Intelligent Tutoring Systems for Ill-Defined Domains - Philippe ...
Intelligent Tutoring Systems for Ill-Defined Domains - Philippe ...
Intelligent Tutoring Systems for Ill-Defined Domains - Philippe ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
5 Discussion<br />
Using Expert Decision Maps 78<br />
The experimental group was able to view and interact with the expert decision maps<br />
with minimal instruction and was better able than the control group to arrive at the<br />
correct diagnosis. It should be noted, however, that the experimental group<br />
outper<strong>for</strong>med the control group <strong>for</strong> Case 1 (be<strong>for</strong>e viewing any of the maps). It is<br />
thus not clear how much of their superior per<strong>for</strong>mance can be directly attributed to<br />
their interactions with the maps. Use of the maps did however seem to be associated<br />
with increased confidence in a correct diagnosis. For correct diagnoses, the<br />
experimental group became more confident over time, while the control group was<br />
less confident <strong>for</strong> Case 3 than <strong>for</strong> Case 2. There was only one incorrect diagnosis<br />
from an experimental group student <strong>for</strong> either Cases 2 or 3, which was submitted with<br />
low confidence (.40), while there were 4 incorrect diagnoses from the control group<br />
<strong>for</strong> these two cases. These 4 diagnoses were submitted with relatively high average<br />
confidence levels (.75 <strong>for</strong> Case 2 and .72 <strong>for</strong> Case 3), suggesting that the control<br />
group students were less able to assess their own per<strong>for</strong>mance. Further analysis of<br />
how often students who submitted the 'incorrect' diagnosis updated their confidence<br />
levels may assist us in identifying specific misconceptions, which can then be<br />
addressed by providing appropriate feedback in the context of the activity.<br />
The case questionnaire responses demonstrated a great deal of variability in the<br />
students' baseline reflection and self-assessment abilities. Previous research has<br />
shown that it is difficult <strong>for</strong> students to reliably assess their current level of<br />
knowledge and skill [37, 38], but also that these capabilities can be developed with<br />
appropriate training [39]. In general, the response to the case questionnaire after<br />
Cases 2 and 3 show less evidence of reflective comments than the Case 1<br />
questionnaire. We suspect that the combination of solving three cases and completing<br />
three case questionnaires was overly ambitious within a two hour time period.<br />
Additionally, minimal instruction was provided to the students as to what types of<br />
self-reflective comments were expected. This issue is confounded, however, by the<br />
fact that Cases 2 and 3 were perceived by both groups as markedly less difficult than<br />
Case 1. Perhaps these students were not as motivated to per<strong>for</strong>m self-reflection<br />
activities when they were more confident in their per<strong>for</strong>mance. Further analysis of<br />
the relationship between confidence, per<strong>for</strong>mance and quality of self-assessment may<br />
help us to untangle these issues, as well as streamline our experimental design <strong>for</strong><br />
future studies.<br />
The visual representations seemed to make some students in the experimental<br />
group more aware of diagnostic reasoning as a process rather than an outcome.<br />
Though there were number of non-responses to this question, this may be attributed to<br />
the fact that the interaction with the expert decision maps was an additional task that<br />
the experimental group per<strong>for</strong>med, as compared to the control group. This task took<br />
both additional time and required additional cognitive resources, both of which may<br />
have impacted on the group's ability to provide evidence of ‘deep’ reflection within<br />
the case questionnaire. In addition to communicating expectations to both groups<br />
about the amount and level of reflection expected, future studies will need to ensure<br />
that tasks <strong>for</strong> both conditions are designed to be equivalent in terms of both time and<br />
difficulty. Further studies can then investigate how self-assessment and reflection can<br />
be better scaffolded within the environment itself.