Positivism

Positivism is, in short, the notion that all knowledge (and therefore all intelligence) is based entirely on the data of experience, that the world is a world of facts, and that these facts follow the rules of pure logic and pure mathematics. Positivism is therefore an antimetaphysical position; it takes the position that, in short, everything that is known is experienced. It is clear that as far as the study of cognition is concerned, the idea of an intelligent soul, or some divinely granted power of intellect, is thrown out the window. One can come to understand how one thinks by examining the mind and/or the brain.

More important, however, is the second of the above premises, that the world is made of facts. The philosophies of Descartes, Hobbes, and Leibniz had established a rational, atomistic tradition in which all phenomena can be described as complicated combinations of simple elements (Descartes), these elements were formal in nature and related by purely syntactical operations, so reasoning could be reduced to complex calculation (Hobbes), and that the process of understanding is the process of reducing to primitives (Leibniz). This position took its ultimate form in Ludwig Wittgenstein's early Tractatus Logico-Philosophicus, in which:

  • The world is made of facts.
  • Facts can be reduced to primitive objects.
  • If all primitive objects are understood, all facts are understood.
  • The mind is engaged in "making pictures of facts."
  • Complex symbols in the mind are combinations of these primitive objects.

GOFAI, then, takes as its premise the idea that AI is the search for primitive elements and primitive logical relations in the world, and the replication of these elements and relations in a computer‹just as human intelligence is the replications of the elements and relations of the world in the brain.

This seems a reasonable claim, especially since this is precisely how digital computers work, and they can certainly produce very complex behavior. The Dreyfus brothers, however, remind us of Heidegger, who pointed out that traditional philosophy is defined by focusing on the facts of the world while "passing over" the world as such. Since Western philosophy's beginnings, it has been taken for granted that understanding something means having a theory of it‹that is, relating objective, context-free elements in terms of abstract rules, programs, and principles. What is perhaps most ironic about the relationship between AI and positivism is that in 1953, when AI was just taking off, Wittgenstein published his Philosophical Investigations, in which he fully repudiated the symbolicist position he had taken up earlier in the Tractatus. As will be discovered more thoroughly in the Phenomenology section, what is missing from the positivist position is the notion of everyday know-how; for instance, though commonsense physics is extremely hard to spell out as a set of facts and rules, everybody has some idea of commonsense physics which enables them to move through their daily lives.

The Dreyfus brothers point out that symbol-based, positivistic AI has been widely discredited precisely because it has been put to the empirical test and, beyond a given level of complexity, failed. As a result, the connectionist paradigm‹creating a holistic model of a brain‹became the predominant area of AI research.

back