Publications des agents du Cirad


VWA: ViewpointS web application to assess collective knowledge building

Lemoisson P., Rakotondrahaja C.M.H., Andriamialison A.S.P., Sankar H.A., Cerri S.A.. 2019. In : Nguyen Ngoc Thanh (ed.), Chbeir Richard (ed.), Exposito Ernesto (ed.), Aniorte Philippe (ed.) Trawinski Bogdan (ed.). Computational collective intelligence : 11th International Conference, ICCCI 2019, Hendaye, France, September 4¿6, 2019, Proceedings, Part I. Cham : Springer, p. 3-15. (Lecture Notes in Computer Science, 11683). International Conference on Computational Collective Intelligence (ICCI 2019). 11, 2019-09-04/2019-09-06, Hendaye (France).

DOI: 10.1007/978-3-030-28377-3_1

Collective intelligence is one major outcome of the digital revolution, but this outcome is hardly evaluated. By implementing a topological knowledge graph (KG) in the metaphor of a brain, the ViewpointS approach attempts to trace and assess the dynamics of collaborative knowledge building. Our approach relies on a bipartite graph of resources (agents, documents, topics) and time stamped ¿viewpoints¿ emitted by human or artificial agents. These viewpoints are typed (logical, mining, subjective). User agents feed the graph with resources and viewpoints and exploit maps where resources are linked by ¿synapses¿ aggregating the viewpoints. They reversely emit feedback viewpoints which tighten or loosen the synapses along the knowledge paths. Shared knowledge is continuously elicited against the individual ¿systems of values¿ along the agents' exploitation/feedback loops. This selection process implements a rudimentary form of collective intelligence, which we assess through innovative metrics. In this paper, we present the exploitation/feedback loops in detail. We expose the mechanism underlying the reinforcement along the knowledge paths and introduce a new measure called Multi Paths Proximity inspired from the parallel neural circuits in the brain. Then we present the Web prototype VWA implementing the ViewpointS approach and set a small experiment assessing collective knowledge building on top of the exploitation/feedback loops.

Documents associés

Communication de congrès

Agents Cirad, auteurs de cette publication :