What Do We Mean by Theory in Information Retrieval?

Theory comes in many forms. In some fields, mathematical statements are used to communicate theory, while in others, verbal statements are used. Some fields rely heavily on models, such as physical, mechanical or stochastic, while other fields rely on simulation. As an area of study, information retrieval (IR) addresses topics and problems that can be informed by a wide-variety of theories and models, including those used to describe the actions of machines, as well as those used to explain the behaviors of humans and systems. While theoretical research is often presented as being at odds with empirical research, in reality, they cannot be separated. In this talk, I will review several theories and models that have guided select IR research, and discuss the ways researchers have exercised, explored and tested theories. I will also discuss the role of theories in IR research, and present criteria that can be used to reason about whether an IR theory is useful. I will argue that we should demand more theory from IR research. In particular, while the absence of theory does not prevent us from doing research, it does restrict our findings to a narrow slice of time.