Page 454 - Deep Learning

Notes to Pages 248–256 437

56. Brownell (1947) and Brownell and Moser (1949).
57. Regarding the amount of classroom time required for each subtraction with
regrouping, see Leinhardt (1987). Regarding estimates of learning rate, Kurt
VanLehn located 11 strategy changes in a 90-minute problem-solving effort,
indicating a rate of 1 learning event every 8 minutes (VanLehn, 1991). However,
VanLehn (1999) observed a much lower rate of learning in an analysis of 9 students
who spent an average of 4.5 hours studying a textbook chapter in physics. Only
28 learning events were identified in the students’ verbal protocols, which gives
the much lower rate of 1 event per 1.5 hours of studying. It is more likely that a
learning event fails to leave any trace in a verbal protocol than that a seasoned
researcher thinks he is seeing a nonexistent event, so the higher learning rate is
likely to be the one closest to the truth.
58. Researchers who work on intelligent tutoring systems (ITSs) form a commu-
nity of their own, distinct from the broader field of computer-based education.
ITS researchers have their own conferences and their own scientific journals.
The ITS field came into its own in the late 1970s; the book edited by Sleeman
and Brown (1982) was the defining document. Wenger (1987) reviewed the first
wave of work. See Woolf (2009) for a recent appraisal of the field. The technical
details of my own approach to ITS design are available in multiple research pub-
lications (Buchanan et al., 1995; Fossati, Di Eugenio, Brown & Ohlsson, 2008;
Langley, Wogulis & Ohlsson, 1990; Lu et al., 2007; Mitrovic & Ohlsson, 1999;
Mitrovic, Ohlsson & Martin, 2006; Ohlsson, 1986, 1987c, 1991, 1992d, 1992h,
1996b; Ohlsson & Langley, 1988; Ohlsson & Mitrovic, 2006, 2007).
59. The first paper on SQL-Tutor was Mitrovic and Ohlsson (1999). For descriptions
and analyses of constraint-based tutors, see Mitrovic, Ohlsson and Martin (2006)
and references in preceding Note 58 and Note 60 following.
60. See Mitrovic, Suraweera, Martin and Weerasinghe (2004) and Mitrovic, Martin
and Suraweera (2007) for general discussions of the database suite. SQL-Tutor is
described in Mitrovic, Martin and Mayo (2002), while Suraweera and Mitrovic
(2004) and Mitrovic (2002, 2005) cover database design and normalization,
respectively.
61. For empirical evaluations of constraint-based tutoring systems, see preceding
Notes 58 and 59. For a study of the micro-engineering of a tutor’s feedback mes-
sages, see Zakharov, Mitrovic and Ohlsson (2005).
62. Conway and Siegelman (2005).
63. Wiener (1948).

Chapter 8. Error Correction in Context
1. Reason (1990, p. 203).
2. From “Prologue: A Tale of Two Farms,” from Collapse: How Societies Choose to
Fail or Succeed by Jared Diamond, copyright © 2005 by Jared Diamond. Used by
permission of Viking Penguin, a division of Penguin Group (USA) Inc.
3. Chi, Glaser and Farr (1988), Ericsson (1996), Ericsson, Charness, Feltovich and
Hoffman (2006), Ericsson, Krampe and Tesch-Romber (1993), Ericsson and
Lehmann (1996) and Feltovich, Ford and Hoffman (1997).