Incorporating rhetorical and plausible reasoning in a system for simulating argumentation
Introduction
Modern argumentation theory seeks to understand the nature and structure of argumentation, which is a very common type of human communication. Its underlying principles are much more complex than those suggested by formal logic models of reasoning, as it uses a wide range of techniques, from analogies to emotional appeals, each technique having its own appropriate context.
The aim of this article is to explore the value of modern argumentation theory for our work building a system for simulating argumentation, focussing particularly on the extra value provided by a small sample of non-logical, rhetorical concepts. Our method is to record and transcribe a meeting, to code it according to the Toulmin scheme [1] of claims, warrants and premises, and to write the warrants as rules in our system, and then to run the system to generate a simulation of the meeting.
Previous work by the authors has involved simulation of a business meeting for which the system has been tested by running it with premises available in the same order as the arguments recorded during the meeting [2]. That simulation was compared with the transcript and with the responses of 55 business graduates (giving their reasons) over a test set of 23 arguments [3]. The present article involves simulation of a second meeting, a university seminar.
Section snippets
Argumentation theory
Following van Eemeren et al. [4]: the aims and objectives of the field of argumentation theory can be characterised as: (1) at a prescriptive level, to develop principles which inform people how to argue well; (2) at a descriptive level, to model the processes of argumentation underpinning such phenomena as setting and shifting a topic, holding the floor in conversation and negotiating social identities; (3) at a normative level the aim of the field is to develop a framework in order to
Computational models of argumentation
Any computational model of argumentation raises a number of questions. What is meant by relevance and how is the relevance of a piece of evidence established [99], [100]? How many different levels does an argument operate on—the method and order of presentation, the rhetorical devices used, the emotional appeal of the message, and so on [24]? How should these be represented? When are arguments descriptive, normative or prescriptive [25]? How do prescriptive statements affect descriptive ones?
Applications of computational argumentation
Within the legal and design domains the aim of argumentation systems is principally to decide which of several cases provides what kind of precedent [114]. A more ambitious system is HYPO [99], [100], [102], [115], used in the domain of trade secret law. It is also possible to generate hypothetical situations used as evidence to test arguments on a ‘what would happen if …’ basis [116]. HYPO provides information about relevant similarities and differences between cases (factors, dimensions, and
Significance of computational argumentation
Do people take account of argumentation in organisational situations or are decisions based on prior preferences? It has been found that there are two approaches to persuasion—the central route (via argumentation) relevant to those whose level of arousal or ‘need for cognition’ [58] is high—and the peripheral route, when the lazy and persuadable subject is vulnerable to selling and persuasion techniques such as suggestion and repetition and influenced by liking and affiliation [59].
However,
Debate simulation
Fig. 1 shows the debate text used as the source for the simulation, and Fig. 2 shows the system inputs and outputs.
The right hand column of Fig. 2 shows the user inputs [and in parentheses, the previous claims available as premises]. The middle column of Fig. 2 shows the warrants or rules used to justify the claims (shown in the left hand column).
Domain dependence
One potential criticism of the system is that of necessity the warrant set available to agents is domain dependent. The rules were used to represent a particular debate—the specifics of an example argument—which leaves open the question of whether or not they could be used in any automatic system or how effective they would be in live argumentation. In the example debate, there were 21 warrant rules and 15 facts. Even this, however, is a start and is no barrier to practical applications—Farley
References (119)
Argumentation dynamics: justifying shifts in focus and scope
Journal of Pragmatics
(1995)Multi-agent conflict resolution: a computational framework for an intelligent argumentation system
Knowledge-Based Systems
(1994)- et al.
Argument representation for editorial text
Knowledge-Based Systems
(1990) Co-operative structuring of information: the representation of reasoning and debate
International Journal of Man–Machine Studies
(1985)- et al.
Meditator: towards a negotiation support system
European Journal of Operations Research
(1987) - et al.
Hierarchical formalisations
International Journal of Man–Machine Studies
(1991) - et al.
Multimedia in a design rationale decision-support system
Decision Support Systems
(1995) Would electronic argumentation improve your ability to express yourself?
- et al.
Supporting collaborative processes with conversation builder
Computer Communications
(1992) - et al.
The logic of plausible reasoning: a core theory
Cognitive Science
(1989)
Foundations of AI: the big issues
Artificial Intelligence
Rigor mortis: a response to Nilsson’s logic and artificial intelligence
Artificial Intelligence
Thinking may be more than computing
Cognition
The uses of argument
A formal model of organisational argumentation: applications of informal logic to a committee meeting
Cybernetics and Systems: An International Journal
Speech acts in argumentative discussions: a theoretical model for the analysis of discussions directed towards solving conflicts of opinion
Arguers as editors
Argumentation
The rhetoric of economics
Journal of Economic Literature
Reasoned use of expertise in argumentation
Argumentation
The theory of communicative action
Rhetorical criticism as argument
Quarterly Journal of Speech
Understanding arguments: an introduction to informal logic
Thinking about thinking, or do I sincerely want to be right?
Straight and crooked thinking
Cognitive psychology and its implications
The structure and process of understanding: implications for managerial action
Sensemaking in organizations
The new rhetoric: a treatise on argumentation
What makes a strong argument?: Emotions, highly-placed values, and role-playing
Communication and Cognition
A design for information systems which can adapt to changing organisational requirements
Software Maintenance: Research and Practice
The intention-inference problem in argumentation understanding
Communication and Cognition
Organisational and behavioural issues raised by intelligent argumentation systems
Behaviour and Information Technology
Argumentation, self-consistency, and multi-dimensional argument strength
Communication and Cognition
Using argumentation to overcome hypertext’s HCI failings
A multi-agent reasoning model for negotiation support
Dynamic memory
In-depth understanding
Cited by (7)
Frame-based argumentation for group decision task generation and identification
2005, Decision Support SystemsIssues of feasibility, coherence, and robustness in a premise-to-claim model of argumentation: Results from four experiments
2001, European Journal of Operational ResearchComputer-mediated communication: Problems and potentials of argumentation support systems
1999, Decision Support SystemsThe verification and modification of applicability of Lewicki & Bunker's trust model in VC-E's practice of China
2012, International Conference on Management Science and Engineering - Annual Conference ProceedingsA discussion information-structuring model based on the toulmin formalism
2008, e-Forensics 2008 - Proceedings of the 1st International Conference on Forensic Applications and Techniques in Telecommunications, Information, and Multimedia and WorkshopA discussion information-structuring model based on the Toulmin formalism
2008, Proceedings - 1st International Workshop on Knowledge Discovery and Data Mining, WKDD