What is an informational environment? 93
At first blush, none of these accounts seem to include a role for information
to play in determining the set of relevant conditions or to be part of these conditions. With respect to the distinction among external, ecological and selective
environments, this observation is correct, yet the exclusion of information on this
level does not amount to an omission. Instead, it warrants a conceptual generality
that is fully adequate: many organisms neither are able nor need to collect much
information on many or most of the relevant conditions in their environment in
order to meet their adaptive needs. To an external observer though, the traits and
demographic performance of those organisms may provide information on conditions in their environments, in terms of covariance with the values of some
relevant variable or variables in that environment. Conversely, the definition of an
ecological environment, in terms of the set of conditions relevant to an organism,
is impartial to whether and how information is collected about these conditions
by its inhabitants. What matters first is that the respective conditions are relevant.
With respect to the construction/constitution case, Godfrey-Smith actually does
not omit information from his account of the relation between environmental complexity and the function of cognition of which that distinction is part. Instead, he
reserves his use of a vocabulary of information for what can be modelled in terms
of Lewis- Skyrms style sender-receiver games (Godfrey-Smith 2013a; 2013b; see
the discussion in Chapter 2). Natural information will not fall within the extension of this more restricted usage, where “unsent signs” figure as “cues” instead.
Terminological preferences aside, Godfrey-Smith is duly concerned with the role
of what I have been referring to as natural information (aka cues) in environments
as a constituent of cognitive processes. His view echoes Fred Dretske’s theory of
information in some respects:
Cognition is useful in an environment which is characterised by:
(i) variability with respect to distal conditions that make a difference to
the organism’s well-being, and by
(ii) stability with respect to relations between these distal conditions and
proximal and observable conditions.
[. . .] Cognition is most favored when there are (i) environmental conditions
salient to the organism, which are not directly observable, and which are not
stable or predictable in advance, and when there are also (ii) highly reliable
correlations between these distal states and states which the organism can
observe or detect more directly.
(Godfrey-Smith 1996a, 118)
It is the “highly reliable correlations” between distal and proximal affairs that
count as information in this context, where the distal affairs are the conditions at
the source and the proximal affairs are the signals. Variability at the source is as
important to the role of information in cognition as stability and unequivocality
of the correlations (for the terminology used in this reconstruction, see NI-4 and
At first blush, none of these accounts seem to include a role for information
to play in determining the set of relevant conditions or to be part of these conditions. With respect to the distinction among external, ecological and selective
environments, this observation is correct, yet the exclusion of information on this
level does not amount to an omission. Instead, it warrants a conceptual generality
that is fully adequate: many organisms neither are able nor need to collect much
information on many or most of the relevant conditions in their environment in
order to meet their adaptive needs. To an external observer though, the traits and
demographic performance of those organisms may provide information on conditions in their environments, in terms of covariance with the values of some
relevant variable or variables in that environment. Conversely, the definition of an
ecological environment, in terms of the set of conditions relevant to an organism,
is impartial to whether and how information is collected about these conditions
by its inhabitants. What matters first is that the respective conditions are relevant.
With respect to the construction/constitution case, Godfrey-Smith actually does
not omit information from his account of the relation between environmental complexity and the function of cognition of which that distinction is part. Instead, he
reserves his use of a vocabulary of information for what can be modelled in terms
of Lewis- Skyrms style sender-receiver games (Godfrey-Smith 2013a; 2013b; see
the discussion in Chapter 2). Natural information will not fall within the extension of this more restricted usage, where “unsent signs” figure as “cues” instead.
Terminological preferences aside, Godfrey-Smith is duly concerned with the role
of what I have been referring to as natural information (aka cues) in environments
as a constituent of cognitive processes. His view echoes Fred Dretske’s theory of
information in some respects:
Cognition is useful in an environment which is characterised by:
(i) variability with respect to distal conditions that make a difference to
the organism’s well-being, and by
(ii) stability with respect to relations between these distal conditions and
proximal and observable conditions.
[. . .] Cognition is most favored when there are (i) environmental conditions
salient to the organism, which are not directly observable, and which are not
stable or predictable in advance, and when there are also (ii) highly reliable
correlations between these distal states and states which the organism can
observe or detect more directly.
(Godfrey-Smith 1996a, 118)
It is the “highly reliable correlations” between distal and proximal affairs that
count as information in this context, where the distal affairs are the conditions at
the source and the proximal affairs are the signals. Variability at the source is as
important to the role of information in cognition as stability and unequivocality
of the correlations (for the terminology used in this reconstruction, see NI-4 and