Uncertainty in Information Architecture

Below are excertps from a longer essay I wrote several years ago that never got published, mixed with a passage from Designing Web Navigation. In thinking about breadth vs. depth recently, I returned to this line of thinking and wanted to share it.

——————————————-

Uncertainty in Information Seeking

Formal associations of information and uncertainty date back to Shannon and Weaver (1949), where the presentation of information itself was believed to reduce uncertainty. Later, Nicholas Belkin (1980) focused on the notion that seekers, sometimes even experts in a given information system, are not able to properly formulate queries to access the information they need. He calls this “anomalous states of knowledge,” or ASK for short. Here, uncertainty underlies the basic information seeking.

Kuhlthau’s (1993) work on uncertainty and information seeking is perhaps the most extensive. She proposes uncertainty as a principle for information seeking, defined as follows:

“Uncertainty is a cognitive state that commonly causes affective symptoms of anxiety and lack of confidence. Uncertainty and anxiety can be expected in the early stages of the ISP. The affective symptoms of uncertainty, confusion, and frustration are associated with vague, unclear thoughts about a topic or problem. As knowledge states shift to more clearly focused thoughts, a parallel shift occurs in feelings of increased confidence. Uncertainty due to a lack of understanding, a gap in meaning, or a limited construct initiates the process of information seeking” (Kulthau, 1993, p. 111).

Wilson et al. (2002) explored the relationship between uncertainty and information seeking. They found that the Uncertainty Principle as outlined by Kuhlthau indeed serves as a useful variable in understanding and predicting information-seeking behavior. Though not conclusive, this research points towards uncertainty as a universal aspect of information seeking.

Uncertainty common in earlier stages is caused by the introduction of new information that stands in conflict with the user’s prior understanding of the material. Significant to this principle is a typical secondary peak of uncertainty in the process, or the “dip” in confidence mentioned above. Optimism at the beginning phases of information seeking often gives way to doubt and uncertainty in the middle phases. In other words, confidence does not lie on a steadily increasing linear scale, as previously believed, but rather can rise or fall as new information is uncovered. Acquiring more information in initial stages increases rather than decreases uncertainty.

Note, however, that it is the perception of complexity, rather than the actual objective complexity of a task, that causes feelings of uncertainty (Kuhlthau, 1999). Perceived complexity is often the cause of the secondary peak of uncertainty, doubt, and confusion in information seeking.

Uncertainty in Breadth vs. Depth

An example of information uncertainty can be found in the issue of breadth vs. depth of information structures, an important issue in web design. Researchers typically have studied search time, disorientation, error rates, and even satisfaction. A good summary can be found in Larson & Czerwinski (1998). The general design recommendation from such research is to increase breadth to reduce search time and errors, as well as increase satisfaction. It is believed that the time spent scanning menu items in a broader structure is less than the time spent drilling down into a deeper structure. In the latter, menu terms are necessarily more general and therefore more ambiguous.

Unfortunately, most breadth vs. depth studies test relatively symmetrical structures, for example 4x4x4 structures (Snowberry et al. 1983), and thus do not account for naturally occurring irregularities in hypertext shapes. One exception is a study by Norman and Chin (1988), in which constant structures were compared to irregular shapes (increasing, decreasing, convex, concave). The researchers found that the concave structure (8x2x2x8) performed best.

Michael Bernard (2002) more recently tested information structures with both symmetrical and asymmetrical schemes as well. He confirmed that broader structures do indeed perform better, but also found that deeper, asymmetrical structures perform better than symmetrical structures of the same depth. For example, 4x4x4x4 structures performed not only worse than asymmetrical shapes of the same depth (e.g. the concave 6x2x2x12) but also worse than deeper concave structures (e.g. 3x2x2x2x12). He concludes that the performance of the structures is determined in part by the properties of the hypertext shape, namely the perceived complexity of the information space and information uncertainty.

A concave information architecture indeed seems to match a decrease in certainty users often experience when seeking information as described by Kuhlthau. At the top level of a concave structure, seekers need orientation without being overwhelmed. A balance of well-selected, mutually exclusive categories serves as an efficient, satisfying starting point. The middle levels are best restricted in breadth, thus reducing uncertainty and feelings of doubt or frustration while making choices. The broader, bottom level of a concave structure, however, provides maximum information scent and a sense of “arrival” as the seeker begins gaining confidence again. As Bernard (2002) writes, “at the terminal level, broad menus reduce the information uncertainty.” At this point in the structure the users are able to handle more complexity.

