Resurrecting Dretskean information 21
a conditional probability involved in this definition, the relation in question must
concern types of signals r that compose all the various r-tokens which obtain
when s is F, so that a statistical correlation between r-tokens and F-conditions at
s can be identified. Singular, non-repeatable co-occurrences of r, s and F cannot
establish probabilities. This condition does not imply that informational relations
can only hold for types of F-conditions at s, but it implies that there must be types
of r-events that occur when some s assumes some property F. Thus, F may be
a singular occurrence, or s may be an individual. The presence of informational
content is strictly dependent on the presence of a correlation between r, viewed as
members of types, and s being F.
At the same instance, Dretske’s quest is for the content of a specific, individual
signal, over which one cannot average, plainly and simply because it is an individual. It is either the case or it is not the case that s is F when some r-token is
present. The conditional probabilities established for types of r-signals supposedly govern the individual instances. On Dretske’s view, the correlations involved
are of a strict kind, so that the conditional probability in his definition is p = 1.
Whenever it turns out that s, instead of being F, is G or that t is F instead of s, any
r that seemingly signals s as being F does not convey information, hence is not a
signal, and accordingly has no content either.
To back up this claim, a second necessary condition for the presence of informational content is introduced, which is not contained in Dretske’s definition but
explicitly stated in the context (Dretske 1981, 73, 76f ): r must be tied to the affair
of s being F by some, at root nomological, regularity. Coincidentally parallel
transformations of values at s and r will not count, even if the probability, in terms
of frequency, is p = 1. Information is transmitted if and only if the conditions at
the source affect the conditions on the receiving side in some regular way, that
is in some way that stays the same over the full extension of instantiations of r.
If r may be caused by other conditions at the source than s being F (say G, H,
etc.), or if F-conditions may hold for something other than s (so that z is F) but still
cause r, or if s being F only intermittently causes r, no informational relation with
the content that s is F obtains. Hence, causal relations are not sufficient for establishing informational relations, nor are they always necessary (Dretske 1981, 33–36).
In turn, however, the relation between source s and signal r need not be a relation of
condition F at s causing r. A causal relation between s and r may be indirect (Dretske
1981, 38). For example both s and r might have a presumably distal common causal
antecedent that ensures covariance between these otherwise independent affairs.
If conditions at the source are to be reliably matched on the receiving side, stable “channel conditions” are required under which r-signals can be detected by R
(Dretske 1981, 111–121). The presence and the nature of channel conditions will
account for both the possibilities of information uptake by organisms and the kind
and amount of informational noise that intervenes between source and receiver.
It would not suffice for r to reliably covary with s being F if many or most signals could not reach a receiver R, for being lost in transmission and thus creating
equivocation, or if the intervening noise added seeming r-signals that, in fact, did
not originate at the source.
a conditional probability involved in this definition, the relation in question must
concern types of signals r that compose all the various r-tokens which obtain
when s is F, so that a statistical correlation between r-tokens and F-conditions at
s can be identified. Singular, non-repeatable co-occurrences of r, s and F cannot
establish probabilities. This condition does not imply that informational relations
can only hold for types of F-conditions at s, but it implies that there must be types
of r-events that occur when some s assumes some property F. Thus, F may be
a singular occurrence, or s may be an individual. The presence of informational
content is strictly dependent on the presence of a correlation between r, viewed as
members of types, and s being F.
At the same instance, Dretske’s quest is for the content of a specific, individual
signal, over which one cannot average, plainly and simply because it is an individual. It is either the case or it is not the case that s is F when some r-token is
present. The conditional probabilities established for types of r-signals supposedly govern the individual instances. On Dretske’s view, the correlations involved
are of a strict kind, so that the conditional probability in his definition is p = 1.
Whenever it turns out that s, instead of being F, is G or that t is F instead of s, any
r that seemingly signals s as being F does not convey information, hence is not a
signal, and accordingly has no content either.
To back up this claim, a second necessary condition for the presence of informational content is introduced, which is not contained in Dretske’s definition but
explicitly stated in the context (Dretske 1981, 73, 76f ): r must be tied to the affair
of s being F by some, at root nomological, regularity. Coincidentally parallel
transformations of values at s and r will not count, even if the probability, in terms
of frequency, is p = 1. Information is transmitted if and only if the conditions at
the source affect the conditions on the receiving side in some regular way, that
is in some way that stays the same over the full extension of instantiations of r.
If r may be caused by other conditions at the source than s being F (say G, H,
etc.), or if F-conditions may hold for something other than s (so that z is F) but still
cause r, or if s being F only intermittently causes r, no informational relation with
the content that s is F obtains. Hence, causal relations are not sufficient for establishing informational relations, nor are they always necessary (Dretske 1981, 33–36).
In turn, however, the relation between source s and signal r need not be a relation of
condition F at s causing r. A causal relation between s and r may be indirect (Dretske
1981, 38). For example both s and r might have a presumably distal common causal
antecedent that ensures covariance between these otherwise independent affairs.
If conditions at the source are to be reliably matched on the receiving side, stable “channel conditions” are required under which r-signals can be detected by R
(Dretske 1981, 111–121). The presence and the nature of channel conditions will
account for both the possibilities of information uptake by organisms and the kind
and amount of informational noise that intervenes between source and receiver.
It would not suffice for r to reliably covary with s being F if many or most signals could not reach a receiver R, for being lost in transmission and thus creating
equivocation, or if the intervening noise added seeming r-signals that, in fact, did
not originate at the source.