22 Informational environments
In an ideal situation, the channel neither subtracts information generated at the
source nor adds extraneous information. In such an ideal situation, any transformation of conditions at the source would be matched by a transformation of the
signals, and any distribution of probabilities for different outcomes at the source
would be matched by an identical distribution of probabilities for different signals
to obtain. This is, in fact, as much of an idealisation as a friction-less plane or point
masses in the context of scientific modelling – whereas the central question for the
mathematical theory of communication, as Dretske acknowledges, did not concern
ideal situations but real-world problems caused by precisely those noisy channel
conditions which one might just wish to idealise away: how much redundant information should be produced at the source so as to ensure that enough information
will get through a channel affected by a specified amount of noise? That measure
does not concern the reliability of information in terms of the probability with
which r is generated in relation to s being F. Instead, it concerns the conditions
for successful transmission of that information – which do not generate additional
information about s. If channel conditions are variable, additional information on
these conditions (rather than on what happens at s), ideally available through independent channels, will be useful to the receiver – that is if it is available.
It is the question of what happens in less-than-ideal situations that separates Dretske’s theory from its main critiques. The critics claim either that the informationtheoretical terminology of signals, channel conditions, noise and equivocation
he mobilises in his account of informational content is misguiding in terms of
bearing no tenable relation to the mathematical theory of communication from
which it was borrowed (let me call this the “misnomer critique”) or that Dretske’s requirements for informational relations to obtain are too restrictive to meet
real-world conditions (let me call this the “restrictiveness critique”). These two
critiques are related and are best considered in relation to each other.
According to the “misnomer critique”, as brought forward by authors such as
Kenneth Sayre (1983), Dretske’s focus on the content of a signal rather than the
conditions for its successful transmission amounts to changing the topic, so his
reference to that theory is tenuous or even purely rhetorical, and his liberal redefinitions of information-theoretical terms are prone to misguide the reader (Millikan expresses a similar view in personal communication). On slightly different
grounds, but to similar effect, Brian Skyrms (2010, 34) argues that Dretske’s semantic notion of information, in which informational content is conceived of in propositional form, relegates the insights of information theory to secondary importance.
The mathematical theory of communication confined itself mostly to measuring the signal strength and channel conditions required for successful transmission of some signal, whatever meaning it may have. Determining that meaning
was considered an affair of senders and receivers of signals, not of objective relations between world affairs that do not per se presuppose the presence of a sender
or receiver. The relations Dretske had in mind could be detected by a receiver,
and they might be communicated by a sender, but in order to do so and convey
information to them, the relations between source conditions and signals will have
to be of a certain kind, independently of any sender or receiver.
In an ideal situation, the channel neither subtracts information generated at the
source nor adds extraneous information. In such an ideal situation, any transformation of conditions at the source would be matched by a transformation of the
signals, and any distribution of probabilities for different outcomes at the source
would be matched by an identical distribution of probabilities for different signals
to obtain. This is, in fact, as much of an idealisation as a friction-less plane or point
masses in the context of scientific modelling – whereas the central question for the
mathematical theory of communication, as Dretske acknowledges, did not concern
ideal situations but real-world problems caused by precisely those noisy channel
conditions which one might just wish to idealise away: how much redundant information should be produced at the source so as to ensure that enough information
will get through a channel affected by a specified amount of noise? That measure
does not concern the reliability of information in terms of the probability with
which r is generated in relation to s being F. Instead, it concerns the conditions
for successful transmission of that information – which do not generate additional
information about s. If channel conditions are variable, additional information on
these conditions (rather than on what happens at s), ideally available through independent channels, will be useful to the receiver – that is if it is available.
It is the question of what happens in less-than-ideal situations that separates Dretske’s theory from its main critiques. The critics claim either that the informationtheoretical terminology of signals, channel conditions, noise and equivocation
he mobilises in his account of informational content is misguiding in terms of
bearing no tenable relation to the mathematical theory of communication from
which it was borrowed (let me call this the “misnomer critique”) or that Dretske’s requirements for informational relations to obtain are too restrictive to meet
real-world conditions (let me call this the “restrictiveness critique”). These two
critiques are related and are best considered in relation to each other.
According to the “misnomer critique”, as brought forward by authors such as
Kenneth Sayre (1983), Dretske’s focus on the content of a signal rather than the
conditions for its successful transmission amounts to changing the topic, so his
reference to that theory is tenuous or even purely rhetorical, and his liberal redefinitions of information-theoretical terms are prone to misguide the reader (Millikan expresses a similar view in personal communication). On slightly different
grounds, but to similar effect, Brian Skyrms (2010, 34) argues that Dretske’s semantic notion of information, in which informational content is conceived of in propositional form, relegates the insights of information theory to secondary importance.
The mathematical theory of communication confined itself mostly to measuring the signal strength and channel conditions required for successful transmission of some signal, whatever meaning it may have. Determining that meaning
was considered an affair of senders and receivers of signals, not of objective relations between world affairs that do not per se presuppose the presence of a sender
or receiver. The relations Dretske had in mind could be detected by a receiver,
and they might be communicated by a sender, but in order to do so and convey
information to them, the relations between source conditions and signals will have
to be of a certain kind, independently of any sender or receiver.