Overview:
The overall aim of AASTM is to enhance today's
generation of argumentation formalisms and implementations in order to
become suitable for a wider variety of real-life applications, such as
reasoning about trust. This requires a unified theory that integrates
the various forms of argumentation related functionality, as well as
efficient proof procedures and sound and scalable software components.
The specific application of the argumentation theory to the field of
trust management is the task of a PhD student who is to be employed to
work on the project.
Objectives:
AASTM is focused on knowledge-based services for the Information
Society, based on semantically rich logic formalisms called
Argumentation Systems [Dung 1995, Vreeswijk 1997]. The overall aim of
AASTM is to enhance today's generation of argumentation formalisms and
implementations in order to become suitable for a wider variety of
real-life applications, such as trust management. The objectives of
AASTM are:
Actions and means involved:
Over the last ten years,
interest in argumentation has expanded dramatically, driven in part by
theoretical advances but also by successful demonstrations of a wide
range of practical applications. AASTM builds on a common framework
recently developed in the European Sixth Framework IST project ASPIC
(2003-2006), which establishes on the one hand a formal foundation to
support the creation, deployment and validation of practical
argumentation systems, and on the other hand a suite of software
components for integrating these components with knowledge (e.g.
semantic web) resources and legacy systems.
Though argumentation technology is currently being applied in fields as medical advising systems [Fox & Das 2000] and machine learning [Mozina et al 2005], present implementations have a limited functionality. AASTM therefore develops argumentation services using advanced approaches like defeasible priorities [Prakken & Sartor 1997], Socratic-style arguments [Caminada 2004] and hierarchical arguments [Modgil 2006], which have thus far not been formalized to such an extent that led to software components could be developed and implemented. A research assistant and an assistant are employed for the following three actions:
Structuration of the project:
Essentially, AASTM is
organized around the interaction among the research assistant and the
assistant, in which the research assistant is developing the formal
model, and the assistant is developing the trust application and the
implementation. Their interaction is arranged via the trust application,
which:
The project leader has several years of experience with managing projects in multi-agent systems and enterprise architecture with this organization, and will see to it that it remains as effective as in previous projects.
Besides, the organization of the project is structured along three work packages following the three actions of the project. The level of accomplishment of the objectives will be measured and assessed during the project’s lifetime, by providing a direct correspondence with the Work packages structure.
Action 1. Basic research.
Task 1.1 Advanced semantics of argumentation frameworks, using semantics
of abstract argumentation frameworks [Dung 1995, Baroni et al 2005,
Coste-Marquis et al 2005], using the connection between logic
programming and argumentation [Dung 1995, Caminada 2006a]
Task 1.2 Conceptual analysis of argumentation forms and methodology,
extending abstract argumentation frameworks [Amgoud & Cayrol 2002,
Bench-Capon 2003].
Action 2. Application to trust management.
Task 2.1. Building on results in the the TRIAS project, the
argumentation theory is used to argue about the reliability of sources.
Task 2.2. Trustworthiness based on the arguments used in dialogue, a new
application domain thus far not studied in the literature.
Action 3. Software development.
Task 3.1. The suitability of particular forms of formal arguments (like
trees [Vreeswijk 1997, ASPIC 2005], lists [Prakken & Sartor 1997]
and sets [Besnard & Hunter 2001] for trust management is analyzed,
and proof procedures and software implementations of the argumentation
theory are developed.
Task 3.2. Scientific documentation and dissemination material.
Present + contextual situation:
From an everyday point of view arguments are "reasons to believe and
reasons to act". Until recently the idea of "argumentation", the process
of creating arguments for and against competing claims, was a subject
primarily of interest to philosophers and lawyers. In recent years,
however, there has been an enormous growth of interest in the subject
from formal and technical perspectives in Computer Science and
Artificial Intelligence, and wide use of argumentation technologies in
practical applications. The benefits of argumentation are a form of
logic that is less "brittle" than classical logic while still being
highly expressive and formally sound.
In computer science and artificial intelligence, argumentation is viewed as a mechanical procedure for interpreting events, organising and presenting documents and making decisions about actions. From a theoretical perspective, argumentation offers a novel framework casting new light on classical forms of reasoning (including logical deduction, induction, abduction and plausible reasoning), communication explanations of advice, supporting discussion and negotiation in computer-supported cooperative work) and learning. From a human-computer interaction point of view argumentation is a versatile technique that facilitates natural system behaviour and is more easily understood by human users and operators. Generally speaking, argumentation capabilities have the potential to add value to any computer-assisted system that provides information and advice to human users or other agents.
