Models, theories, and methods in image research: A discussion and a look to the future

Authors


Introduction

Library and information science (LIS) researchers, though often criticized for not using theoretical frameworks and established methods enough, are beginning to either develop such frameworks and methods or adopt those developed within other fields. However, several questions are still unanswered. This interactive panel will raise questions related to models, theories, and models applicable to image retrieval research for discussion by the audience. This panel follows the discussion started in the companion image research panel, which will have presented current image research projects and framed them in the context of a broader image retrieval research framework. Specific aims of this panel, which is sponsored by SIG VIS, include:

  • To outline the origins and foundational models, theories, and methods used in image research
  • To begin a discussion among researchers interested in all aspects of image research regarding the models, theories, and methods commonly used
  • To discuss the pros and cons of the models, theories, and methods used in image research and to identify the issues/gaps present

Audience

All researchers and professionals whose main concern is the full spectrum of issues in library and information science (LIS) that apply to image creation, organization, storage, retrieval, semantics, and use would benefit from the panel's discussions. Members of the audience will have an opportunity to network and form future collaborative research teams based on their common interests.

Models and Theories

Conceptual models may be considered a representation of major concepts and the interrelations amongst these concepts. They are more general in nature than theories. Conceptual models provide a framework for thinking, and help researchers to systematize knowledge and guide their research. In essence, they may provide a map of reality. The domain of images is complex due to the elements of subjectivity and context, and the fact that the subject content is more prone to interpretation than textual information. It is important to understand and explore the value of conceptual models in this domain. This is one of the several aspects of image research this panel seeks to address.

Humans' innate biological, intellectual, and emotional responses to visual information necessitate the inclusion of theories from multiple disciplines, such as cognitive science and psychology. For example, visual perception theories address how the human brain processes and interprets low-level visual information. This primitive processing level ties to all higher levels of image understanding. Attention theories may explain how humans select focal points; understanding the selection of these focal points may inform user needs for representation. The theory of the basic level notes the human tendency to categorize objects at consistent levels of description. Theories discussing the existence of five or six basic emotions could direct a need for users to index and search for affect within an image retrieval system.

Image researchers have used many of these interdisciplinary models and theories as the basis for studying multiple dimensions associated with images, users, and systems. However, little discussion exists regarding the appropriateness of their use, the limitations encountered, or a synthesis of the various theories employed in image research.

Methods

Any solid research needs a set of established and widely used methods in order for its results to have acceptable levels of both internal and external validity and reliability. The full spectrum of issues that library and information science (LIS) is concerned with (generation, collection, organization, interpretation, storage, retrieval, dissemination, transformation and use of information) demands different sets of methods to investigate and understand fully the phenomena being studied. Issues related to image research are no exception. Only reasonably valid and reliable results would be useful in our efforts to design user-centered, efficient, and effective image indexing and retrieval systems.

A brief look at image research literature reveals that researchers select methods appropriate only for the topic or research problem at hand. As a result, any comparison of results is very difficult. There is a need for common methods of sample selection (materials and subjects or users) and procedures of data collection and analyses. This calls for an examination of past practices and a discussion about the way forward.

Questions and Topics for Discussion

The following questions and topics would guide the discussion by the panelists and the audience:

  • 1.A discussion of the methods, models, & theories currently used by image researchers in general and the cognitive and emotional aspects of interactions with visual information in particular
  • 2.The origins (LIS or other fields of study) of these methods, models, & theories
  • 3.How widely are they used (both within and outside LIS)?
  • 4.What methods, models, & theories are appropriate for image research?
  • 5.What are the pros and cons of these methods, models, & theories?
  • 6.Discussion on the need for new methods, models, & theories
  • 7.What are we looking for in these new methods, models, & theories?
  • 8.Do conceptual models play a crucial role in image research?

Ancillary