Change search
ReferencesLink to record
Permanent link

Direct link
Agent-Based Argumentative Diagnostic Reasoning in Dementia
Umeå University, Faculty of Science and Technology, Department of Computing Science.
2012 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

This thesis describes the development of a source-based argumentation dialogue system in the medical domain of dementia. Its aim is to enable a physician and a computer agent to establish the presence or absence of a dementia condition or diagnoses for a specific patient. The participants take turn in making the moves in a dialogue, every move consists of an argument or a critical question (a challenge) and during the process of arguments and counterarguments an argumentation-tree, in which a winner may be declared, is constructed.The following four particular problems are formulated and solved: 1. how could an appropriate argumentation graph be designed enabling a dementia debate; 2.  how could an algorithm automatically select attacks in the argumentation graph on behalf of the computer agent using ACKTUS, a semantic web-application for modeling knowledge (1); 3a. how could the reasoning in an argumentation graph be improved by using inquiry dialogues introduced by Black & Hunter (2) to ensure a correct winner election; 3b how could it handle the mutual attack problem, and 4. how could the solutions for these issues be implemented?Through a literature study, ACKTUS review, discussions with expertise and a trial and error approach, solutions to the above problems were suggested. The design of schemes, critical questions and their relation was constructed following a source-based approach. An attack algorithm using a ranking- and a selection procedure was invented where the selection was based on likelihood of winning, degree of user-friendliness, profit, essentiality in the discussion and user preferences. A defeasible logic adapted by Black & Hunter was introduced for the arguments in the designed argumentation graph. After applying special techniques, Black & Hunter’s inquiry dialogue system were used to create dynamic arguments in a new argumentation tree, which enabled a comparison with the designed argumentation tree. This ensured the correctness of the winner election in that argumentation tree. Also an ignorance concept, where the latest attack always is considered a winner, was used to deal with the mutual attack problem. A web-based implementation of the system was accomplished.The results were evaluated regarding functionality. However, a presentation of the system to physicians with further evaluation, still remains. Limitations of the system are that it uses a subset of ACKTUS content that does not include the handling of conflicting arguments, and the mutual attack problem is limited. These issues are all subjects for further development and improvement. Another future prospect is an introduction of Artificial Intelligence algorithm/s such as CBR (Case Base Reasoning) which might improve the system’s performance.One conclusion was that even more refinements were needed before using the system in health care. Another conclusion was that the work had shown how a system for dementia could be developed using an argumentation dialogue between a computer agent and a human participant with an algorithm for automatically attacks and an improvement of reasoning ensuring the accuracy of the winner election.

Place, publisher, year, edition, pages
, UMNAD, 894
National Category
Engineering and Technology
URN: urn:nbn:se:umu:diva-52092OAI: diva2:496619
Educational program
Master's Programme in Computing Science
Available from: 2012-02-10 Created: 2012-02-10 Last updated: 2012-02-10Bibliographically approved

Open Access in DiVA kB)183 downloads
File information
File name FULLTEXT01.pdfFile size 18829 kBChecksum SHA-512
Type fulltextMimetype application/pdf

By organisation
Department of Computing Science
Engineering and Technology

Search outside of DiVA

GoogleGoogle Scholar
Total: 183 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

Total: 216 hits
ReferencesLink to record
Permanent link

Direct link