Date of Award
Doctor of Philosophy (PhD)
Computer Engineering and Sciences
Arguments are essential objects in a debate. In the DirectDemocracyP2P system, they occur in association with signatures for petitions. The arguments of a signer on a given issue are grouped into one single justification, are classified by the type of signature, in our case supporting or opposing, and can be subject to various types of relations such as refute, subsume or more recent. Given those signatures and relations, the addressed problems are: (i) what makes a good supporting or opposing justification (ii) how to recommend the best justification to a new voter, (iii) how to recommend a compact list of justifications subsuming the majority of known arguments for or against an issue. In this dissertation, we investigate solutions based on weighted bipartite graphs within the context of the DirectDemocracyP2P (DDP2P) system. We describe algorithms based on computing transitive closures covering different types of edges in those graphs, refutes and subsumes. Due to the decentralized design of DDP2P, the system can be in a constantly inconsistent state as peers come up and down, organization are created and destroyed, users join and leave them. It therefore has a probabilistic nature. We describe a Bayesian network model to generate data and compare the algorithms. We conclude with a brief discussion on new research avenues that this thesis opens. We argue that our approach takes a small step towards making voters more knowledgeable and their decisions better supported.
Roussev, Roussi, "Identifying The Most Relevant Arguments From User Meta-Data" (2018). Theses and Dissertations. 833.