Bachelor thesis projects

Project: A XML Transformation Tool for Argumentation Patterns

TYPE: bachelor thesis

Engagement: 10 weeks for 3 days a week

PROBLEM:
An argumentation framework [Dung95] is a graph whose nodes are called arguments and whose edges are called attack. Argumentation patterns are descriptions for how to solve design problems of argumentation frameworks that can be used in many situations. In other words, they are a description or template for how to solve a problem that can be used in many different situations. They describe relationships and interactions between arguments, without specifying the final application arguments that are involved. Graphical notations are important and useful in design, but they are not precise enough to define how argumentation patterns can be combined. Argumentation patterns need to be flattened into argumentation graphs where the nodes are distinguished into auxiliary nodes and focal nodes and they represent the arguments and the directed edges represent the attack relation among arguments.

The thesis is addressed to the implementation of a XML transformation tool for the argumentation patterns already defined in the literature.

METHODOLOGY:
develop XML transformation tools to implement argumentation patterns. The basis of these transformation tools is to show a clear distinction between:

  • representation: a XML file
  • visualization: visualization of the XML file
  • reasoning: main focus of the argumentation patterns, e.g., how to combine patterns.

Further applications of the developed XML transformation tools are addressed towards the NAO robots of the interdisciplinary research group Individual and Collective Reasoning (ICR) of the University of Luxembourg.

EXAMPLES:
Assume that a modeler believes that the argument "Jones is liable" is accepted if "Jones has a contract" and "Jones has breached the contract." This is an instance of the conjunction pattern: the former argument is accepted only if each of the latter is accepted.

COLLABORATIONS: Guido Boella and Serena Villata, University of Turin

FURTHER INFORMATION: Leon van der Torre, ICR, University of Luxembourg

Bachelor / master thesis projects

Project: Norm simulation

Type: bachelor thesis, or master thesis

Problem: apply normative reasoning methodologies developed in the field of multiagent systems to the field of agent based simulation, for example to the simulation of economic systems.

Methodology: tools such as jES (the Java Enterprise Simulator) and SLAPP (Swarm-Like Agent Protocol in Python) developed at the University of Torino.

Examples:

  • bachelor thesis: apply tools to example of choice like
  • master thesis: use simulation to study normative reasoning

Collaborations: Guido Boella, University of Torino

Further information: Leon van der Torre

Master thesis projects

Project: Geometric Rule Updating

TYPE: master thesis

PROBLEM:
Suppose we have a system whose "information state" is represented as a bounded 2-dimensional plane divided into rectangles, with each rectangle being assigned one of two colours, red or blue (which could stand for "true" and "false", or "accept" and "deny"). Each pair of (rectangle, colour) may be thought of a rule saying "all points within (rectangle) are assigned (colour)". Now suppose new information comes along in the form of a new rectangle together with a colour, saying that all the points within this rectangle are to be re-assigned to that colour. This new information may well conflict with the initial state, in that the new rectangle may overlap with rectangles having the opposite colour.

The main problem to be solved in this project is to design and implement an algorithm which will update any given initial information state to incorporate the new information, subject to the following constraints:
(i) the new state must again be a partition of rectangles, and
(ii) it should be kept "as similar as possible" (in some sense to be determined) to the initial one.

This simple problem is relevant to problems of the dynamics of real systems which can be modelled as sets of geometric rules, for example Firewall Rule bases.

PREREQUISITES:
Programming skills essential.
Familiarity with logic and proof desirable.
Familiarity with belief revision a bonus, but not essential.

WORK PLAN AND METHODOLOGY:
1. Formulate the problem and think about and write down in a formal way some desirable properties for the update operation, including properties for iterated update.
2. Design and implement an algorithm which does the job. Show that it satisfies all the desirable properties.
3. If time permits, think about extending the problem to higher dimensions, e.g. cubes instead of rectangles. Such an extension could lead to applications in firewall rule update, where the number of dimensions is typically 5.

REFERENCES:
- An axiomatic approach to firewall rule update

FURTHER INFORMATION: Richard Booth

Project: Argumentation Patterns: reasoning and decision making

TYPE: master thesis

ENGAGEMENT: 15 weeks for 5 days a week

PROBLEM:
An argumentation framework [Dung95] is a graph whose nodes are called arguments and whose edges are called attack. Argumentation patterns are descriptions for how to solve design problems of argumentation frameworks that can be used in many situations. In other words, they are a description or template for how to solve a problem that can be used in many different situations. They describe relationships and interactions between arguments, without specifying the final application arguments that are involved.

Graphical notations are important and useful in design, but they are not precise enough to define how argumentation patterns can be combined. Formal semantics is needed to define the argumentation patterns and their use. Moreover, argumentation patterns need to be flattened into argumentation graphs where the nodes are distinguished into auxiliary nodes and focal nodes and they represent the arguments and the directed edges represent the attack relation among arguments. Starting from the example of argumentation schemes [Reed&Walton05], argumentation as dialogic model is seen as an underlying metaphor so it is well suited for explanation and interaction. The challenge is to use the argumentation patterns, instead of argumentation schemes which are only a linguistic description, in order to have an hybrid system which generates arguments from a reasoning schema with the aim to support intelligent agents in decision making.

The thesis is addressed to the theoretical definition and implementation of argumentation patterns to be embedded in a multiagent system for supporting decision making.

METHODOLOGY: theoretically define and develop with a XML transformation tool argumentation patterns. The basis of these transformation tools is to show a clear distinction between:

  • representation and visualization of the XML file
  • reasoning and decision making: the main aim of the argumentation patterns to be defined: the derivation of goals in practical reasoning and the need to explain decisions

The achieved results in reasoning and decision making will be addressed towards the NAO robots of the interdisciplinary research group Individual and Collective Reasoning (ICR) of the University of Luxembourg.

Examples of argumentation pattern: assume that a modeler believes that the argument "Jones is liable" is accepted if "Jones has a contract" and "Jones has breached the contract." This is an instance of the conjunction pattern: the former argument is accepted only if each of the latter is accepted. An example of goal generation schema is the following Toulmin scheme where the claim is "You should buy our whitening toothpaste!", the data is "Studies show that teeth are 50% whiter after you use this toothpaste!" and the warrant is "Duh, people want whiter teeth!"

COLLABORATIONS: Guido Boella and Serena Villata, University of Turin

FURTHER INFORMATION: Leon van der Torre, ICR, University of Luxembourg