The European Sixth Framework IST project ASPIC (2003-2006) has developed a common framework recently developed. The services that have emerged as core functions of the argumentation paradigm include reasoning, decision-making, learning and communication. ASPIC establishes a formal foundation to support the creation, deployment and validation of practical argumentation systems ASPIC has developed a suite of software components based on this framework and a development platform for integrating these components with knowledge (e.g. semantic web) resources and legacy systems.
In ASPIC, the notion of autonomy is used to denote the requirement that the software must have some ability to decide for itself which goals it should adopt and how these goals should be achieved. The methodology of the project uses argumentation in order to permit agents to treat the reasons that justify alternative goals and actions (arguments for and against) as first-class objects that can be explicitly analysed, questioned and rebutted.
Argumentation provides a flexible way of dealing with competing hypotheses in such circumstances, accommodating a wide range of conventional uncertain reasoning methods as well as more advanced inference techniques like defeasible and non-monotonic reasoning and paralogical systems (systems that are tolerant of inconsistency and contradiction). In non-monotonic and uncertain reasoning, argumentation systems have been used to define inference systems for existing non-monotonic logics or to define a non-standard (most often non-monotonic) consequence relation for a particular logic based on a notion of argument. Although these works differ in technical detail and the underlying formal languages they use all of them combine arguments in some way, by aggregation of sets of arguments or undercutting and defeat among arguments, to arrive at cases for competing claims or propositions which can be directly compared. This is a flexible and natural way for agents to reason, make decisions, and select or modify plans in a complex and changing world.
Complex technical systems and services increasingly require several autonomous agents that have to collaborate and communicate in order to achieve required objectives, because of the inherent interdependencies and constraints that exist between their goals and tasks. Increasingly they depend upon complex conversations concerned with negotiation or persuasion where agents have different capabilities and viewpoints. Such dialogues have at their heart an exchange of proposals, claims or offers. What distinguishes argumentation-based discussions from other approaches is that proposals can be supported by the reasons that justify (or oppose) them, roughly equating to explanations for why a proposal or counter-proposal was made. This permits greater flexibility than in other decision-making and communication schemes since, for instance, it makes it possible to persuade agents to change their view of a claim by identifying information or knowledge that is not being considered, or by introducing a new relevant factor in the middle of a negotiation or to resolve an impasse.
Description of the project:
AASTM aims to contribute to both the theoretical state of the art of
formal argumentation, as well as to the available software
implementations. The latter is definitely not a trivial task, as it
involves issues of computability, complexity and efficiency in a broad
sense.
As for the theoretical part, AASTM aims to make contributions in "Dung-base" as well as in "Dung-instantiated" (see the General Technical Description). As for Dung-base, Caminada has recently introduced the concept of semi-stable semantics [Caminada 2006c], which is a particularly promising alternative for domains where traditionally stable semantics has been used. However, current proof procedures for semi-stable semantics are very limited and the issue of computational complexity has barely been touched. AAS aims to deal with these issues in order to come up with an efficient software implementation.
As for "Dung-instantiated", we aim to study which kinds of instantiated arguments are best suited for particular domains. Prakken, for instance, has argued that arguments composed of strict and defeasible rules are best suited for the legal domain. Caminada, in his PhD thesis [Caminada 2004] has discussed the difference between epistemical and constitutive reasoning and its consequences for the validity of default contraposition. AAS aims to put these considerations in a broader perspective and deal with the often neglected issue of which formalisms should be applied under which particular circumstances and for which domains.
In the TRIAS project at the University of Luxembourg, the problem of a knowledge seeking agent to evaluate the reliability of knowledge sources and knowledge items is studied, which is necessary to plan its search, to handle fusion, in particular conflicts, and to guarantee the quality of its informational output. The basic question is which degree of belief and/or which judgments the agent should attach to evidence it receives from a particular source. In a nutshell, this depends on two factors. First, on its prior belief about the subject matter, e.g. how plausible it considers certain assumptions. Secondly, on its assessment of the reliability of the knowledge provider in the given situation, i.e. on how much the agent trusts the source of specific information. To determine the appropriate degree of trust, the agent has to explore its partial know-ledge about the web of science and its constituents, do some reasoning, and possibly even initiate some additional probing or background checks. In the AASTM project, we use this problem to define the first set of requirements and evaluation of the developed argumentation techniques.
