Peter van Eijk Department of Computing Science Rijks Universiteit Groningen Summer 1982 English translation: Oct 1995
Contents:
Introduction - Dennett's Tools - My Tools - How
does man see the world
A Hofstadter-ian strange loop - Conclusion - References.
"If you meet the Buddha
on the road,
kill him!"
Information processing has become the magic word for a number of sciences that in one way or another have the human as their subject of enquiry (in this paper I will abbreviate the stem of information processing and similar words to ip, in this case iping). With this, the concept of iping has acquired the status of scientific paradigm. It becomes interesting therefore to investigate the tools that are in use to understand iping. In this essay I will attempt to contrast Dennett's tools [Dennett 78], which have a psychological and philosophical background, with my ideas, which have their background in the core computer science curriculum. One can also say that I try to rephrase a part of Dennett's ideas into my own words.
Dennett introduces the concept of intentional system. With that he hopes to have a good starting point for philosophical excursions. The behaviour of intentional systems can be described in terms of a set of "beliefs" and "desires". This set, the so-called beliefsystem, then forms a description of the system at the intentional level (stance). Dennett emphasizes that this description is relative to the describer and his goals. In addition to the intentional level Dennett distinguishes the design level and the physical level. A description at the design level is in terms of the functions of the components of the system, while a description at the physical level only uses the physical properties of (the realisations of) these components. It will be clear that at these latter two levels the descriptions will become much more complex.
Dennett claims that an attribution of intentionality corresponds to an attribution of rationality. Because of this, the concept of intentionality can be used to show where a theory of rational human thinking is using homunculi (little humans in the human), namely there where the theory uses intentional terms for description.
Belief systems can be seen as a kind of world models, they contain information about the environment: facts, dependencies, etc. We can therefore consider the intentional level as a kind of meaning level. Through the use of belief systems the system assigns a certain meaning to specific events, for example whether these events should be avoided (are dangerous).
Software is a collective name for descriptions of ipers, in particular those that can be realised by automata (or machines). Computer science studies, amongst others, the construction of software.
Continually recurring concepts in computer science are "representation" and "structuring". Input, output, and state of an iper all have their own internal representation. Important elements of a representation are codings (for example red=1), where meanings are mapped to numbers, and structurings (the way of decomposing such representations). A good structuring makes it possible to study a representation at several levels, and to abstract from codings. This explains the importance of the terms hierarchy and level: a hierarchy is a structuring that is divisible in levels. Besides these, we have the concept of interpretation. This encompasses the concept decoding.
So: a coding maps meanings to numbers (or bits), a representation maps meanings to meanings, and an interpretation maps numbers to meanings.
Most computer programs convert information in one representation into another representation. If these representations are not adequate, this process will usually involve a search process. If, for example, a quadratic equation is represented in polynomial form, its zeros can be computed, whereas with a representation in functional form they need to actually be searched for with some kind of approximation method. This conversion adds nothing to the amount of meaning: in a closed ipsystem, the total amount of information can only decrease. Stated differently, the input plus the computerprogram are a redundant description of the output. A classic example of information reduction is in so-called management information systems, where megabytes of data on the ups and downs of the enterprise are condensed into a few well-organized tables and graphs on which decisions are based.
This leads us to the conformity principle. It is a fact of experience that the structure of an iper is related to (is in conformity with) its environment. The iper models in a certain sense reality, its environment. This modelling is naturally relative with respect to the goals because, in general, it will be imperfect. Illustrations and applications of this principle:
As Dennett already states in chapters 1 and 12, a lot of computer programs of the more complex kind lend themselves to an approach on the intentional level. In my practice, the clearest examples of these are compilers and operating systems. In using these naturally a lot of errors are made when the knowledge of the user is insufficient. The typical scheme of an explanation that puts the ignorant back on the right track is then: "Yes, but if you do this the computer then thinks that you mean that, and since it tries to accomplish such it reacts so." One therefore acts as if the computer assigns meaning to our conduct, like we assign meaning to the conduct of other people. Dijkstra calls this antropomorphizing and opposes it. I call it intentionalising and see it as an attempt to understand something with the tools that are natural to humans.
This tendency to assign meanings to events is a human quality. Apart from the intentionalising as mentioned before it appears as a characteristic of religions. Some examples of these are seeing
In this context, sociologists talk about a "social and religious definition of reality".
For me the ultimate religion is the Zen doctrine, especially because it questions this structuring of reality. The opening quote of this paper is explained in [Kopp 72], in which the search of psychotherapeutic patients for meaning is illustrated with examples from the world literature:
'The Zen master warns: "If you meet the Buddha on the road, kill him!" This admonition points up that no meaning that comes from outside of ourselves is real. The Buddhahood of each of us has already been obtained. We need only recognize it. Philosophy, religion, patriotism, all are empty idols. The only meaning in our lives is what we each bring to them. Killing the Buddha on the road means destroying the hope that anything outside ourselves can be our master.' (page 188)
Stated differently: when it appears that something or somebody starts to bring meaning into your life, you should put an end to that really quick.
The structuring of the world in terms of meaning has, according to the conformity principle, its counterpart in the structure of human thought. For this Newell and Simon have introduced the "physical symbol system hypothesis" [Newell Simon 76] [Simon 81], and what is a symbol besides being a carrier of meaning? This hypothesis states that symbol systems are necessary and sufficient for intelligence in general. According to Simon a computer is also a symbol system, but by now it should be clear that according to me this is more a property of humans than of computers.
How arises this structuring of human thinking (by humans)? The conclusion thrusts itself upon us via the conformity principle: our vision on our structure is a result of that structure. Stated differently: the fact that we think that we think in terms of meaning is caused by the fact that we think in terms of meaning. Apparently this self-involvedness is fundamental in psychology. At least, that is what I have heard about [Linschoten 64].
The concept of meaning is the most fundamental human tool. As derivatives of that operate the concepts of information and information processing. That is because information is created by the favour of attaching meaning to certain patterns, physical or otherwise. One question then remains: how is it that we get these patterns?
This is a translation from the original 1982 paper, which was written in Dutch as a student paper for a course given by prof Michon. Translation and hypertext links by the author. © 1995 Peter van Eijk.
This translation tries to mimic the original Dutch as closely
as possible. As I understand that this is rather terse, and not
particularly accessible, I am in the process of producing a more
up to date version. All comments are welcome, also by e-mail.
Home page Peter van Eijk.