Paul B. Kantor © 2002
Just as communication may fall into the error of imagining that because we learn about the world and about other people through communicative processes, communication is more fundamental than the world that is communicated about, so also library and information science may fall into the error of imagining that the theory of organizing information about the world is more fundamental than the world thus described. You can see that error worked out in various ways in the readings. While epistemology (the theory of knowledge) is a noble endeavor (as is LIS), they are not the same. So let’s try and formulate a theory, at the level that Dyson proposes.
Recall that for Dyson, “the germ theory of disease” is a theory. Specifics such as the role of the immune system, or of public health, are more at the level he would call models. They are important, but they are intentional oversimplifications, introduced so that we can make progress. In other words they address parts of a complex whole, in a way that permits practical advance.
So, for a theory of LIS. Let’s introduce two fundamental abstractions: “text” and “meaning”. These are rather like “germ” and “disease”. The former is more concrete, but still varies a lot. The latter is more abstract, and may be very hard to pin down. However, we don’t need to pin them down in order to make a fundamental statement about theory. But we do need to introduce one more concept. [This does not make our theory more complex, or better, or worse, than the GTOD]. I will call this concept the “Index”. The fundamental idea is that one may work with (or on) a “text” and produce from that work an “index entry”. The index entry is supposed to represent the text.
So now for the fundamental theory: Texts and their meanings are not unrelated. [Sounds a little like Foucault – with that negative]. I choose this negative form because if I take a positive form, and say that “they are related” you will immediately ask me how. And then I have to descend into one or another model. But we all can have an idea of what “not unrelated” is supposed to mean. One way to say it (this is the Shannon view, by the way) is that “related” means: if you know something about a text, you also know something about its meaning.
But I said I needed another concept (the Index, for those who were not paying attention) and now I’ve gone ahead and not used it at all. So what’s going on? Well, there is in fact one more concept which is really needed too. This is the concept of the “reader” (of the text). You may call this reader the “conctructor” or “deconstructor” or whatever. I don’t much mind. The point is that the “meaning” does not reside in the text, but resides in the mind (? Well, if readers have minds) of the reader, as a result of interaction with the text. To show that I have read some math, I will write Meaning=Meaning(Reader, Text). This means that a particular instance of Meaning is a function (the “Meaning function”) of both the reader and the text. This means that it depends, in an essential way, on both.
So the claim of this fundamental theory is that this Meaning function really does depend on the text, and you cannot ignore it. [An example would be to say that f(x,y)=xy. If either variable, x or y changes, not only does the function change, but even its dependence on the other variable changes.
And what about the index? An index can be thought of also as a kind of “meaning”. We can emphasize that by writing Index=Index(Agent, Text). Now, why didn’t I write Reader instead of Agent. This is a cute question. Librarians (who are such agents) are trained, in a sense, to act in predictable ways. This is different from the notion of the reader. In truth, is the very fact that librarian-agents follow “rules” that gives us any hope of developing non-human indexing agents (currently, computer programs; perhaps one day we can train pigeons or dolphins to index texts).
So what does this framework give us? It gives us a basis from which to describe the various models that are represented in the packet of readings. We will find that these models often present themselves as focused on (or gaining their strength from) one of those perspectives. Work based on a theory of knowledge could be said to concentrate on the Meaning function. So-called system oriented work is focused on the Index function. Rule-based methods seek to foreground the Agent. Cognitive methods foreground the Reader.
What should we ask of a theory regarding a practice? A good starting point here is the work by H. Simon on the “Sciences of the Artificial”. He discussed the fact that sciences studying human constructions are not the same as sciences studying the naturally occurring world. In particular, constructions can be changed, as needed, to make them perform better, and they are generally understood in a context where there are defined “goals”.
