Browse Books

Reviewer: Jeffrey B. Putnam

Nowadays, many of us spend a fair amount of time performing Web searches. We may be looking for products, better pricing, answers to medical questions-really, almost any sort of information. We have, of course, search engines to help, as well as online directories and portals. But how well do we actually do this This book is an attempt to formalize both information seeking behaviors and how well they perform under various assumptions about how the places we look up are structured. Primarily, the idea is to note the parallels between informavores (information-using agents) and the foraging behavior of various kinds of animals. Chapter 1 establishes the similarities between information foraging and food seeking, and serves as an introduction for the subsequent chapters. Chapter 2 covers foraging models in more mathematical detail; examples are included. Chapter 3 examines some of the structure of the Web and some experimental results of how people do searches. Chapter 4 discusses the notion of "information scent" in a Web context-that is, the use of cues, such as the text associated with hyperlinks, the structure of Web sites and pages, and similar indications of where a sought-after answer might be found. Chapter 5 discusses an artificial intelligence (AI)-style model, built in ACT-R, and compares the results from that model with those from humans involved in the same kinds of information-seeking activities. Chapter 6 considers an interaction technique called "scatter/gather" and analyzes how well this technique works. Chapter 7 considers the information foraging problem as a Markov model, and then uses this to model user interaction with a Web site. Chapter 8 discusses aspects of information gathering when groups are involved. Chapter 9 provides some Web design heuristics. It presents an overview of Bloodhound, a Web-based service that uses the Markov model to simulate user traversal of a Web site when the user is seeking specific information. Chapter 10 summarizes much of the book and discusses future work. The book is an interesting mix of mathematical analysis and data gathering. The models presented are certainly interesting, but it is not clear to me how effective they may be in the current, rapidly changing Web. Still, the better the models we can build, the better we can understand just what it is that we are building, including how it will be used when we build a Web site. This book certainly contributes to this knowledge base, making it a must-read for anyone interested in the field. The extensive bibliography should also prove very valuable, although it might have been better if it was placed at the end of the book, instead of at the end of each chapter, as many of the citations in different chapters come from the same works. Sometimes, the models do seem to come out of thin air, and their derivations are sometimes a bit sketchy. There are a few places where it is not entirely clear how the experiments performed relate to the material covered. In addition, the chapter on "scatter/gather" seems to be a bit of a tangent. It looks like an interesting way to do searches, but it is far from clear why any real Web user would want to use this and it isn't clear what it adds to the discussion. On the whole, this was a rewarding and very interesting-though occasionally puzzling-read. It should be useful to the more theoretically inclined Web designers and to researchers involved in how the Web is actually used. Online Computing Reviews Service

Reviewer: H. Van Dyke Parunak

A number of metaphors have been invoked to explain how to bring people together with the information they need. Information retrieval presumes that the information has been stored in an orderly structure, in which it can be looked up. Data mining views it as latent, scattered among irrelevant dross from which it must be extracted. Information farming expects that the information does not yet exist, but that it can be coaxed into existence by manipulating the institutions that have the power to generate it, as a farmer coaxes crops out of seeds, soil, sun, and water. Pirolli's innovative monograph introduces yet another metaphor, that of an animal foraging for food. Like any metaphor, this one suggests some useful perspectives on the problem: information is disorganized (as in data mining, but not information retrieval), it can be associated with a "scent" that guides users to it, and users' desire to search further decreases as their appetite is sated. In addition to these informal insights, the work provides a point of contact with research in mathematical ecology into formal models of animal foraging behavior. Much of the progress of science has come from transferring insights derived in one discipline to another by way of an insightful analogy. Information foraging is a very promising instance of this historic process. The first two chapters of Pirolli's work introduce the metaphor, survey the related biological research, and introduce the methodological approach. The methodology is based on Anderson's notion of rational analysis, which focuses on the structure of the task at hand rather than observed behaviors. This analysis forms the basis for construction of a series of production-rule models of task execution in various derivatives of the adaptive control of thought-rational (ACT-R) environment, and comparisons between these models and observed human performance occupy much of the remainder of the book. Chapter 3 reports on some early quantitative analyses of how users search for information on the Web, and chapter 4 offers a rational analysis of Web foraging based on the notion of "information scent," clues in the environment (the page currently viewed by a user) as to where to look for further information. A spreading activation process propagates information among cognitive structures representing cues available in the environment. Chapters 5 and 6 implement cognitive models of information gathering in the ACT-R framework, and compare their behavior with human subjects. Chapter 5 models general information foraging, while chapter 6 models the scatter-gather document browser. Chapter 7 introduces an alternative approach to studying information foraging: stochastic models implemented as Markov processes. Thus far, the book has been concerned with a single user seeking information. Chapter 8 discusses information foraging that takes place in a community, where one user's actions may contribute to another user's search process. Chapter 9 summarizes the study's applications in the form of design recommendations for information systems. Chapter 10 identifies future areas of research stimulated by this study. The book's references are organized by chapter, not as a single integrated bibliography, but there is an integrated index to researcher names, as well as a comprehensive subject index. The volume is a careful, competent exposition of a novel approach to bringing people together with the information they need, and merits careful attention by the information science community. Online Computing Reviews Service