Paper Reading #23

Comments:
Comment #1
Comment #2
Reference:
Title: Evaluating Automatic Warning Cues for Visual Search in Vascular Images
Author: Boris W. van Schooten, Betsy M.A.G. van Dijk, Anton Nijholt, Johan H.C. Reiber
Venue: IUI 2009/2010
Summary:
This group conducted an experiment using MRA images to analyze the effects of visual searches. Namely, the aim was to determine the level of detriment or effeciency in a false positive, false negative, no suspicious areas, and perfect detection situation. The users were presented with images of vessels, and they were supposed to verify the automatic segmentation. They "compared user time and error performance as well as subjective preference for the following conditions: no warning highlights, only false positives (paranoid), only false negatives (conservative), and perfect highlighting." The results indicated that the users performed significantly faster with the paranoid highlighting than with no highlighting. On top of that, they made less errors as well, although by a relatively insignificant margin. Lastly, the group found that the users preferred the presence of suspicious areas over no suspicious areas, and "appear" to prefer paranoid over conservative highlighting. This is very different than what has been found in other similar studies. Other studies showed detrimental effects for false positive rates approximately the same as in this study, but this experiment indicates some potential positive effects for those false positives. Figure 1 shows an example from the experiment.

Discussion:
This was my first or second short paper, and I basked in all its glory. The experiment was easy to understand, so I think the author did a good job explaining it. I feel sorry for the people who had to participate in the study because this seemed like a pretty boring experiment to take part of. Since visual search is so common, whether it be in airports, medical fields, or even the military, I can see how this type of study could be useful. I would be interested in seeing if other groups could replicate these results on larger scales and with different applications.