Chapter 6: Conclusions

In general, the reported experiments found support for the HuWI cycle. Recall the three primary hypotheses derived from the cycle were:

• H₁: If SK directs interaction on a website, then the paths taken through the hyperlink structure will be predicted by the SK (Experiment 1).

• H₂: If the information sampled is goal-relevant, then the type of information sampled (content or structure) while interacting with a website, and subsequently recalled, will be predicted by goals (Experiment 2).

• H₃: If the goal-relevant information sampled from a website modifies SK, then information will be found more efficiently as experience using a website is gained (Experiments 1) and more goal-relevant than goal-irrelevant information will be in SK (Experiment 2).

The following sections summarize the findings of Experiments 1 and 2 with respect to each of these hypotheses.

System Knowledge Directs Interaction on a Website

Experiment 1 demonstrated that participants’ interactions with the website were directed by SK. When SK was consistent with the website, participants located information more efficiently than they did when SK was inconsistent with the website. 

In addition, each replicate of the same task resulted in finding information more efficiently. Therefore, if it is assumed that participants modified their SK for the website as they used it, then their SK directed their interaction more efficiently in each replicate.

One unexpected finding of Experiment 1 was that the initial SK (i.e., that which was learned first) appeared to have a long lasting influence on directing interaction. Specifically, finding information that was incongruent with the initial SK was less efficient than finding congruent information, even after experiencing four replicates of the exact same set of tasks. Apparently this persistency was the result of remnants of the old incongruent knowledge elements remaining in SK, which were occasionally activated in order to direct interaction.

Future research should investigate the persistent effects of incongruent knowledge further in order to quantify the changing influence of the incongruent knowledge remnants. This research would lead to a mathematical model for predicting web navigation when incongruent knowledge elements exist in SK.

Information Sampled from the Website is Goal-Relevant

Experiment 2 demonstrated that the information that was sampled from the website was goal-relevant when the information related to content. However, when the goal and information related to structure, there was no strong evidence to suggest that structure information was sampled. There are three possibilities for not obtaining an effect on the structure SK measures. First, the manipulation may have been too weak to expose any true differences. Second, the measure of structure SK may have been insensitive to structure SK. Finally, the effect itself may be so small that it is either difficult to measure or it is unimportant.

Therefore, the conclusion, based on the available evidence, is that goals may be able to predict the amount of content SK that is acquired, but there is no evidence to suggest that goals can predict the amount of structure SK that is acquired. In addition, there is no strong evidence to suggest that structure SK can be acquired at all, at least at levels that impact interacting with a website.

Because structure exists in websites (i.e., webpages are linked), and many authors have speculated that structure knowledge can be acquired (e.g., Edwards & Hardman, 1989; Kim & Hirtle 1995), it is curious that there is no current empirical evidence to suggest that it can be acquired. Nevertheless, due to the lack of evidence for the modification of structure SK, it is reasonable to eliminate structure SK from the cycle. However, as will be discussed in the next section, this decision could be challenged if it is established that structure SK is important for human-web interaction.

Goal-Relevant Information Sampled from a Website Modifies System Knowledge

Experiment 1 and Experiment 2 demonstrated that content SK can be modified to accommodate content information. However, there is no strong evidence to suggest that the same is true for structure. Due to the lack of evidence that structure SK is important for the HuWI cycle, the decision was made to remove structure information and structure SK from the cycle. If it could be established that it is possible to acquire structure SK, it would have to be demonstrated that structure SK is necessary for the HuWI cycle to function. Specifically, if it were possible to remove all structure information from a website would performance on the website be the same as it would be if it was included?

It would not be possible to investigate performance in navigating a website without structure (i.e., links and separate webpages would have to exist). However, if one could establish methods that would provide different levels of navigation cues then these issues might be investigated. For example, by providing structure aids such as sitemaps (i.e., graphical overviews of a website’s structure) could provide users with additional structural information. However, research on the effectiveness of sitemaps is debatable, since some studies have found that they help performance (Billingsley, 1982; McDonald & Stevenson, 1998) and others have found that they hinder performance (Dias & Sousa, 1997; Stanton, Taylor, & Tweedie, 1992). In addition, the issue of why sitemaps work in some instances and not in others has not been addressed sufficiently.

Farris et al. (2002) speculated that since users seem to only remember content information from a website, then it is content that serves as a surrogate mental representation for structure. Therefore, it is possible that sitemaps enhance navigation efficiency when content matches structure and that they hinder performance when content and structure are incongruent. Alternatively, Stanton et al. (1992) speculated that sitemaps hinder performance when they distract from the primary information finding task. In addition, they proposed that sitemaps work when they are used as a navigation aid and that they do not work when they are overly relied on to find information. However, this issue has not been directly investigated. Resolution of this sitemap issue might give insight into how structure SK can or cannot be included in the HuWI cycle.

Status of the HuWI Cycle

Overall, both of the reported experiments tended to support the HuWI cycle. Specifically, the available evidence suggests that SK directs interaction with the website (Experiment 1). The user’s interaction with the website then samples the available content information in the website, but this is obvious only when the goal relates to content (Experiment 2). Finally, the sampled information modifies the user’s content SK (Experiments 1 and 2), but apparently does not influence structure SK, as it was operational zed in the reported studies.

Given its implications for web design, and the omnipresence of the web in modern society, the HuWI cycle should continue to be refined as new evidence emerges. In addition, several useful models can be derived from the HuWI cycle. For example, a model of learning on the web could be developed by furthering our understanding the SK modification process. Similarly, a model of web navigation could be developed when the issues surrounding structure SK have been addressed. These models such as these will have the potential to improve the learning outcomes for distance education courseware and to enhance the user experience of websites.  

Document Last Updated December 31 1969 19:00:00.