Page 449 - Deep Learning

432 Notes to Pages 208–212

9. The current version of this micro-theory is the descendant of a prior theory that
was first published in Ohlsson (1987b). It was re-formulated in Ohlsson (1993a,
1996a, 1996b); see Ohlsson (2007b), Ohlsson, Ernst and Rees (1992) and Ohlsson
and Rees (1991a, 1991b) for various applications. The current version is augmented
with auxiliary hypotheses about how the error correction mechanism interacts
with other learning mechanisms (Ohlsson, 2008a).
10. Allwood (1984), Allwood and Montgomery (1982) and Anderson and Jeffries
(1985).
11. See Ohlsson (1994) for the origins and definition of this distinction. It was intro-
duced into cognitive science by Anderson (1976, pp. 116–119) and Winograd
(1975). It is ultimately rooted in common sense: There is a clear subjective differ-
ence between knowing that something is the case and knowing how to perform a
certain type of task. Behaviorally, the distinction is expressed in the fact that the
ability to recite from memory a description of a desired task performance does not
automatically bring with it the ability to produce that performance without prior
practice (“In order to land a 747 airplane, …). Neves and Anderson (1981) pro-
posed a computational theory of the translation from a verbal recipe to an execut-
able skill and its application in the domain of geometry. Neuroscience research
also supports the distinction between declarative and practical knowledge (Squire,
1987). Indeed, the distinction has turned out to be helpful in understanding the
patterns of impairment and ability in diverse classes of patients with brain damage
(Eslinger & Damasio, 1986; Glisky, 1992; Timmerman & Brouwer, 1999).
12. Philosophers have explored various ideas about truth, including that truth is the
relation between an assertion and a material state of affairs (the correspondence
theory, see O’Connor, 1975); that truth consists in an assertion cohering with all
other assertions already accepted as truth (the coherence theory; see Walker, 1989);
that truth is whatever supports successful action (the pragmatic theory; see Smith,
1978, and Thayer, 1982); and others as well (see, e.g., Kirkham, 1992). The prolifera-
tion of theories is sufficient evidence that philosophers do not yet have a theory of
truth: “The problem of how to relate truth to human desires, beliefs, intentions,
and the use of language seems to be the right one to concentrate on in thinking
about truth. It also seems to me this problem is not much nearer a solution today
than it was in [John] Dewey’s day” (Davidson, 1990, p. 280). Because the attempt at
explication has been going on for at least 2,500 years, the lack of success becomes
an indictment of the enterprise and suggests that the question is not framed well.
13. Evans (2007), Johnson-Laird (1999) and Shafir and LeBoeuf (2002). But see Rips
(1994), especially Table 5.1 (pp. 151–152), for an example of a logic-based theory
that predicts certain aspects of logical reasoning with high accuracy.
14. Perkins (1981, pp. 128–129).
15. The dissociation between action and judgment, generation and evaluation, has
been observed in a wide variety of behaviors, from pushing buttons in a speci-
fied sequence (Kay, 1951) to speech (Mattson & Baars, 1992; Postma, 2000). The
developmental experiments regarding number knowledge were done by Gelman
and Meck (1983): “The range of set sizes to which the children know that these
principles [of counting] apply is much greater than the range that they can suc-
cessfully count” (p. 357).