Wolfram Science Group
Phoenix, AZ USA
Registered: Aug 2003
There are two aspects of the question. One is about human languages, how they actually manage to work, how they are learned and passed on, how anyone ever learns others beyond their native one. The second is about any abstract system of signs and its computational potential, and what that means for the problem of communication more broadly considered.
On the first, we know a fair amount about how languages work, but the whole field of linguistics is testimony that it is not a trivial problem. While a fair amount of work has been done with abstract languages - which in a sense are already models - detailed formal modeling of natural languages is still a relatively new idea. NKS is being used on the problem e.g.
The broader subject is discussed in the NKS book in the context of the somewhat related problem of "Intelligence in the universe", which starts on page 822 and runs to the top of page 840. How one might recognize intelligence without a shared historical background of conventional signs is the question that ties this issue in to the subject of communication. In the notes, "Theories of communication" on page 1181 is directly on topic. There are about 50 index entries about languages, formal human and computer, you might look up.
Empirically, there are over 5000 named human languages. But about half the people on earth speak one or another of just half a dozen largest ones. There is an obvious benefit to using the same language as lots of other people, but details of language spread interact with tons of messy details of political, cultural, and commercial history.
In practice only a handful of linguistic specialists make it their business to learn languages wholesale, rather than learning a handful of particularly useful ones, useful for widespread native use, important literary accomplishments, or as "overlap" languages used for commerce and such ("lingua francas", we say, though technically that was an individual example of a kind of "pidgin").
Ethnographers use grammar structure charts and build dictionaries gradually, sometimes over periods as long as years. External reference e.g. isolating things, pointing, drawing, etc, help them get started. Human languages share many features even where shared past history can practically be ruled out, suggesting some biological basis for elementary aspects (existence of words, thing-concepts, schemes of aggregation or classing, etc). Linguists themselves debate some of these (action words, ways of handling tense, etc), as interpreting the role some feature of a language plays and mapping it to a similar role in a different language, is a largely intuitive business with considerable "play".
Philosophically, this has been discussed as the "problem of translation". The basic issue is whether the structure of relations in a shared external world does or does not ensure one proposition in some system of signs can be checked for full equivalance to a proposition, or set of propositions, in another system of signs. (This is sometimes distinguished from ordinary day to day translation issues in practical life, as "radical translation", which is supposed to exhaustively capture previous objectively verifiable meanings).
This came up in part because positivist philosophers wanted to map all meaningful statements to objectively verifiable referrents, and reject anything left over as "meaningless". Most think that project failed. The problems with it inspired a "linguistic turn" to that strand of philosophy. In some cases the reaction went over to a "relativist" antithesis, which effectively denies translation occurs, except as "interpretation," in the translator's own mental structures.
The basic issue was recognized quite early, that in general the meaning of a set of conventional signs depends on the internal (abstract) relations among other members of that set, not just on the (external, "pointer-like") relations between elements of that set and external referrents. Different schools have made more or less from this, sometimes frankly blowing it out of proportion.
One thing PCE says on the subject may not be immediately obvious. It may seem arbitrarily hard to capture structures that could arise in somebody else's universal formal system. But if you have a universal formal system yourself, it is in principle always possible. That is, claims that there are "thoughts" (small subsets of internal, formal relations) expressable in formal system A that cannot be formally mimicked in formal system B, are untenable if B is universal.
If a "meaning" includes a map to other relations, that might shift. And how easy it is to formulate some set of relations, may vary widely from formal system to formal system. In practice, it might be too difficult a computational task to capture some simple expression in A, in a B that is particular clumsy at that relational structure. But if B is universal then it is always possible in principle. This aspect of PCE detracts from "relativist" conclusions. (There is a short note about that on page 1131).
I hope this helps.
Report this post to a moderator | IP: Logged