Detecting Lost in Information Retrieval Based on Cursor Movement Analysis
Kotomi Higuchi (AY 2013)
In our daily lives, it is common to use the Internet as a means of searching for something, and search engines play an important role. However, it is not always easy to find the desired information when using search engines. In this situation, Kuhlthau (1993) defined the Information Search (ISP) model, which divides people's information-seeking behavior in search engines into the first to sixth stages, and shows the process of determining the necessary information as the stages progress, as well as the emotions involved in each stage. The ISP model has uncertainty, or "hesitation," and the detection of this "hesitation" is important for the development of adaptive systems. Based on previous studies that have detected "hesitation" in Japanese-English translation tasks, this study focuses on the number of round trips of the mouse cursor and aims to detect "hesitation" in information search query formation from navigation and information tasks, which are the main information search behaviors of humans. The experiment was conducted using a single topic and a single round-trip.
The experiment consisted of 6 topics, with 2 informational and 2 navigational tasks per topic, for a total of 24 information search sessions. First, participants were given a piece of paper with a story, a question about the story, and a query to search for the answer to the question. If they could not predict the results, that was okay, and if they had an idea, they were asked to look at the screenshot of the search results with that query and think of a new query that would find the answer to the question they thought would be better while looking at the search results list page. They were also asked to complete a post-test questionnaire after completing each task.
The results of the analysis showed that the number of round-trips was slightly higher for the Informational task than for the Navigational task, but there was not much difference. The results also showed that there was no correlation between the number of cursor round-trips and each task (navigational and informational tasks, with and without hesitation), while the combination of the two attributes (navigational with and without hesitation, informational with and without hesitation, and navigational with and without hesitation) showed no difference, informational with and without hesitation), it was found that in the navigational with hesitation task, the number of cursor round-trips was lower when there were two elements of hesitation: "I could not decide which keyword is the best" and "I do not think that the keyword I considered will produce a good result". We found that the number of cursor round-trips was low.
As a future direction, by changing the device to an eye-tracking device and comparing eye movements under the same experimental conditions, it will be possible to investigate why the number of cursor round-trips is low when the above hesitation factors are present in the navigation task with hesitation. We can also compare the results with those of this study to see the relationship between cursor and eye movement.
(Translated by DeepL)