Please tell us what you think of the Bulletin interactive pdf!  Feedback

Bulletin, April/May 2008

IA Column
Exploring Exploratory Search

by Mark Nolan

Mark Nolan is an information architect in Washington, D.C. He can be reached at mc_nolan<at>

An expert’s information search goals are simple:

  1. Show me something I don’t already know. 
  2. Among the 500 + n³ or so articles published last month in or around my area of expertise, rank the ones I should read first.

Expert users already have proficient, polished search skills. It doesn’t take them long to learn how to tap the value of an information system. They have a knack for “gaming” a system (in a positive manner) to generate productive results. In the process they are quick to determine if the system may be a valuable resource for what they seek or whether they need to find another resource. Moreover, they tend to be highly educated and smart enough to resist falling into any predetermined cognitive style. If one approach doesn’t work, they find another. 

Working with power users on several projects provided my teammates and me the catalyst to investigate the latest innovations in search research. What we discovered was exploratory search.

Experiencing Information through Search
Research underway at leading institutions is breathing new life into our understanding of how we search for information. Exploratory search is an emerging discipline that encompasses the examination of dense or rich information domains via three main facets: “Lookup,” ”Learn” and “Investigate.” Researchers at Microsoft Research, Johns Hopkins University and other leading institutions have discovered that relationships are not linear among these facets, nor is there a definitive parent-child relationship in how we explore information. That is, we do not browse, learn and investigate in a linear process as we, say, watch a movie. For example, behaviors within the “Investigate” activity may precede those from “Learn” or “Lookup,” while activities under “Investigate” may be paired with one or more from “Learn.” Rather, it is the combination and recombination of the different relationships in how information is presented and how we seek for it that holds the potential for expediting new discoveries. It is these very relationships that have the potential to produce the “Ahas” or exclamations of “Eureka” by researchers. The goal for information architects is to support the user’s ability to build relationships among discovered information items [1]. In the design of search results and interfaces for browsing rich information resources we need to design a certain degree of elasticity into the product to give users more control over the results.

In his article “Exploratory Search: From Finding to Understanding,” Gary Marchionini outlines three main facets of search behavior – “Lookup,” “Learn” and “Investigate” – and further illustrates the differences among the three modal search states and makes the case for a new approach to researching search [2]. Each of the three modal search states has associated behaviors and sub-modes. 

  • Under “Lookup” Marchionini proposes the following: fact retrieval, known item search, navigation, transaction, verifications and question/answering.
  • Under “Learn” we find knowledge acquisition, comprehension/interpretation, comparison, aggregation/integration and socialization. 
  • Under “Investigate” the group contains accretion, analysis, exclusion/negation, synthesis, evaluation, discovery, planning and forecasting, and transformation. 

In “Lookup” mode, many of us locate people, places and things every day on Yahoo, Google or Ask. We know or suspect the thing we’re looking for is out there – we just have to find the right combinations of search terms to locate it. In the process, we may “Learn” about new items, relationships among people, places and things, provided that a reasonable portion of the domain is well organized (and well managed). The success rate on “Learn” is inconsistent, however, because it is hard to plan and manage. Efforts to date assume a logical transition from “Lookup” to “Learn” in a user’s search behavior. What the research challenges is the assumption that to get to “Learn” users must start with “Lookup.” The research shows that in fact users may begin in “Learn” mode and transition to fact retrieval and verification under “Lookup.” 

Supporting Exploratory Search
“Investigate” is the most challenging of the three primary tasks to work on and the crux of our design challenge. As the authors of “Supporting Exploratory Search” put it, “What do we do if we want to locate something from a domain where we have a general interest but no specific knowledge?” [3]

It’s a good question. In other words

  • How do we discover what we don’t know how to find? 
  • How do we locate information beyond our area of expertise that may be important to us? 
  • How do we browse topics and search resources for unknown quantities that we cannot define or even describe? 
  • How else do we "Learn?”

To continue our scenarios, users may begin in “Investigate” mode, where they are seeking to exclude sources, data and information from a task, as does our expert researcher cited above. They then proceed to “Learn” once they have isolated a series of articles they suspect will help them keep current in their field. Imagine being able to isolate articles for our expert persona that use terms he or she commonly searches on but which are used in a slightly different context. For instance, a chemist may research published results of experiments that use similar methods or materials, though in different disciplines such as microbiology and molecular physics. Matches may generate leads that may hasten collaboration of research, leading to important discoveries in the parallel disciplines. Today, collaboration across disciplines is staggeringly important in research on subjects such as the human genome, intelligence and materials science. 

