114 Environments of intelligence
conditions of persistence and reliability (“glue and trust”), which supposedly
rule out overly ephemeral coupling relations (Clark 2010a, 83f ). If the coupling
in question has a historically established function, the issue of temporary nonoccurrence or non-performance of that coupling relation, rather than being
resolved by means of such auxiliary conditions, is actually dissolved: it does not
matter if the function in question, for temporary loss of one component, turns
out not to be performed in many instances, as long as the coupling occurs with a
degree of reliability and according to an unequivocal regularity so as to become
part of the explanation for the reproduction of the organic and, in some cases, the
environmental components involved. To use a non-cognitive case of extension for
comparison (see Menary 2010b, 14): it does not matter if a spider loses her web
once in a while, or perhaps even very often, as long as the however intermittent
production and use of webs provides part of the explanation for the presence of
web-making mechanisms in the spider population.
Third, and most profoundly perhaps, a historical account of functions will counter Mark Sprevak’s critique of classical machine-state functionalism (2009) which
targets the “Martian intuition” that an identical or analogous kind of function can
be realised by physically or biologically very differently constituted systems. On
that intuition, which is basically the Lewisian one (1980), if we can specify a set
of functions for certain traits and behaviours in human beings and Martians, we
should be able in principle to identify which concrete structures within or around
the human being or Martian will play what sort of causal role in accomplishing
those functions. However, Sprevak objects, one could easily imagine the conditions of persistence and reliability of the respective structures and the coupling
relations involved to be violated in manifold ways, and hence fail to meet the
specifications laid out by extended functionalism. Playing on the Martian intuition
in this way will probably kill the Martian – but, if we follow the historical account
of functions, it will not kill the intuition. Any coupling with whatever structure
that were as unreliable or otherwise insufficient or even freakish in nature as
described by Sprevak would militate against the imagined organism’s reproductive chances, and against the factual (rather than conceptual) possibility of his
existence. Conversely, Sprevak’s Martian may be imagined to be in command of
capabilities and enter into environmental coupling relations so unfamiliar that we
could not find any remote analogue in the specifications of human cognition. But
then, the entire functional analogy will not hold. After all, there might be hypercognitive functions to Martians that human beings are insufficiently equipped to
imagine in the first place.
Clark and Chalmers’s reservations against history diagnosed here become
explicit when they introduce a set of criteria that is supposed to be necessary and
sufficient for the function of internal or external resources as components of cognitive processes (1998, 17; see also Clark 2010b):
(1) The resource should be of reliable availability, and should be used on a
frequent basis.
(2) The information provided by that resource should be directly available.
conditions of persistence and reliability (“glue and trust”), which supposedly
rule out overly ephemeral coupling relations (Clark 2010a, 83f ). If the coupling
in question has a historically established function, the issue of temporary nonoccurrence or non-performance of that coupling relation, rather than being
resolved by means of such auxiliary conditions, is actually dissolved: it does not
matter if the function in question, for temporary loss of one component, turns
out not to be performed in many instances, as long as the coupling occurs with a
degree of reliability and according to an unequivocal regularity so as to become
part of the explanation for the reproduction of the organic and, in some cases, the
environmental components involved. To use a non-cognitive case of extension for
comparison (see Menary 2010b, 14): it does not matter if a spider loses her web
once in a while, or perhaps even very often, as long as the however intermittent
production and use of webs provides part of the explanation for the presence of
web-making mechanisms in the spider population.
Third, and most profoundly perhaps, a historical account of functions will counter Mark Sprevak’s critique of classical machine-state functionalism (2009) which
targets the “Martian intuition” that an identical or analogous kind of function can
be realised by physically or biologically very differently constituted systems. On
that intuition, which is basically the Lewisian one (1980), if we can specify a set
of functions for certain traits and behaviours in human beings and Martians, we
should be able in principle to identify which concrete structures within or around
the human being or Martian will play what sort of causal role in accomplishing
those functions. However, Sprevak objects, one could easily imagine the conditions of persistence and reliability of the respective structures and the coupling
relations involved to be violated in manifold ways, and hence fail to meet the
specifications laid out by extended functionalism. Playing on the Martian intuition
in this way will probably kill the Martian – but, if we follow the historical account
of functions, it will not kill the intuition. Any coupling with whatever structure
that were as unreliable or otherwise insufficient or even freakish in nature as
described by Sprevak would militate against the imagined organism’s reproductive chances, and against the factual (rather than conceptual) possibility of his
existence. Conversely, Sprevak’s Martian may be imagined to be in command of
capabilities and enter into environmental coupling relations so unfamiliar that we
could not find any remote analogue in the specifications of human cognition. But
then, the entire functional analogy will not hold. After all, there might be hypercognitive functions to Martians that human beings are insufficiently equipped to
imagine in the first place.
Clark and Chalmers’s reservations against history diagnosed here become
explicit when they introduce a set of criteria that is supposed to be necessary and
sufficient for the function of internal or external resources as components of cognitive processes (1998, 17; see also Clark 2010b):
(1) The resource should be of reliable availability, and should be used on a
frequent basis.
(2) The information provided by that resource should be directly available.