Proverbio Árabe.
2.4 Galileo, Observando Manchas Solares
2.4.1 La Elipse En La Astronomía Leyes de Kepler
To understand users’ response for a given retrieval system, it requires monitoring users’ information seeking behavior while performing a search task. There seem to exist two prominent methods used for analyzing user behavior: (1) Experimental method and (2) Log analysis. An experimental method is a lab based user study, which involves users (participants) to perform specific search tasks with the retrieval system being tested. While during log analysis, users’ interaction with the system over a time period is
2.4. User Behavior recorded for analyses.
2.4.1.1 Experimental method
Answers to certain research questions can be obtained through experimental method. For instance: Do people in Glasgow drink more during winters than in summers? Do people become crazy after writing PhD thesis? Experimental methods or laboratory studies allows researchers to examine relationship between two or more variables. Dur- ing an experimental method, environment is deliberately manipulated to ascertain what effect one variable has on the other [Field and Hole,2003].
In experimental method, answers to research questions are obtained by observing the “cause” and “effect” of certain factors. For instance, to answer research question: Does listening to music reduces stress? Here, the cause (music) can be measured by observing its effect (stress level).
Experimental studies are often conducted in labs so that the researcher has the maxi- mum control over the experimental situations. Advantage of conducting a laboratory studies is that, it offers explanation for the observed behavior [Kelly,2009].
In past, many researchers have conducted experimental studies to observe effects of cer- tain features of retrieval system on users’ information seeking behavior (For example, [Joho and Jose,2008;Villa et al.,2009;Thomas et al.,2010;Hopfgartner et al.,2010]). However, the experimental methods also has certain limitations, and they are primarily due to the fact that these methods typically involve subjective elements, small number of participants, and have large variance in the results due to sampling bias. It is hard to generalize the results from experimental methods due to the small number of participant [Zhang et al.,2009].
Furthermore, experimental methods have limited benefits because it does not provide intervention-free studies of user-initiated searches, where no observers were present to monitor end-user searching, no researchers assigned tasks to end users, and no inter- viewers questioned users before, during, or after the search [Markey,2007a].
Nevertheless, experimental studies are good with respect to the amount of control re- searchers have over the study situation. This is useful when trying to isolate the impact of one or more variables [Kelly, 2009]. Next, a naturalistic method (log analysis) to investigate users’ information seeking behavior is described.
2.4. User Behavior 2.4.1.2 Log analysis
Search engine companies and the researchers often use transactional log to understand the user behavior during information seeking process (For example, [Spink et al.,2002;
Jansen and Spink, 2005;Agichtein and Zheng, 2006;Jansen, 2006;White and Morris,
2007a;Duarte Torres et al.,2010]).
A search engine’s transactional log is an electronic record of interactions that have oc- curred during a searching episode between a Web search engine and searchers who are seeking information on that Web search engine. Therefore, the use of data stored in transaction logs of Web search engines, Intranets, and Web sites can provide valu- able insight into understanding the information-searching process of online searchers [Jansen,2006].
Although, transactional log data continues to be useful in uncovering search behavior patterns and is used by many researchers and search engine companies, there exist some limitations and social issues associated with it.
[Kurth,1993] identified the limitations of the transactional log analysis and suggested that, although transaction log data effectively describe what searches patrons enter and when they enter them, but they do not reflect, except through inference, who enters the searches, why they enter them, and how satisfied they are with their results. Further- more, [Murray and Teevan,2007] reported following social and technical challenges of transactional log analysis:
• Can query log data be safely collected and analyzed? Should it? • Can query log data be anonymized and shared? How shall it be done? • Can we establish standards of practice for query log analysis?
Key points and suggestions from the discussions for the above questions were that, using transactional log data is crucial for training researchers in this field. In addition, there is a strong potential for search engine companies to benefit from the findings of the studies from these log data. In order to make fair use of these data, setting up review boards and clear guidelines were suggested. Furthermore, preserving logs was suggested as it provides a reflection of the character, values, fears, hopes and desires of the people who issue them.
In-spite of the above limitations and challenges, transactional logs continues to be a valuable resource for analyzing user behavior and search patterns. It allows to learn
2.4. User Behavior
how end users search online IR systems left to their own devices and unaffected by potential biases such as the presence of an observer, their knowledge that a reviewer would scrutinize their search at a later time, their aptitude for or the potential biases of a researcher-assigned task [Markey,2007a]. In the following Sections findings of user behavior analysis from experimental methods and log analysis are presented.