In decision making, negotiation, and other kinds of practical reasoning, it is necessary to model preferences over possible outcomes. Such preferences usually depend on multiple criteria. We argue that the criteria by which outcomes are evaluated should be the satisfaction of a person’s underlying interests: the more an outcome satisfies his interests, the more preferred it is. Underlying interests can explain and eliminate conditional preferences. Also, modelling interests will create a better model of human preferences, and can lead to better, more creative deals in negotiation. We present an argumentation framework for reasoning about interest-based preferences. We take a qualitative approach and provide the means to derive both ceteris paribus and lexicographic preferences.

Goals are not only used to identify desired states or outcomes, but may also be used to derive qualitative preferences between outcomes. We show that Qualitative Preference Systems (QPSs) provide a general, flexible and succinct way to represent preferences based on goals. If the domain is not Boolean, preferences are often based on orderings on the possible values of variables. We show that QPSs that are based on such multi-valued criteria can be translated into equivalent goal-based QPSs that are just as succinct. Finally, we show that goal-based QPSs allow for more fine-grained updates than their multi-valued counterparts. These results show that goals are very expressive as a representation of qualitative preferences and moreover, that there are certain advantages of using goals instead of multi-valued criteria.

The research reported on in this thesis is part of a larger research project that aims to develop a negotiation support system called the Pocket Negotiator. This thesis focuses on the question how such a system can represent and reason about a user’s preferences between the possible outcomes of a negotiation. In real-world negotiations, there are many negotiation issues which can have many different values, resulting in a large space of complex outcomes. A negotiation support system needs to have a model of the user’s preferences over this outcome space. Although most current negotiation support systems use numerical measures such as utility to represent preferences, such quantitative preferences are hard to specify for human users, and so it would be more natural to model the user’s preferences in a qualitative way. Moreover, due to the exponential size of the outcome space, it is not feasible to specify a preference ordering directly. Therefore, we aim to represent the preferences in a more compact way by aggregating multiple evaluation criteria that influence preference. The main research objective of this thesis is to develop a framework for the representation of, and reasoning about such qualitative multi-criteria preferences. The thesis makes the following contributions. • We propose strategies to derive preferences from incomplete or uncertain information about the objects to be compared. The decisive and safe strategy for incomplete information is based on the notion of least and most preferred completions of objects. The strategies for uncertain information are based on an ordinal representation of the certainty levels of facts. • We argue that instead of negotiation issues, the negotiators’ underlying interests should be chosen as criteria, especially if the issues are not preferentially independent. We show that the use of interests as criteria is more flexible than modelling conditional preferences, and provides a better explanation of the derived preferences. • We present a general framework for the representation of qualitative, multicriteria preferences, called Qualitative Preference Systems (QPS). The framework defines outcomes as value assignments to a set of variables which can have arbitrary domains, includes a knowledge base that can impose (hard) constraints and define new (abstract) concepts, and defines three types of criteria that can be combined in a tree structure. We show that the QPS framework is expressive, as it can model conditional preferences and underlying interests, goal-based preferences, bipolar preferences, and preferences represented in two other well-known approaches that are representative for a large number of purely qualitative preference modelling approaches. Moreover, we show that the goal-based variant of QPS is just as expressive. • For all proposed preference representation frameworks we define corresponding argumentation frameworks that include a logical language, a set of inference rules, and a defeat relation. Some of the argumentation frameworks also provide the possibility to reason with background knowledge to derive information about the values of variables by default. • We propose a mechanism to generate explanations for preferences represented in a QPS. We use the intuition that preferences can be explained by the criteria that are deciding in the overall preference. Moreover, we show how a system can use user-provided explanations to update its current preference model. • Finally, we introduce a modal logic, called Multi-Attribute Preference Logic (MPL), that provides a language for expressing several strategies to qualitatively derive a preference between objects from property rankings. Three such strategies from the literature on prioritized goals are modelled. The additional value of the logic is that it is possible to reason not only about which objects are preferred according to a certain ordering, but also about the relation between different orderings.

Introduction A key challenge in the representation of qualitative, multi-criteria preferences is to find a compact and expressive representation. Various frameworks have been introduced, each of which with its own distinguishing features. In this paper we introduce a new representation framework called qualitative preference systems (QPS), which combines priority, cardinality and conditional preferences. Moreover, the framework incorporates knowledge that serves two purposes: to impose (hard) constraints, but also to define new (abstract) concepts. QPSs are based on the lexicographic rule studied in [1]. This rule is a fundamental part of the framework presented as it offers a principled tool for combining basic preferences. We believe this ability to combine preferences is essential for any practical approach to representing qualitative preferences. It is needed in particular for constructing multi-criteria preferences. It is not sufficient, however, since more expressivity is needed and useful in practice. Therefore, QPSs in addition provide a tool for representing knowledge, for abstraction, for counting, and provide a layered structure for representing preference orderings. QPSs are able to represent various strategies for defining preference orderings, and are able to handle conditional preferences. Logical Preference Description language (LPD; [3]) can be embedded into the QPS framework and that there is an order preserving embedding of CP-nets [2] in the QPS framework. These embeddings provide a representation that is just as succinct as the LPD expressions and CP-nets.