Thereafter, we study a new question in trust management. Assuming sources are able to argue themselves, and we observe this argumentation, what can we deduce from the trustworthiness of the source? For example, when a source would provide conflicting information, then we would quickly deduce that the source is unreliable. Other reasons to distrust a source if there is a bias towards particular outcomes. For example, if we consider information sources, the provided information may be colored by political or ideological ideas. In general such biases are not easily detected, but when advanced systems can be asked or forced to give an argumentation for their delivered services, such evaluation becomes within reach. As far as we know, such studies in trust management have not been described in the literature yet.
Finally, we develop software which can be used in the platform developed in the ASPIC project, using state of the art service technology.
Cost/benefit ratio:
Argumentation is a relatively new
paradigm for semantically rich logic systems. It builds on important
European strengths, notably in mathematical logic and computational
logic, and promises to resolve a number of important limitations on
logic-based technologies. Argumentation makes use of the concepts of
computational logic (e.g. logic programming, logic databases, constraint
logics, description logics) but provides a natural way of introducing
knowledge of the world, where "knowledge" is frequently inconsistent,
subjective and relativistic. Classical logics are intolerant of these
realities and argumentation theory provides a powerful but essentially
conservative way of extending classical logics to make them more
flexible and robust, through greater tolerance of uncertainty (e.g. by
assessing collections arguments for propositions rather than "proofs")
and ambiguity (by allowing internally consistent arguments to be
constructed from mutually inconsistent knowledge resources and resolving
the inconsistency through negotiation). Argumentation technologies
therefore have the potential to significantly extend the range of
application of logic-based technologies for both research and practical
use.
By embodying these research results in a set of standard software components, and providing a development platform for integrating the components with other software and resources such as web resources, we believe that AAS has the capability to catalyse commercial and industrial exploitation of argumentation techniques. By making these components available in open source form this will facilitate development of argumentation based technologies and products. Implementation of argumentation services or argumentation based agent technology in software applications could significantly strengthen the software industry and provide a valuable technological advantage.
Argumentation research is also likely to have significant theoretical impact. For example, some argumentation systems have been developed as "rational" reasoning and decision-making systems and promise to resolve long-standing problems of pure quantitative approaches (notably probability and statistical decision-theory) and pure logic-based approaches (e.g. predicate logic, non-monotonic logics).
A particularly prominent beneficiary will be the software agents' community. An increasing number of software applications is being conceived, designed, and implemented using the concept of autonomous agents, ranging from email filtering, through electronic commerce, to large medical, regulatory and industrial applications. Argumentation technology provides important new capabilities for agent systems through new ways of exploiting knowledge in practical reasoning, more flexible decision-making from better management of uncertainty, more sophisticated negotiation capabilities between agents which need to share responsibility for complex tasks, and integration of machine-learning techniques into autonomous systems, which currently have very limited abilities in this respect.
The cost-benefit ration of this project is very good, since the costs are limited to costs for attracting personnel, and the expected scientific output in number of scientific papers is very high. Costs are restricted to costs for personnel. We do no expect there to be a need for expensive software development environments (much argumentation software has been written in freely available platforms such as Ruby and Perl) and having most of the programming done by the proposed PhD student. Moreover, the project will develop a substantial expertise in formal and computational argumentation. The proposed research assistant has been one of the most active participants within the ASPIC project. The ASPIC project has led to many new insights in the field of formal argumentation. Some of these insights have not yet been completely elaborated on. We expect there to be an opportunity for AAS to fill in these gaps. In this way, AAS could with relatively little resources obtain quite interesting results. Based on previous experiences, it is expected that AAS will result in a significant number of high-level publications. The proposed research assistant as well as the project leader have much relevant experience in the field of formal argumentation, as well as an excellent international network in the field of research.
General technical description:
The approach of formal
argumentation emerged in the 1990s [Vreeswijk 1993, Dung 1995, Pollock
1992+1995, Simari & Loui 1992] as a response on developments
regarding defeasible logic in the 1980s (like [Reiter 1980, McCarthy
1986]) The unique approach of argumentation for defeasible reasoning is
that instead of a monlithic approach for defeasible entailment as is
observed in the field of traditional nonmonotonic logics (such as
default logic or autoepistemic logic), entailment under the
argumentation approach is defined in a modular and quite intuitive way.
Based on the information in a knowledge base, arguments are constructed.
It is then examined how these arguments interact and attach each other.