There is still an enormous amount of work to be done in improving the execution of designs and search solutions that will help lead the user to a greater understanding and investigation of a topic. For instance, information architects working on solutions associated with distance learning online instruction may find themselves butting up against the challenges of creating pre-coordinate information displays to facilitate “Learn” and “Lookup.” 

Peter Morville’s prescient observation that “we do not realize the potential of what we can do until we discover the possibilities” [4] haunts us every day. The observation cuts right to the heart of the challenge in designing information architecture solutions for experts: designing solutions to facilitate “Investigate,” “Learn” or “Lookup.” To date, search engine technology remains the only viable option for supporting explorations of rich information collections, particularly ones with a high volume of contributed content. 

Improvements in search engine technology are offering new opportunities. Search engine technology is the next frontier in the quest to bridge the gap between data mining tools, translation techniques and etymology. Improving search engine technology by applying findings from exploratory search research has already begun. Some efforts have proved successful while others have generated inconsistent outcomes to date, but work is underway to improve results. For instance, engineers are adding different usage metrics to the mix to sharpen rankings for relevancy. In the near future, the system will not only know what was already searched on (as some search engines do today), but also whether to show that information to the users, given the context of the their searches, their search patterns and their history of adding or eliminating terms based on their review of prior search results. More to the point, emerging search technologies such has Twine and Readware have the power to explain why certain results were produced and give the user the option to adjust, tweak, expand or winnow the criteria on-the-fly to a degree unprecedented in the past. These new tools are transforming search engines into engines of discovery. New solutions will open up reams of information resources, reducing the time it takes to “Learn” and “Investigate” a topic.

Information architects influence the production of content. Metadata is helpful when and where metadata can be applied or when it is part of a formalized production process. However, information is often not formally processed. The level of inconsistency in the production and capture of supplemental information provided in a proof for a research experiment, for example, is astounding in this day and age. It is in these small corners of rich information domains that important discoveries may sit for months, even years.

Rich information domains vary in structure and organization. Nevertheless, they have the equivalent of fingerprints – they produce their own heuristics, however inconsistent. Some resources reflect the way information is captured or stored; others may reflect a process or technique used for capturing the information, such as a publishing process. Yet others reflect the disposition of the professionals who organized them or built the appliance that captures the information. The means of production of content directly affects a system’s ability to locate, draw meaning from or build relationships among its contents. 

To enable exploratory search, the IA community needs to exert more influence in the design and development of the solutions that capture, originate and produce content. For example, at the object level, in the planning stages, information architects could effect efficiencies by providing requirements for how information should be managed at its creation. We have data and data architects; we have information and information architects – and we have confusion in the industry on the role each plays in the design and development of systems. This situation needs to end. 

Exploratory search research opens new opportunities for information architects to grow the profession. Research discoveries are unleashing profound enhancements in search engine technology. To increase the value of the findings of the research on exploratory search, information architects need to explore new methods and approaches to designing information displays for expert systems. As professionals, we need to exert more influence over the devices, appliances and software that govern the production of content. Enhancements to the production of content will cyclically increase our understanding of designing information solutions for experts, adding additional value to the user interfaces designed for less expert solutions.

Resources Cited in the Article
[1] Gersh, J., Lewis, B., Montemayor, J., Piatko, C., & Turner, R. (2006, April). Supporting insight-based information exploration in intelligence analysis. Communications of the ACM, 49 (4.), 63-68. Abstract and references available at Full electronic text available to subscribers or for purchase.

[2] Marchionini, G. (2006, April). Exploratory search: From finding to understanding. Communications of the ACM, 49(4), 41-46. Abstract and references available at Full electronic text available to subscribers or for purchase.

[3] White, R.W., Kales, B., Ducker, S.M., & Schraefel, M.C. (2006, April). Supporting exploratory search. Communications of the ACM, 49(4), 36-39. Abstract and references available at Full electronic text available to subscribers or for purchase.

[4] Torenvliet, G. (2007, May + June), Review of "Ambient Findability by Peter Morville," O'Reilly Media, 2006; ISBN 0-596-00765-5; $29.95. Interactions,14(3), 50-51. Electronic text available to subscribers or for purchase at