# The FOA language game

Lurking at the core of the entire FOA enterprise is the fundamental question of SEMANTICS : what do the words in our language \mean? Computer scientists are most familiar with artificial languages (formal grammars, programming languages, etc.), for which precise semantics in terms of a particular machine, are absolutely necessary. Many philsophers of language, notably Frege and Ludwig \Witt, have advocated that a similarly precise semantics of natural language is also possible: words are predicates about states of the world: either they apply or they don't.

An alternative point of view says that such a precise and abstract semantics can never be achieved. What language means is what it means to us, the language users. That is, words' meanings cannot be separated from the Forms of Life of which they are a part. As it happens the same Ludwig has argued forcefully on this side of the debate!

was published in 1922 and is a primary reference for what we now think of as early \Witt''; {\em Philosophical Investigations} [REF640] was published in 1953 and characterizes the late \Witt.'' In the interim \Witt taught elementary school, played music and quit philosophy more than once. But even more striking than the passage of time between these two great works is how diametrically opposed the arguments put forward in{\em Tractatus} and {\em Investigations} are. N. Malcolm [Malcolm67] expresses just how unusual a state of affairs this is: A considerable part of the {\em Investigations} is an attack, either implicit or explicit, on the earlier work. This development is probably unique in the history of philosophy -- a thinker producing, at different periods of his life, two highly original systems of thought, each system the result of many years of intensive labors, each expressed in an elegant and powerful style, each greatly influencing contemporary philosophy, and the second being a a criticism and rejection of the first. [p. 334] (Terry Winograd's turn-around concerning appropriate applications of natural language procesing (NLP) technology, between his dissertation through 1983 [REF358] and his 1986 book with Flores [Winograd86] almost qualifies for early-Winograd vs. late-Winograd, however!:)

You can imagine my reluctance to attempt to characterize just what it was that changed his mind \about, in a short sidebar! Quite roughly then, early-\Witt thought that Language was the perfect philosopher's tool. He aspired to a universal language, shared by all careful users, that could positively and uniquely allow careful {\em naming} of things. Just as numbers point to essential categories and mathematics builds these into theorems about how numbers are related, simple words name simple categories of objects (events, states, ...),and more complicated linguistic expressions name more complicated categories. An utterance of language means the same thing wherever and whenever it is said, just as $2$ does, and just as $a^2 + b^2 = c^2$ remains true wherever. By the time of his {\em Investigations}, \Witt had given up hope that the convenient naming system of mathematics was possible elsewhere. Lanugage for the late-\Witt depended critically on the {\em context} of the utterance. Naming things was only one possible language game; there are many others people play all the time. Without understanding the {\em purposes} to which an utterance is being applied, we can't really understand its meaning. The meaning of a sentence is its use (Gebrauch), its {\em employment} (Verwendung), its {\em application} (Anwedung) [Malcolm67] . This book proceeds on the assumption that \Witt got it right the second time, and focuses especially here on language serving the FOA language game. Your mileage may vary. }

One of most useful devices for getting across his theory of language was his notion of the LANGUAGE GAME ({\em Sprachspiele} in German) [REF640] . \Witt gives many varieties of language games, from chidren's games as simple as ring-around-rosy'' (\S7) to such adult'' games as: \item forming and testing hypotheses \item making up a story; and reading it \item Asking It is interesting to compare the multiplicity of the tools in language and of the ways they are used. (\S23) Certainly FOA counts as another example of a language game, but one with special rules.

Another interesting aspect of theory is how well it anticipates the models of language meaning arising from modern machine learning techniques. The common cause is that \Witt too was centrally concerned with learning, by children. This is evident in his ring-around-rosy'' example, and in his explicit attention to consequences of learning that apply equally well to our algorithms: [Consider] two pictures, one of which consists of colour patches with vague contours, and the other of patches ... with clear contours. The degree to which the sharp picture can resemble the blurred one depnds on the latter's degree of vagueness.... Won't you then have to say: 'Anything and nothing is right.' And this is the position you are in if you look for the definition corresponding to our concepts in aesthetics or ethics. In such a difficulty always ask yourself: How did we learn the meaning of this word [vague]? From what sort of examples? In what language games? Then it will be easier for you to see that the word must have a family of meanings. (\S76,77)

Our current versions of the FOA language games are tied to the technologies by which we are currently allowed to communicate with one another For now centralized search engines are in the center of this dialog. Authors write and sometimes try to influence the audeinces their documents reach. Later, readers use a few of the first words that come to mind to tease out some possible answers. Search engines do their best to connect these two vocabularies.

Reading and writing are the primitive language games on which FOA is based. The tools available to help writers and readers are currently strong constraints in the FOA rules. People can only express what they are allowed to express. If only simple query languages are available, only simple questions will be asked. If all documents are treated interchangeably, as context-free samples of text, then the tacit context assumed by the author is not available.

And so to our abilities to automatically learn what the words really do mean to authors and to readers will change as the evidence the WWW dialogs do. Especially unclear at the present are guarantees about COMMUNICATION PRIVACY AND SECURITY : If we believe all our words are for everyone's ears, then many things will never be said via the Net. If search engines watch over our shoulders as we browse, should we be grateful because it will understand what we \mean, or should we send them a bill for the valuable training data we have provided? As companies like \texttt{Amazon.com} use new technologies which allow them to eavesdrop'' on commercial transactions , consumers must ultimately decide what their privacy, and more effective indexing, is worth to them personally.

## Subsections

FOA © R. K. Belew - 00-09-21