Based on the total picture of the set of arguments and the defeat
relation between these arguments it is then determined, using a
particular argument-based semantics [Dung 1995, Baroni & Giacomin
2006, Coste-Marquis et al 2005, Caminada 2006c] what the different
extensions of arguments are. Based on these extensions, one then obtains
the conclusions that can be seen as justified, using a sceptical or
credulous approach [Baroni & Giacomin 2006].
Although some approaches to formal argumentation already existed in the early 1990s [Simari & Loui 1992, Vreeswijk 1993] argumentation was given a great impuls by the work of Dung [Dung 1995] on the semantics of abstract argumentation frameworks. The approach of Dung received a lot of following, both by those who applied the abstract semantics as one of the components in their argumentation formalism (like [Prakken & Sartor 1997, Caminada & Amgoud 2005]), as well as by those who proposed alternative argument-based semantics given a Dung-style argumentation framework (like [Baroni et al 2005, Caminada 2006, Coste-Marquis et al 2005]) and by those who have proposed to extend Dung's notion of an argumentation framework with new functionality (like [Bench-Capon 2003, Amgoud et al 2004]. Using terminology borrowed from the EU funded ASPIC project one can thus distinguish the following levels of research regarding the functionality of formal argumentation:
Organisation description:
AAS is proposed as a project involving three persons:
Essentially, the project is organized around the interaction among the research assistant and the assistant, in which the research assistant is developing the formal model, and the research assistant is developing the trust application and the implementation. Their interaction is arranged via the trust application, which:
The project leader has several years of experience with managing projects in multi-agent systems and enterprise architecture with this organization, and will see to it that it remains as effective as in previous projects.
Task T1.1: Advanced semantics of argumentation frameworks
This task has as main purpose some of the currently unsolved
argumentation issues in the field of ``Dung-base’’ (see General
Technical Description). The key research question is, given a set of
arguments and a defeat relation between these arguments, which
extensions of arguments can be constructed. Current approaches include
well-known semantics such as stable, grounded, preferred and complete
semantics [Dung 1995], as well as the more recently developed CF2
semantics [Baroni et al 2005] and prudent semantics [Coste-Marquis et al
2005]. A recent and promising approach is semi-stable semantics as
propoese in [Caminada 2006]. The unique property of semi-stable
extensions is that they are equal to the stable extensions in situations
where there is at least one stable extension, while at the same time
the existance of semi-stable extensions is guarenteed, even in
situations where no stable extensions exist. Other desireble properties
also apply [Caminada 2006]. Semi-stable extensions thus provides an
interesting and fault tolerant alternative for domains where
traditionally stable semantics has been applied.
One important issue that needs to be resolved, however, is the development of eficient proof procedures for semi-stable semantics, especially for query-based applications. The approach as has been developed until now often requires the generation of all preferred extensions [Caminada 2006] which is hardly feasible for practical purposes. The first objective of task T1.1 is therefore the development of more efficient and practical proof procedures for semi-stable semantics.
One field where stable semantics is still widely in use is logic programming and answer set programming. An interesting research issue is therefore how the concept of semi-stable semantics can be applied to obtain a level of fault-tolerance for logic programming and answer set programming. The application of semi-stable semantics to logic programming and answer set programming is therefore the second objective of task T1.1 and will be persued both in a declarative was, as well as in the form of executable proof procedures.
This task will provide part of the theoretical underpinning for AAS. However, as well as being sound, the theory must be computationally tractable for the purpose of software development.
Task T1.1 will be mainly the responsibility of the research assistant (Martin Caminada), who already has experience both on semi-stable semantics as well as on logic programming and answer set programming. Task T1.1 will start with the development of proof procedures for semi-stable semantics as applied in abstract argumentation. This will result in a scientific publication, which is scheduled to be submitted before month 9.
The results of this publication will then be used as a starting point for the application of semi-stable semantics in the field of logic programming and answer set programming. This will again result in a scientific publication (for which Caminada will take responsibility) which is scheduled to be submitted before month 18.
Task T1.2: Conceptual analysis of argument forms and methodology
Many forms of formal argument have been stated in the literature,
including lists [Prakken & Sartor 1997], sets [Besnard & Hunter
2001] and trees [Vreeswijk 1997, ASPIC 2005]. One particular point that
has often been overlooked is which of these specific argument forms is
most applicable to which particular argumentation domain. As an
illustration, some argument forms allow the principle of default
contraposition, which is particularly problematic in the field of legal
reasoning. The first objective of task T2.1 is therefore to come up with
a comprehensive theory on which kinds of formal arguments are
applicable to which domains (legal reasoning, trust management, ...).
