|
|
|
-- Blackmon, Kitajima, Polson, & Lewis (2002)Blackmon, M.H., Kitajima, M., Polson, P.G. & Lewis, C. (2002). Automated Cognitive Walkthrough for the Web (AutoCWW). A position paper presented at the workshop on Automatically Evaluating the Usability of Web Sites in CHI'2002.
Automated Cognitive Walkthrough for the Web (AutoCWW)
How AutoCWW Can Promote Universal Usability of Websites Press reports and government studies describe the enormous potential of the World Wide Web to serve the information and learning needs of a broad spectrum of citizenry, including educational outreach in such areas as basic skills tutoring or medical/health information. These hopeful visions, however, are compromised by evidence suggesting that relatively high levels of education are a prerequisite for navigating and comprehending informational/instructional texts on current-generation complex websites. The Pew Internet Project (2001) survey found a dramatic, two-to-one disparity in Internet use between persons who never attended college and those with less education. Among American citizens with a high school education or less, only 37% use the Internet. In stark contrast, 71% of Americans who have completed some college use the Internet, and that percentage rises to 82% for the subset who have completed a college or graduate/professional degree. Three goals for closing the digital divide emerged from the ACM Conference on Universal Usability 2000 Conference, and our research group is addressed the goal of "bridging the gap between what users know and what they need to know" (http://www.UniversalUsability.org/definition/introduction.html). Many citizens have difficulty navigating websites because of low reading ability, insufficient background knowledge, and/or a cultural heritage differing significantly from the dominant culture. For the next three years I will be funded half time by NSF to conduct research that contributes to closing these knowledge and skill gaps by extending the Cognitive Walkthrough for the Web (CWW). Web developers have trouble predicting the behavior of such users, because developers do not themselves experience difficulties navigating complex websites to find information. Thus, the AutoCWW fills a role not filled by other available automated usability evaluation tools. Both CWW and the AutoCWW rely on Latent Semantic Analysis (LSA) to (1) objectively estimate the degree of semantic similarity (information scent) between representative user goal statements and heading/link texts on each web page, and (2) assess the probability that a particular user group can effectively learn from texts they discovered by navigating the website. Under NSF funding I will conduct a series of experiments that will first parameterize the CWW for college-educated users navigating informational/instructional websites and then parameterize the CWW to validly simulate navigation behaviors of diverse user groups, including users with 6th-, 9th-, and 12th-grade reading levels in English. This is feasible because LSA currently has semantic spaces representing these levels of educational attainment for American-educated individuals. Carrying out CWW analyses by hand is time-consuming, particularly if a web developer wants to predict the navigation behavior of several diverse user groups for a wide variety of user goals. There is only one realistic solution to this: automating the CWW to create the AutoCWW. Latent Semantic Analysis (LSA) is a key component of the AutoCWW, because LSA already a computer-based and because it is possible to program submission of texts to LSA, automatically generating predictions of which heading a user will focus on and which link within that heading a user will select when pursuing a given goal.
Full Texts |