Features of Computerized Multimedia Environments that Support Vicarious Learning Processes - New Science of Learning - page 61

acquisition, free-recall questions on topics presented in dialog and monolog formats
were administered.
Results yielded a significant difference in favor of the dialog condition as
compared to the monolog only among learners in the deep questions condition.
Differences between dialog and monolog were small in the other conditions, slightly
favoring topics in the dialog format in one condition and the monolog in the other.
Clearly then, deep questions included in the dialog, rather than questions per se or
simply presenting dialog, accounted for the difference in favor of the dialog con-
dition in Exp. 1 and in Craig et al. (2000). Presenting deep questions prior to each
content statement is, of course, easily implemented in computerized multimedia
environments.
Findings lending indirect support for the conclusions of Driscoll et al. (2003)
were obtained by Lin et al. (2005). They presented three groups of college students
with an animation depicting the parts, locations, and functions of the human heart.
Learners in one group were presented with written descriptive advance organizers
along with static pictures prior to dynamic animations. Those in a second condition
received written shallow question organizers, along with the static pictures, prior
to the animations, and in a control group learners only viewed the animations. The
descriptive organizers specified parts and functions of the heart. The question orga-
nizers asked shallow questions such as “What is the heart muscle that controls...?”
(see Lin et al. 2005, Fig 4.2). As would be predicted by the Driscoll et al. (2003,
Exp. 2) findings, there were no differences between the three groups on any of
four tests used to assess learning. As Lin et al. (2005) pointed out, “The advance
organizers cued the learner to critical information but did not create the neces-
sary environment for students to expend the effort needed to rehearse, internalize,
and synthesize information in meaningful ways necessary for higher-order learning
to occur.” In terms of the Driscoll et al. (2003) conclusions, neither the descrip-
tive organizers nor the shallow question organizers supported deep-level learning
processes.
Craig et al. (2006) conducted related research on deep questions. Learning gains
from several vicarious learning conditions were contrasted with those obtained by
learners in interactive tutoring sessions with an intelligent tutoring system, called
AutoTutor. AutoTutor yields learning gains of about 1.0-2.1 standard deviation
units when compared to various controls (Graesser et al., 2004; VanLehn et al.,
2007). AutoTutor helps students learn by holding a conversation in natural language.
Each topic begins with an information delivery , followed by a question presented to
the learner. An ideal answer to the question is decomposed into a set of key concepts
(sentences), called expectations . Latent semantic analysis (e.g., Graesser, Person, &
Harter, 2001; Landauer & Dumais, 1997) assesses the learner's progress by com-
paring their contributions to the content of each expectation. After all expectations
for the topic are covered, a brief summary is presented before AutoTutor moves on
to the next topic.
In the Craig et al. (2006) interactive tutoring condition there was interactive
tutorial dialog between the learners and AutoTutor for 35-40 min on 12 topics
concerned with computer literacy. The video and audio from each learner in this