Conversely, convex structures present more choices at the middle levels than on the ends (e.g. 2x8x8x2) and thereby contradict a normal pattern of cognitive and emotional user needs in information seeking: there is more uncertainty after navigating has begun. This could mean an increased likelihood of a hesitation in the search process, and feelings of apprehension and frustration may set in.

Therefore, why the performance of varying hypertext shapes is given by perceived complexity and uncertainty. This means that in evaluating or creating information architectures, affective considerations can play a potential role in predicting their overall success, namely feelings of uncertainty and confidence.

Uncertainty in the Scent of Information

Jared Spool and his colleagues (2004) have popularized the notion of the scent of information. Scent refers to how well links and navigation match a visitor’s information need and how well they predict the content on the destination page. There are potentially many aspects of navigation design that contribute to scent, including position on screen, labels, icons, color, descriptive texts, and so forth.

But ultimately scent is more complex and subtle than how links are displayed. It really has to do with creating a sense of confidence in navigating. The researchers explain:

Usually, however, scent is invisible. It is a product of how well the designers understand the site’s users, those users’ needs, and how the users access the site.

In fact, the best way to detect scent is to measure the users’ confidence…When the scent is weak, users are not confident at all. They doubt their choices. They tell us they are making “wild guesses.” They click hesitantly, hoping the site will magically come through for them. More important, they rarely find what they are seeking.

When scent is strong, however, their confidence builds as they draw closer to their content. They traverse the site with little hesitation. Moreover, they find what they are seeking.

Trigger words emerge to be the most critical aspect in creating information scent. These are navigation labels and texts that match a visitor’s need on the page. Discussed further in Chapter 5, labels are what people are scanning for when the first land on a page. Scanning for trigger words is a consistent pattern Spool and his team found across user types, across tasks, and across sites:

We’ve noticed that people looking for information all exhibit similar patterns. They first scan for their trigger words—words or phrases they associate with the content they’re seeking—in an attempt to pick up the scent.

Trigger words help to indicate they are on the right track. They reduce uncertainty and give confidence in navigating further.

While labels and trigger words certainly play a leading part in reducing uncertainty and providing a sense of confidence, I would also argue that the shape of the information architecture itself also plays a significant role.

References

1. Shannon, C.E., & Weaver, W. (1949). The mathematical theory of communication. Urbana, IL: University of Illinois Press.
2. Belkin, N. J. (1980). Anomalous states of knowledge as the basis for information retrieval. Canadian Journal of Information Science, 5, 133-143.
3. Bernard, M. L. (2002). Examining the effects of hypertext shape on user performance. Usability News, 4.2. Accessed online March 10, 2003 at http://wsupsy.psy.twsu.edu/surl/usabilitynews/42/hypertext.htm.
4. Kuhlthau, C. C. (1993). Seeking meaning: A process approach to library and information services. Norwood, NJ: Ablex.
5. Kuhlthau, C.C. (1999). The role of experience in the information search process of an early career information worker: Perceptions of uncertainty, complexity, construction, and sources. Journal of the American Society for Information Science, 50(5), 399-412.
6. Larson, K. & Czerwinski, M. (1998). Web page design: Implications of memory, structure and scent from information retrieval. Proceedings of the Assocaitions for Computering Machinery’s CHI 1998, 18-23.
7. Norman, K. L. and Chin, J. P. (1988). The effect of tree structure on search performance in a hierarchical menu selection system. Behaviour and Information Technology, 7, 51-65.
8. Snowberry, K., Parkinson, S., & Sisson, N. (1983). Computer display menus. Ergonomics, 26, 699-712.
9. Spool, Jared, Christine Perfetti, & David Brittan (2004). Design for the Scent of Information, User Interface Engineering.
10. Wilson, T. D., Ford, N., Foster, A., & Spink, A. (2002). Information seeking and mediated searching. Part 2. Uncertainty and its correlates. Journal of the American Society for Information Science and Technology, 53(9), 704-715.