RCode vs CCode

jmc@gang-of-four.stanford.edu (John McCarthy)
Date: Sat, 8 Sep 90 17:53:49 -0700
From: jmc@gang-of-four.stanford.edu (John McCarthy)
Message-id: <9009090053.AA11796@Gang-of-Four.Stanford.EDU>
To: interlingua@venera.isi.edu
Subject: RCode vs CCode
I think

1. A CCode must contain some kind of RCode, although the RCode could
have limited expressivity.  Circumstances arise in which the recipient
of a message knows very little about what the sender is trying to say.

2. The state of the art might dictate that a particular CCode contain
only a limited RCode.

3. A CCode would be more useful if it also permitted communications
that were not complete assertions.  There are two reasons for this.

	a. Many communications are not assertions.  The theory of
speech acts deals with this.  I have been working on a language
called Elephant 2000 (promised for the year 2005) that contains
an input-output language based on speech acts.  While there is
a preliminary draft, I don't want to try to divert Interlingua in
the direction of Elephant speech acts.  The present effort can
result in many useful things.

	b. This second point seems more relevant.  Suppose you ask
me "What is your name?"  I may reply "John McCarthy" and don't
have to make a complete sentence out of it.  You then say,
"Address?", and I tell you my address.  It seems to me that a
CCode should include the possibility of such communication.
Among other things, it would permit communication with programs
that are much less intelligent than would otherwise be possible.

	Maybe it would be convenient to regard all such communication
as abbreviated RCode.  The question might be taken as an abbreviation
of "I would like you to tell me your name".  I'm doubtful that this
is the right way to look at it.

	I conjecture that the simplest and most basic form of
natural language communication involves the utterance of single
words.  Two people involved in a common task often communicate
in single words, and sentences of which these words might be
abbreviations are never formed.  Children begin by communicating
in single words.  Finally, most computer-computer communication
is on a low level equivalent to communicating single words, which
the programs then interpret appropriately.  Communication in
sentences is required when the sender cannot make strong assumptions
about the receiver's mental state.

	For all these reasons, it seems to me that at some point,
the Interlingua group should consider how far it wants to go in
the direction of providing for highly context dependent communication
that doesn't consist entirely of complete sentences.  Doing it may
require imagining that the communicators have mental states.