Supporting Collective Intelligence through a Recursive Annotation Tool Federico Cabitza University of Milano-Bicocca Viale Sarca 336, 20126 Milan, Italy cabitza@disco.unimib.it Marco P. Locatelli University of Milano-Bicocca Viale Sarca 336, 20126 Milan, Italy locatelli@disco.unimib.it Carla Simone University of Milano-Bicocca Viale Sarca 336, 20126 Milan, Italy simone@disco.unimib.it ABSTRACT The nature of the cooperative setting in which the phe- nomenon of Collective Intelligence is considered influences the mechanisms by which Collective Intelligence emerges from that setting and how this process can be supported by a computer-based technology. This paper focuses on tight communities and specifically on the collaborative discourses that their members articulate therein around documental ar- tifacts. The paper presents a collaborative annotation tool endowed with discourse oriented functionalities whose char- acteristics emerged from our investigations in the agricul- tural and healthcare domains. In addition, the tool supports those activities that are aimed at building a representative and digested summary of the opinions gathered collabora- tively in a community, and those aimed at collaboratively building a modified version of the original document ac- cording to those opinions and standpoints. The tool and an example of its customization according to the requirements collected from the agricultural domain are presented. 1. INTRODUCTION In the past year we were involved in research projects in the agricultural and in the healthcare domains to investi- gate how to support collaboration among agricultural grow- ers that belong to a regional association in the former case, and among doctors, nurses and patients in the latter case. In both cases annotations of specialized artifacts was one of the relevant practices we observed. The main outcomes of our investigation in the healthcare domain are reported in [7, 12, 8]. In this paper we briefly report on the outcomes of our investigation in the healthcare domain and more deeply in the agricultural domain since they nicely complement each other. From this empirical research we derived a set of requirements that we are currently implementing in an annotation tool called RAT (Recursive Annotation Tool) that serves as sand- box for testing the feasibility of the functionalities to be engineered in future projects. The paper positions this research effort within the broader theme of CI to identify the scope of application of the pro- posed technology and presents the issues uncovered in the healthcare domain and the setting and the initial solution developed for the project in the agricultural domain. Then it discusses a set of more sophisticated, discourse oriented functionalities that emerged from our interactions with the involved stakeholders in both domains. The concluding re- marks position the proposed annotation tool in the frame- work of a more comprehensive research agenda and hint to the next steps of our research in this field. 2. OUR FOCUS WITHIN THE CI THEME The theme of Collective Intelligence (CI) is strictly related to the theme of cooperative knowledge work: this is espe- cially true when CI is combined with Online Discourse, that is with a conscious [17] interaction among the involved ac- tors. The cooperative and social nature of a work setting influences the mechanisms by which CI is created as well as they ways such a creative process can be supported by a mediating technology. Two orthogonal dimensions can be considered: first, the links existing between the involved ac- tors 1 ; second, the kind of issues that trigger and exploit the phenomenon of CI. In regard to the first dimension, we can distinguish between situations where actors gather in loose communities: in this case, the links between the commu- nity members are weak, since members can come to inter- act for the first time or because their previous interactions did not build a recognizable common background [22]; and situations where actors constitute tight communities, whose members are mutually acquainted with each other since they share previous experiences or rely on sufficiently understood and agreed-upon notions and conventions. The second di- mension is related to what triggers social interactions: this can be a wicked problem [30] that requires exploration in unknown territories; or the collection of (and agreement on) professional practices that are found to be the “best ones” to deal with some situations; or the collection of opinions “about and around” an artifact in order to reach some con- sensus about them, facilitate their spreading and adoption or continuous refinement and improvement. This list is open ended: we mentioned some initial items that we can de- rive from the literature and our direct research experience 1 this dimension is reminiscent of the concept of “work cou- pling” used in [26] “for defining the intensity or demand of the work for information sharing or level of communication required” and in [28] to discuss how distance influences col- laboration