Let me be clear about this. In current scientific thinking, planets, for example, do not have goals. Space exploration, on the other hand, does seem to have goals. When we move to the social sciences it becomes a little less clear. Do human constructs (institutions, societies, fraternities, juries, …) have goals? People who try to apply a goal oriented view to such institutions are usually forced to admit that there will be at least 3 layers of goals: goals of the individual participants; goals of the institution itself, which may be more or less explicit, and finally goals of the surround in which it has been brought into existence. Of course the three kinds of goals rarely agree, but they often have some common elements, which makes it possible for the individuals to participate effectively in the institutions on behalf of the surround.
Now let’s come back to LIS. How do the papers that we read here situate themselves with respect to this framework and “theory”. Generally they are “models” in the sense that they focus on one or another aspect in order to make progress.
Some of them are “orienting models” -- that is, they propose a general thing to think about, but not a very concrete way of thinking about it. Others are “reducing models”, they replace one or another abstraction appearing in the general picture by an “operationalization” that makes it possible to specify very concrete lines of research.
It is worth mentioning here that the norms of scientific discourse differ in one very important regard form the norms of social discourse. In social discourse it is generally desired to avoid error (“better to keep still and be thought a fool, than to open your mouth and remove all doubt”). This is desired by speakers, and also by co-participants. So if someone, for example, does not seem to understand what a term like “ethno-methodology” means, the norms of social discourse may keep us from correcting him. But the norms of scientific discourse require that such correction be made honestly and openly. This imposes two other changes in behavior. Second (!) the person who is in error violates the science norms if he takes the correction “personally”, rather than taking it as a correction of his “understanding”. But first (!) and far more important: the speaker acquires some obligation to speak in such a way that he might actually be seen to be wrong.
Thus, socially, it is fine, and even meritorious, to say “conversations between two people are enormously complicated things”. But the scientist contributes more by saying “Conversations between two people simply do not ‘convey information’.” Of course we might suggest that this idea is wrong, and speculate on how Marcia might have answered the question “where is Donny right now?” before and after that fateful phone call. But the speaker has “done right” by “being wrong” because we now know something more about how to describe conversations.
Note also that when we seek to converse in this way tiny distinction become very important. For example, the sentence “Conversations between two people do not simply ‘convey information’” has exactly the same words. And they are in almost exactly the same order. One word has been moved two positions. Yet they say different things. The example given would not refute the second statement. In fact, it would be irrelevant to it. But some other test might be proposed. For example, we might interview Donny and ask him why he did not explicitly ask for help. The answer might lead us to some deeper insights about the function of the conversation. Or we might be so brilliant that we formulate the deeper insights without even asking. Then, of course, it would be our turn to try to formulate those insights so clearly that others might find them to be not correct.
This is the practical meaning of Popper’s concept of “falsifiability” and it is a spectacularly powerful tool for limiting the range of discourse on scientifically definable topics.
Just two final consideerations, and I will let go of the pen. When we talk about a function, such as Meaning(Reader, Text) it is sometimes useful to talk about the “Range” of the function. In mathematics, this will be the “set of values that the function may assume”. We may find it useful to resort to this “Range” concept when we read papers that talk about the “organization of knowledge” (a particular hobby of L-theorists). While we cannot be quite sure what this “knowledge” stuff is, we may suppose that it has some close relation to the set of meanings. It’s probably a subset, and it probably tends to contain meanings that are shared by more than a handful of people.
The other is the kind of simplification that occurs in much experimental or observational research. In these settings, the abstractions (Reader, Text, Agent, Meaning) are replaced by much simpler concrete notions. For example, “relevance theory” says something like: there is one aspect of meaning that could be called the Value of the meaning, and that value in turn depends on (or perhaps is synonymous with) something that we will call the “Relevance of the text for the reader”. Or a researcher may replace the entire reader by a few specific characteristics of the reader: age, area of study, native language, etc. this results in a model equation which has the same form as the fundamental equation but might read: Value=Value(age, education, question; length, reading level, keywords).
I think this is enough for a “warm-up”, and I will invite us to see how well (or poorly) it stands up against the set of readings that we encounter.