saying, Of course, you are right!
[ 08 June 2002, 08:57: Message edited by: James A. Barham ]
IP: Logged
|
|
Danpech
Member
Member # 163
|
posted 14. June 2002 00:34
I have not read any of the literature on this subject and that may be why I either do not know what you are really talking about, or I do gather what it is that you are talking about but simply disagree.
I say that deduction is prior to induction, for, in the absence of any fact that can function as an implicit premise, no induction is possible. What I mean is, the principle of, and objects pertaining to, deductive reasoning must exist before any inductive activity can take place, and that it is the self-awareness of the reasoner that serves as one of the primary premises:
1) "I exist" (the law of identity).
2) "I do not not exist" (the law of non-contradiction)
3) "I can discover truths" (by induction)
1) and 2) are the basis of all deduction because together they constitute the prime implicit deduction of the reasoner: "I exist (A=A), therefore I do not not exist (A/not A)". The formula A=b is what (valid) deduction can be reduced to, simply because valid deduction amounts to the principle of equivalence, which is just a complex mode of the principle of identity:
a) All y have y.
b) Fred is an instance of y.
c) Therefore, Fred has y.
(NOTE: I am not at all confident that I have accurately, or even correctly, expressed above all that I intend to express here, as I am very tired and this is not a subject that I have developed much within my own thinking.)
IP: Logged
|
|
Leonid Andreev
Member
Member # 282
|
posted 21. June 2002 17:25
Danpech:
The context of my question lies apart from the academic basics. In fact this issue is the cornerstone in the AI foundation and is far from being clear. I understand your logic but I cannot accept it, for the devil is in detail – in those details that are ignored in logical inferences of the type you cited, or rather go unnoticed as they are below the noise level. Let us not forget that most of the great scientific discoveries were not made with help of formal logic, and especially so, not based on logic of the Greek sophism, but contrary to it.
Despite all the vast knowledge generated by biological science, little is known about the reality of the origin of life on Earth and about its teleological drive or the so-called evolution thrust. It would be grossly overstating to claim that we know anything definite about the actual mechanisms underlying the functioning of consciousness. Knowing what a neuron looks like, or that synaptic signals are responsible for information transmission, and a lot more about the brain structure and functions does not, however, give us any confidence in our ability to understand the biological, chemical and physical foundations of the phenomenon of consciousness as a whole.
Nevertheless, we, homo sapiens, have an indubitable advantage: being carriers of consciousness, we are able to analyze consciousness. When trying to understand, for instance, why the DNA of eukaryotes, unlike that of prokaryotes, consists of almost totally repetitive sequences, we realize that we would be unable to investigate this issue without recourse to the entirety of life sciences. When we attempt applying our consciousness to analysis of consciousness, we feel that we can do it. After all, when trying to create the artificial intelligence, we are not going to copy the biological system of the brain functions. We are going to design a machine equipped with knowledge and intelligence. However, in doing so, we will not avoid the necessity to adopt the fundamental principles of interrelation between induction, deduction and associative reasoning that are true for human consciousness.
As I see it, the title question of this thread points to the central challenge for researchers in consciousness and AI, for the real brake in understanding the interrelation between deduction and induction is the part of our consciousness that is controlled by deduction.
IP: Logged
|
|
|
Printer-friendly view of this topic