This will result in at least one scientific publication.
Another related point is that of general methodology for the construction of argumentation formalisms. In his thesis, Caminada treated the issue of the construction of an argumentation formalism basically as a modeling problem; the idea is that formal argumentation models a particular field of informal argumentation (legal, medical, epistemical, ...). The issue of methodology then becomes how one can be sure one has adequalely modeled the argumentation and reasoning that goes on in a particular domain. In his PhD thesis, Caminada has identified several ways of justifying particular design decisions in argumentation formalisms, of which the notion of mirror examples is the most striking one [Caminada 2004]. The second objective of task T1.2 is therefore to work out a comprehensive theory on argumentation design decisions. This will result in another scientific publication.
Task T1.2 will be the main responsibility of Caminada. Van der Torre will also cooperate on the topic of formal argumentation as a modeling issue. As much work has already been done in Caminada’s PhD thesis, we do not expect there to be major problems that need to be overcome. The existing work, however, does need to be elaborated on and be related to other work, which requires a reasonable amount of work to be spent on it.
Task T2.1: Arguments about Trust This work will aim to validate proposals for formal argumentation models and implementation techniques and algorithms on trust management. Using appropriate tools we will develop and test algorithms for these argumentation techniques (e.g. theorem provers, dialogues managers, induction systems) with a particular focus on flexibility, scalability and interoperability. The work package will also investigate optimization options and techniques for core algorithms.
Objectives:
For the use of the trust model and scenarios in trust management, we use results obtained in the TRIAS project. The task will be done by the assistant, together with researchers involved in the TRIAS project (including the project leader).
Task T2.2: Trust Application
Objectives:
This will be partly exploratory, given the lack of related research on how the used arguments may indicate the reliability of a source. The task will therefore include the assistant, the research assistant, and active involvement of the project leader.
We start by discussing some scenarios in which the argumentation of a source may indicate the trustworthiness of a source, by rereading some of the existing literature on trust research and explore how explicit argumentation may be used as an additional parameter.
From the scenarios we define criteria, like consistency or biases. Finally we test the trustworthiness using the formal model and a prototype. The material obtained in this task will be the main result of the PhD thesis of the assistant (which he or she will write during the third year of the project).
Task T3.1: Software development
The main objective of task T3.1 is the development of a software
component that is capable of argumentation using at least grounded,
credulous preferred and semi-stable semantics.. Apart from that, the
software should also implement various advanced argumentation features
that have yet been left unimplemented by existing argumentation
software, such as defeasible priorities, Socratic-style arguments and
hierarchical argumentation.
The development of argumentation software is mainly the task of the still to be assigned research assistant. Standard software engineering methodology will apply.
Task T3.2: Exploitation and dissemination
The main objective of task T3.2 is to manage project results
exploitation and dissemination. We will seek to disseminate its research
results to the academic community (to be performed by participating in
conferences, summer schools like the ESSLLI and giving lectures) and to
demonstrate the practical viability of argumentation technology as
illustrated by the developed software of task T3.1.
Task T3.2 is the responsibility of everyone involved in the AAS project (van der Torre, Caminada and the yet to be assigned assistant). We aim to perform at least one tutorial on formal argumentation and write at least one article on argumentation technology for an industrial (non-scientific) audience.
Software dissemination. The AAS components will be documented and made available for research purposes in source code form.
References
[Amgoud & Cayrol 2002] L. Amgoud and C. Cayrol, "A reasoning model based on the production of acceptable arguments". Annals of Mathematics and Artificial 34 (1-3):197-215 (2002)
[Amgoud et al 2004 ] L. Amgoud, C. Cayrol and M.-Ch. Lagasquie-Schiex, "On the bipolarity in argumentation frameworks". NMR 2004 1-9 (2004)
[ASPIC 2005] ASPIC consortium, "Deliverable D2.6 - Final review and report on formal argumentation system" (2005)
[Baroni et al 2005] P. Baroni, M. Giacomin and G. Guida. "Scc-recursiveness: a general schema for argumentation semantics". AIJ 168(1-2):165-210 (2005)
[Baroni & Giacomin 2006] P. Baroni and M. Giacomin, "Evaluation and comparison criteria for extension-based argumentation semantics". COMMA 2006
[Bench-Capon 2003] T.J.M. Bench-Capon, "Persuasion in Practical Argument Using Value Based Argumentation Frameworks". Journal of Logic and Computation 13(3):429-448 (2003)
[Besnard & Hunter 2001] Ph. Besnard and A. Hunter, "A logic-based theory of deductive arguments". AIJ 128(1-2):203-235 (2001)
[Boella et al, 2005a] G. Boella, J. Hulstijn, and L. van der Torre, Argument games for interactive access control. Proceedings of WI05. IEEE, 2005.
