In this paper we provide an account of information and
informational content, show how it accords with certain intuitive
principles of information, and use it to resolve an apparent tension
among those principles.
Our aim is not to provide a semantics for talk about information, but to provide an account of information itself. Still, it will be helpful to begin with some observations about the structure and logical properties of information reports.
(1) The x-ray indicates that Jackie has a broken leg.(2) The acoustic waves from the speaker carry the information that the announcer said, ``Nancy Reagan is irritated.''
(3) The fact that the x-ray has such and such a pattern indicates that Jackie has a broken leg.
(1) and (2) have a structure similar to reports of propositional attitudes. We call an information verb or verb phrase (`shows', `indicates', `carries the information'), together with the preceding noun phrase, an informational context. We shall call the proposition designated by the `that'-clause the informational content. The object designated by the intial noun phrase of a report like (1) or (2) we shall call the carrier of the information; the fact designated by the initial noun phrase of a report like (3) we shall call the indicating fact.
Both styles of information reports are factive. That is, if the report is true, the informational content is true too. If the x-ray indicates that Jackie has a broken leg, then she does. In particular, if the fact that the x-ray has such and such a pattern indicates that her left hind leg is broken, then it is. In this way information reports differ from reports of some cognitive attitudes and linguistic acts, but are similar to others. What is believed or conjectured or asserted or denied need not be true, although what is seen or known must be. And in this way information reports differ from reports of what is possible, and are similar to reports of what is necessary.
Information contexts, like modal contexts and propositional attitude contexts, are clearly not truth-functional. Reports of cognitive attitudes and linguistic acts differ from modal statements, in that substitution of necessarily equivalent statements in the latter preserve truth value, while this is not so with the former. On this issue, information reports are like reports of cognitive attitudes and linguistic acts. For example it does not follow from (1) that the x-ray shows that Jackie has a broken leg and 7+5 = 12.
Like `believes,' and `is necessary,' `indicates' distributes across and but not across or.
If x indicates that P and Q, then x indicates that P and x indicates that Q.x may indicate that P or Q, even though x neither indicates that P nor indicates that Q.
Reports of linguistic acts and cognitive attitudes are notoriously
opaque
[4] --substitution of codesignative terms in the
content sentences does not always preserve the truth value of the
whole report. As we might expect given the emphasis in recent
philosophy of language on the different semantic properties of names
and definite descriptions, it is important to distinguish two kinds of
opacity. Opacity with respect to definite descriptions is
relatively noncontroversial, common, and well understood. Modal,
cognitive, linguistic, and informational reports all exhibit such
opacity, when read in a certain way. Even though Jackie is Jonny's
dog, we cannot infer from (1) that the x-ray indicates that Jonny's
dog has a broken leg, if we take this to mean that Jonny's ownership
of the dog is part of what is indicated.
Opacity with respect to proper names is less common, more controversial, and less well understood. It seems that if Tully was necessarily human, so was Cicero; that if Cicero was possibly the best philosopher of his age, so was Tully. But it seems that someone might well say or believe that Cicero was the best Roman philosopher, while not believing or even explicitly denying that Tully was. Should we say that such a person said or believed that Tully was the best Roman philosopher, simply because he said or believed that Cicero was? This would be misleading. Would it be false? Philosophers differ over whether the opacity is real and semantic or apparent and pragmatic. If the former, then it seems that on this score, information reports are closer to modal statements than to reports of cognitive attitudes and linguistic acts. If the entrails of some animal showed that Tully had a broken leg, then they showed that Cicero did.
Information contexts, then, are factive and not truth-functional; substitution of necessarily equivalent content statements does not preserve truth; they distribute across and but not across or; they may support opaque readings with definite descriptions in the content sentences, but not with proper names.
With our intuitions thus sharpened, let us turn from information reports to information itself.