Over the last few years, new views of intelligence originated by social science researchers have challenged prevailing assumptions within Artificial Intelligence. They have particularly influenced research into the design of interactive systems, my own area of interest. Social science research has brought to light the situated, interactive nature of intelligent activity that AI approaches have de-emphasized. It also has highlighted the key role of tools in intelligent activity.
A new AI research school that aims at designing cooperative systems is influenced significantly by these two bodies of social science work. The goal of such research is to design systems that complement the abilities of people and that augment their performance, rather than duplicate people's abilities and replace them.
Along with colleagues at the MCC Human Interface Laboratory (Hollan et al. 1991), I developed an approach to the design of cooperative systems called collaborative manipulation. It grounds person-computer cooperation on collaborative manipulation of objects in a shared workspace. The principles of collaborative manipulation have been applied to the design of a knowledge editing tool called the HITS Knowledge Editor.
This paper is structured as follows. I first present the relevant social science research and draw implications for the design of interactive systems. I then characterize the collaborative manipulation approach in terms of three design principles and show how they are informed by this research. Finally, I show how these principles are embodied in the HITS Knowledge Editor, illustrate the effectiveness of the system using data taken from user studies, and argue that the effectiveness is due to the design approach.
2 Social Science Influences
Recently, social scientists from fields like ethnomethodology and anthropology have critiqued the mentalistic conception of intelligence prevalent within mainstream AI, which focuses on structures and processes internal to an agent. Through both empirical studies and theoretical arguments (Garfinkel 1967; Lave 1988; Sch?on 1983; Suchman 1983, 1985), they have constructed an alternative picture of activity as being inherently bound up with the details