[Boella et al, 2005b] G. Boella, J. Hulstijn, and L. van der Torre, Argumentation for access control. Proceedings of AI*IA05. Springer, 2005.
[Boella et al, 2006] G. Boella, J. Hulstijn and L. van der Torre, A logic of abstract argumentation. Postproceedings of ArgMAS05, Springer, LNCS, 2006.
[Caminada & Amgoud 2005] M.W.A. Caminada and L. Amgoud, "An Axiomatic Account of Formal Argumentation". AAAI 2005 608-613 (2005)
[Caminada 2004] M.W.A. Caminada, "For the Sake of the Argument; explorations into argument-based reasoning". PhD dissertation Vrije Universiteit Amsterdam (2004)
[Caminada 2006a] M.W.A. Caminada, "Well-Founded semantics for Semi-Normal Extended Logic Programs". NMR 2006
[Caminada 2006c] M.W.A. Caminada, "Semi-stable semantics". COMMA 2006
[Coste-Marquis et al 2005] S. Coste-Marquis, C. Devred and P. Marquis, "Prudent semantics for argumentation frameworks". Proc. of the 17th IEEE International Conference on Tools with Artificial Intelligence (ICTAI 2005) 568-572 (2005)
[Dung 1995] P.M. Dung, "On the acceptability of arguments and its fundamental role in nonmonotonic reasoning, logic programming and n-person games". AIJ 77:321-357 (1995)
[Fox & Das 2000] J. Fox and S. Das, "Safe and Sound. Artificial Intelligence in Hazardous Applications". AAAI Press/MIT Press, 2000.
[Kaci and van der Torre 2006a] S. Kaci, L. van der Torre and E. Weydert, Acyclic Argumentation: Attack = Conflict + Preference. Proceedings of ECAI 2006.
[Kaci and van der Torre 2006b] S. Kaci and L. van der Torre, Preference reasoning for argumentation: non-monotonicity and algorithms. Proceedings of the NMR'06, 2006.
[McCarthy 1986] J. McCarthy, "Applications of circumscription to formalizing common sense knowledge". AIJ 26(3):89-116 (1986)
[Modgil 2006] S. Modgil, "Value Based Argumentation in Hierarchical Argumentation Frameworks". COMMA 2006
[Mozina et al 2005] M. Mozina, J. Zabkar, T.J.M. Bench-Capon and I. Bratko, "Application of Argument Based Machine Learning to Law". ICAIL 2005 248-249 (2005)
[Pollock 1992] J.L. Pollock, "How to reason defeasibly". AIJ 57:1-42 (1992)
[Pollock 1995] J.L. Pollock, "Cognitive Carpentry. A Blueprint for How to Build a Person". MIT Press, Cambridge, MA (1995)
[Prakken & Sartor 1997] H. Prakken and G. Sartor, "Argument-based extended logic programming with defeasible priorities". Journal of Applied Non-Classical Logics 7:25-75 (1997)
[Prakken 2005a] H. Prakken, "A study of accrual of arguments, with applications to evidential reasoning". ICAIL 2005 85-94 (2005)
[Prakken 2005b] H. Prakken, "Coherence and flexibility in dialogue games for argumentation". Journal of Logic and Computation 15 1009-1040 (2005)
[Reiter 1980] R. Reiter, "A logic for default reasoning". AIJ 13:81-131 (1980)
[Simari & Loui 1992] G.R. Simari and R.P. Loui, "A mathematical treatment of defeasible reasoning and its implementation". AIJ 53:125-157 (1992)
[Vreeswijk 1993] G.A.W. Vreeswijk, "Studies in Defeasible Argumentation". PhD dissertation Vrije Universiteit Amsterdam (1993)
[Vreeswijk 1997] G.A.W. Vreeswijk, "Abstract argumentation systems". AIJ 90:225-279 (1997)
Progress reports:
yearly report AASTM 2007