Page 450 - Deep Learning

Notes to Pages 212–221 433

16. Botvinick et al. (2001) and Fiehler, Ullsperger and von Cramon (2005).
17. There is a branch of Artificial Intelligence research that builds constraint satis-
faction systems. However, the formal definition of a constraint proposed here
is different from the one that appears in those models, and constraints are
utilized in a different way in the error correction model than in typical con-
straint satisfaction system. See, e.g., Mackworth (1992), for a review from a
computer science point of view and Holyoak and Thagard (1989b) for an appli-
cation of the traditional constraint satisfaction concept to a cognitive model-
ing problem.
18. Allwood (1984, p. 415).
19. The hypothesis that cognitive processes are driven by anticipations and expec-
tations have been explored in several areas, including sensory-motor behav-
iors (Hoffman et al., 2007), text comprehension (Schank, 1982) and memory
(Neuschatz et al., 2002). It is possible to experimentally verify the generation of
expectations even in the eye movements of babies (Haith, Hazan & Goodman,
1988). Expectations regarding others obviously play an important role in social
interactions (Holmes, 2002).
20. Gensler (1987, p. 78).
21. Holmes (1992, p. 57).
22. Bell (2005), McCann (1978) and Thagard (1992, Chap. 3).
23. Brown and Burton (1978), Brown and VanLehn (1980) and Burton (1982).
24. The system does not have to be all that complex before diagnosis becomes dif-
ficult; see Burton (1982) regarding the issues in diagnosing errors in arithmetic.
The identification of programming errors (Johnson, 1986; Spohrer, Soloway &
Pope, 1985), the troubleshooting of complex equipment (Konradt, 1995; Lajoie &
Lesgold, 1992; Liu & Liu, 2001; Patrick et al., 1999a, 1999b; Patrick, James, Ahmed
& Halliday, 2006) and the diagnosis of sick people (Groopman, 2007; Norman,
2005) are more complicated.
25. Norman (1981), and Reason (1990).
26. Heckhausen and Beckmann (1990), Reason (1990, Chap. 3) and Senders and
Moray (1991, p. 89).
27. Laird and Newell (1993), Lenat (1983), Newell and Simon (1976) and Simon
(1990).
28. Displacement errors are closely related to the concept of capture errors, as intro-
duced by Norman (1981) and Reason (1990). The subtle difference is that the lat-
ter notion emphasizes the strength of a rule or action schema, while the notion
of displacement emphasizes its generality. Both concepts help explain how a rule
grabs control of action even when it should not.
29. Marcus et al. (1992).
30. For example, Anderson and Jeffries (1985) studied errors in problem solving due
to loss of information from working memory. Another malfunction view of error
derives them from repairs, attempts by the cognitive architecture to overcome
impasses that happen during the execution of a flawed, incomplete or inconsis-
tent cognitive strategy (Brown & VanLehn, 1980).
31. See Ohlsson and Rees (1991a, 1991b) for the original statement of the constraint-
based rule specialization algorithm.