Resurrecting Dretskean information 25
So here are the points that I see in Dretske’s theory that are worthy of being
‘resurrected’: first, information is, above all, natural information, in terms of being
present in the environment independently of any means of signal transmission or
processing, and independently of senders and receivers. Second, and conversely,
informational relations serve to ground content for their receivers. Third, information is relationally defined, in terms of lawful covariance between r and conditions
at s, with no reified informational entities being postulated. The term “information” denotes a set of invariant relations in the environment that is foundational
to the possibility of keeping track of objects and events throughout a multitude of
conditions for a variety of differently constituted organisms. Hence, one key task
will be to tell the variant from the invariant relations in this context. However, the
question remains of how to account for the possibility that, under certain conditions, less-than-unequivocal relations (in which p < 1 for s being F, given r) might
be of informational value to such a variety of differently constituted organisms,
and that these conditions are the normal conditions under which perceiving and
cognising organisms operate. Knowledge might not even be the gold standard on
which to evaluate these conditions.
Alternative views of information
In contrast to Dretske’s rigid approach to probabilities concerning informational
content, Skyrms uses a more relaxed and expressly game-theoretic approach of
“signalling games” between senders and receivers (Skyrms 2010, 35). On his
account, the informational content of a signal shall be determined from the probability with which tokenings of the signal, in relation to some world affair, will
influence the behaviour of the receiver (Skyrms 2010, 33–41). The direction and
magnitude of that influence (the “vector”) plainly is the informational content of
the signal. The content vector determines the meaning of a signal within the context of a specific signalling game that applies to a given situation for the organism
using and, in a subset of cases, also the organism producing the information in
question. The direction of the vector might be objectively measurable, but it is a
direction relative to the organism involved. His constitution, his abilities and his
aims need to be taken into account. The information effects something in relation
to him, and what it effects – in most cases, the adaptive response on individual or
population levels – is the content of that information.
If a signal always and exclusively occurs in correlation with one of n possible
states of affairs, its content is unequivocally determined within the context of
the game, as is the response from the receiver. If the relation is less than unequivocal while still being identifiable as a vector of the aforementioned kind, the
information transmitted is weaker, but it does not cease to be information. On
this view, the probabilities for informational content come in frequency-based
degrees. Moreover, the probabilities involved are now all treated on one level. If,
on the repeated occurrence of some signal, there is an effect on the behaviour of
an organism or a population in terms of an increased probability of some kind of
behaviour to obtain in comparison to all available alternatives, and hence if the
So here are the points that I see in Dretske’s theory that are worthy of being
‘resurrected’: first, information is, above all, natural information, in terms of being
present in the environment independently of any means of signal transmission or
processing, and independently of senders and receivers. Second, and conversely,
informational relations serve to ground content for their receivers. Third, information is relationally defined, in terms of lawful covariance between r and conditions
at s, with no reified informational entities being postulated. The term “information” denotes a set of invariant relations in the environment that is foundational
to the possibility of keeping track of objects and events throughout a multitude of
conditions for a variety of differently constituted organisms. Hence, one key task
will be to tell the variant from the invariant relations in this context. However, the
question remains of how to account for the possibility that, under certain conditions, less-than-unequivocal relations (in which p < 1 for s being F, given r) might
be of informational value to such a variety of differently constituted organisms,
and that these conditions are the normal conditions under which perceiving and
cognising organisms operate. Knowledge might not even be the gold standard on
which to evaluate these conditions.
Alternative views of information
In contrast to Dretske’s rigid approach to probabilities concerning informational
content, Skyrms uses a more relaxed and expressly game-theoretic approach of
“signalling games” between senders and receivers (Skyrms 2010, 35). On his
account, the informational content of a signal shall be determined from the probability with which tokenings of the signal, in relation to some world affair, will
influence the behaviour of the receiver (Skyrms 2010, 33–41). The direction and
magnitude of that influence (the “vector”) plainly is the informational content of
the signal. The content vector determines the meaning of a signal within the context of a specific signalling game that applies to a given situation for the organism
using and, in a subset of cases, also the organism producing the information in
question. The direction of the vector might be objectively measurable, but it is a
direction relative to the organism involved. His constitution, his abilities and his
aims need to be taken into account. The information effects something in relation
to him, and what it effects – in most cases, the adaptive response on individual or
population levels – is the content of that information.
If a signal always and exclusively occurs in correlation with one of n possible
states of affairs, its content is unequivocally determined within the context of
the game, as is the response from the receiver. If the relation is less than unequivocal while still being identifiable as a vector of the aforementioned kind, the
information transmitted is weaker, but it does not cease to be information. On
this view, the probabilities for informational content come in frequency-based
degrees. Moreover, the probabilities involved are now all treated on one level. If,
on the repeated occurrence of some signal, there is an effect on the behaviour of
an organism or a population in terms of an increased probability of some kind of
behaviour to obtain in comparison to all available alternatives, and hence if the