The domains of natural information 59
less tightly bounded but more natural conditions. Knowledge may not always
and under all circumstances belong to the conditions of success in perceiving and
acting within an environment. Conversely, setting the admission criteria too loose
might be similarly dissatisfying, because there are innumerable ways in which
all sorts of world affairs are related, with whatever degree of probability. If such
relations could pass as informational without being reasonably unambiguous and
reliable, and perhaps even without being detectable to any organism, the cognitive
value of information would be inflated away.
The previous definitions, in their reference to strict natural regularities and
reference classes, aim at allocating a proper place both to the “nomological” or
“specification” views defended by Dretske and Gibson and the “probabilistic” or
“local” views advocated by Millikan and Skyrms in particular. The mark of distinction between the respective domains of application of those accounts is that
between the description of information as being present in the environment (NI-1)
and as being factually available to a receiver (NI-2 and NI-3). The claim that the
limited and imperfect availability of natural information to an organism leads to
underspecification, as indicated in NI-2b and NI-2c, does not militate against the
specificity criterion in NI-1.
According to NI-1, information basically is specification of the environment.
It is so to the extent that there are invariant relations between world affairs that
could be tracked by some receiver, if only hypothetically. If there were no such
relations, a receiver would have no possibility of keeping invariant aspects of
the environment apart from the variant ones. Nor could he keep rule-governed,
invariant-preserving transformations apart from random ones. On this level of
analysis, the reference class for the informational relations will comprise of the
set of all uniformly regular invariant relations in the world that could be possibly
tracked by an – unspecified – observer. Hence, this part of the definition covers
both Dretske’s requirement of a nomologically governed probability of p = 1 for
information to obtain and the independence of that relation from any particular
observer. By implication, NI-1 excludes from the class of informational relations
such relations between world affairs that are probabilistic, for being governed by
a disjunct set of regularities {i 1 ,. . ., i n } or for including an element of randomness
and thus altogether failing to meet condition I.
Conversely, NI-2 excludes the hypothetical case of relations that, even if they
were properly regular, would be unobservable in principle (although perhaps
manifesting themselves in other, intractable ways). Informationally inscrutable
situations in this sense are part of certain metaphysical or religious doctrines, such
as the Hegelean Weltgeist, cosmic conspiracies, the Judeo-Christian inscrutable
God or the Occultation of the Mahdi in Shia Islam. For these sorts of boundary
case, there would be no reference class identifiable in accordance with NI-2 or
any of its sub-clauses.
Epistemically more mundane and relevant are situations of epistemic uncertainty as they are characteristic of many areas of probabilistic, statistics-reliant
science. In this kind of case, one typically expects some natural regularity or law
I to underlie a set of variant phenomena, while having to rely on signals that are
less tightly bounded but more natural conditions. Knowledge may not always
and under all circumstances belong to the conditions of success in perceiving and
acting within an environment. Conversely, setting the admission criteria too loose
might be similarly dissatisfying, because there are innumerable ways in which
all sorts of world affairs are related, with whatever degree of probability. If such
relations could pass as informational without being reasonably unambiguous and
reliable, and perhaps even without being detectable to any organism, the cognitive
value of information would be inflated away.
The previous definitions, in their reference to strict natural regularities and
reference classes, aim at allocating a proper place both to the “nomological” or
“specification” views defended by Dretske and Gibson and the “probabilistic” or
“local” views advocated by Millikan and Skyrms in particular. The mark of distinction between the respective domains of application of those accounts is that
between the description of information as being present in the environment (NI-1)
and as being factually available to a receiver (NI-2 and NI-3). The claim that the
limited and imperfect availability of natural information to an organism leads to
underspecification, as indicated in NI-2b and NI-2c, does not militate against the
specificity criterion in NI-1.
According to NI-1, information basically is specification of the environment.
It is so to the extent that there are invariant relations between world affairs that
could be tracked by some receiver, if only hypothetically. If there were no such
relations, a receiver would have no possibility of keeping invariant aspects of
the environment apart from the variant ones. Nor could he keep rule-governed,
invariant-preserving transformations apart from random ones. On this level of
analysis, the reference class for the informational relations will comprise of the
set of all uniformly regular invariant relations in the world that could be possibly
tracked by an – unspecified – observer. Hence, this part of the definition covers
both Dretske’s requirement of a nomologically governed probability of p = 1 for
information to obtain and the independence of that relation from any particular
observer. By implication, NI-1 excludes from the class of informational relations
such relations between world affairs that are probabilistic, for being governed by
a disjunct set of regularities {i 1 ,. . ., i n } or for including an element of randomness
and thus altogether failing to meet condition I.
Conversely, NI-2 excludes the hypothetical case of relations that, even if they
were properly regular, would be unobservable in principle (although perhaps
manifesting themselves in other, intractable ways). Informationally inscrutable
situations in this sense are part of certain metaphysical or religious doctrines, such
as the Hegelean Weltgeist, cosmic conspiracies, the Judeo-Christian inscrutable
God or the Occultation of the Mahdi in Shia Islam. For these sorts of boundary
case, there would be no reference class identifiable in accordance with NI-2 or
any of its sub-clauses.
Epistemically more mundane and relevant are situations of epistemic uncertainty as they are characteristic of many areas of probabilistic, statistics-reliant
science. In this kind of case, one typically expects some natural regularity or law
I to underlie a set of variant phenomena, while having to rely on signals that are