60 Informational environments
not unequivocal enough to determine that regularity, or r as an expression of that
regularity with certainty. Intervening factors will be held accountable for blurring the signal but are either of random nature or too many and too complex to
actually be accounted for. What can be achieved instead is partial, approximative
confirmation or refutation of a hypothesis within the bounds of available data and
applicable methods. NI-2c covers this sort of case in particular, with the possibility that NI-2b may apply, too.
For example there is the possibility that someone contracts an infectious disease, let us call it u-fever, while lacking the key symptoms of the illness, which
might be called u-rashes, or even remaining asymptomatic. The condition of s
being F will then be fulfilled, and I-regularly so, but r does not always obtain,
so that false negatives obtain. This is a frequent case in medical practice, and
it belongs to the realm of possibilities that, like the possibility of indiscernible
symptoms with variant aetiologies, educate any responsible physician to being
a careful probabilistic reasoner.
1
If u-rashes could be caused either by u-fever or,
say, morbus y, that symptom by itself would not provide information on the underlying condition within the context of the reference class relevant to the physician.
If, however, the u-rashes symptom occurs only intermittently but in rather exclusive connection to u-fever, and if the proper aetiology of that symptom is in place,
and if the intermittency of the symptom can be ascribed in contextual conditions,
u-rashes will count as a signal of u-fever, and hence provide natural information
about the presence of u-fever in the patient. With respect to intermittency, the
uptake of additional information about s by R, taken up by different means, will
help to reduce uncertainty on the receiver’s side as to whether the illness is present
in the absence of that typical signal.
However, most relevant to the present context are situations of underspecification that occur in perception in a natural environment. On the one hand, conditions in the ambient optic array, to recur to Gibson’s terminology, are specific in
terms of the laws of reflection, refraction and spatial geometry being in place.
On the other hand, visual information may become masked, blurred or otherwise
distorted, or conditions are such that some of the information present will not be
taken up at all. This is the case for NI-2c in perception in natural environments.
Alternatively, some of the information may be reliable only in certain places, at
certain times. This is the case for NI-2b in perception in natural environments.
In reversed order of argument to the previous set of examples, NI-2b is the paradigm case of underspecification in natural perception, where NI-2c will often also
apply. The reference class for the signals involved will not comprise probabilities
near the Dretskean condition of unity in NI-2c cases. Under NI-2b, they might but
need not locally approach it.
In view of this issue, and along the lines of Millikan’s critique of Dretske, Matthieu de Wit et al. (2015), together with Rob Withagen and Anthony Chemero
(2009) suggest a ‘softer’ notion of natural information as the biologically more
realistic one: “epistemic contact” with the environment, in Withagen and Chemero’s wording, typically comes in degrees. Perfect contact, that is full specification,
is practically unattainable for organisms under natural conditions of perception,
not unequivocal enough to determine that regularity, or r as an expression of that
regularity with certainty. Intervening factors will be held accountable for blurring the signal but are either of random nature or too many and too complex to
actually be accounted for. What can be achieved instead is partial, approximative
confirmation or refutation of a hypothesis within the bounds of available data and
applicable methods. NI-2c covers this sort of case in particular, with the possibility that NI-2b may apply, too.
For example there is the possibility that someone contracts an infectious disease, let us call it u-fever, while lacking the key symptoms of the illness, which
might be called u-rashes, or even remaining asymptomatic. The condition of s
being F will then be fulfilled, and I-regularly so, but r does not always obtain,
so that false negatives obtain. This is a frequent case in medical practice, and
it belongs to the realm of possibilities that, like the possibility of indiscernible
symptoms with variant aetiologies, educate any responsible physician to being
a careful probabilistic reasoner.
1
If u-rashes could be caused either by u-fever or,
say, morbus y, that symptom by itself would not provide information on the underlying condition within the context of the reference class relevant to the physician.
If, however, the u-rashes symptom occurs only intermittently but in rather exclusive connection to u-fever, and if the proper aetiology of that symptom is in place,
and if the intermittency of the symptom can be ascribed in contextual conditions,
u-rashes will count as a signal of u-fever, and hence provide natural information
about the presence of u-fever in the patient. With respect to intermittency, the
uptake of additional information about s by R, taken up by different means, will
help to reduce uncertainty on the receiver’s side as to whether the illness is present
in the absence of that typical signal.
However, most relevant to the present context are situations of underspecification that occur in perception in a natural environment. On the one hand, conditions in the ambient optic array, to recur to Gibson’s terminology, are specific in
terms of the laws of reflection, refraction and spatial geometry being in place.
On the other hand, visual information may become masked, blurred or otherwise
distorted, or conditions are such that some of the information present will not be
taken up at all. This is the case for NI-2c in perception in natural environments.
Alternatively, some of the information may be reliable only in certain places, at
certain times. This is the case for NI-2b in perception in natural environments.
In reversed order of argument to the previous set of examples, NI-2b is the paradigm case of underspecification in natural perception, where NI-2c will often also
apply. The reference class for the signals involved will not comprise probabilities
near the Dretskean condition of unity in NI-2c cases. Under NI-2b, they might but
need not locally approach it.
In view of this issue, and along the lines of Millikan’s critique of Dretske, Matthieu de Wit et al. (2015), together with Rob Withagen and Anthony Chemero
(2009) suggest a ‘softer’ notion of natural information as the biologically more
realistic one: “epistemic contact” with the environment, in Withagen and Chemero’s wording, typically comes in degrees. Perfect contact, that is full specification,
is practically unattainable for organisms under natural conditions of perception,