We propose an explanation facility for a qualitative preference representation framework. We show how an explanation can be provided for qualitative, multi-criteria preferences based on the criteria that are used to decide preferences between outcomes. Such a facility provides an important tool for a user to understand how preferences are determined. We show that this facility can also be used by a user to inform the system about its preferences. Such a user-provided explanation can be used for updating and improving a preference model maintained by the system.

A key challenge in the representation of qualitative, multi-criteria preferences is to find a compact and expressive representation. Various frameworks have been introduced, each of which with its own distinguishing features. In this paper we introduce a new representation framework called qualitative preference systems (QPS), which combines priority, cardinality and conditional preferences. Moreover, the framework incorporates knowledge that serves two purposes: to impose (hard) constraints, but also to define new (abstract) concepts. In short, QPS offers a rich and practical representation for qualitative, multi-criteria preferences.

Preferences between different alternatives (products, decisions, agreements etc.) are often based on multiple criteria. Qualitative Preference Systems (QPS) is a formal framework for the representation of qualitative multi-criteria preferences in which a criterion's preference is defined based on the values of attributes or by combining multiple subcriteria in a cardinality-based or lexicographic way. In this paper we present a language and reasoning mechanism to represent and reason about such qualitative multi-criteria preferences. We take an argumentation-based approach and show that the presented argumentation framework correctly models a QPS. Then we extend this argumentation framework in such a way that it can derive missing information from background knowledge, which makes it more flexible in case of incomplete specifications.

This paper presents an argumentation-based framework for the modelling of, and automated reasoning about multi-attribute preferences of a qualitative nature. The framework presents preferences according to the lexicographic ordering that is well-understood by humans. Preferences are derived in part from knowledge. Knowledge, however, may be incomplete or uncertain. The main contribution of the paper is that it shows how to reason about preferences when only incomplete or uncertain information is available. We propose a strategy that allows reasoning with incomplete information and discuss a number of strategies to handle uncertain information. It is shown how to extend the basic framework for modelling preferences to incorporate these strategies.

Preferences for objects are commonly derived from ranked sets of properties or multiple attributes associated with these objects. There are several options or strategies to qualitatively derive a preference for one object over another from a property ranking. We introduce a modal logic, called multi-attribute preference logic, that provides a language for expressing such strategies. The logic provides the means to represent and reason about qualitative multi-attribute preferences and to derive object preferences from property rankings. The main result of the paper is a proof that various well-known preference orderings can be defined in multi-attribute preference logic.

Preferences are derived in part from knowledge. Knowledge, however, may be defeasible. We present an argumentation framework for deriving qualitative, multi-attribute preferences and incorporate defeasible reasoning about knowledge. Intuitively, preferences based on defeasible conclusions are not as strong as preferences based on certain conclusions, since defeasible conclusions may turn out not to hold. This introduces risk when such knowledge is used in practical reasoning. Typically, a risk prone attitude will result in different preferences than a risk averse attitude. In this paper we introduce qualitative strategies for deriving risk sensitive preferences.

In the context of practical reasoning, such as decision making and negotiation, it is necessary to model preferences over possible outcomes. Such preferences usually depend on multiple criteria. We argue that the criteria by which outcomes are evaluated should be the satisfaction of a person's underlying interests: the more an outcome satisfies his interests, the more preferred it is. Underlying interests can explain and eliminate conditional preferences. Also, modelling interests will create a better model of human preferences, and can lead to better, more creative deals in negotiation. We present an argumentation framework for reasoning about interest-based preferences. We take a qualitative approach and provide the means to derive both ceteris paribus and lexicographic preferences.

A formal two-phase model of democratic policy deliberation is presented, in which in the first phase sufficient and necessary criteria for proposals to be accepted are determined (the `acceptable' criteria) and in the second phase proposals are made and evaluated in light of the acceptable criteria resulting from the first phase. Such a separation gives the discussion a clear structure and prevents time and resources from being wasted on evaluating arguments for proposals based on unacceptable criteria. Argument schemes for both phases are defined and formalised in a logical framework for structured argumentation. The process of deliberation is abstracted from and it is assumed that both deliberation phases result in a set of arguments and attack and defeat relations between them. The acceptability status of criteria and proposals within the resulting argumentation framework is then evaluated using preferred semantics. For cases where preferences are required to choose between proposals, inference rules for deriving preferences between sets from an ordering of their elements are given.

Negotiation support systems (NSS) help users in the complex process of reaching agreements about exchange of goods or services. A difficult issue in the development of NSS is how to extract knowledge from qualitative real-life data and embed it into the system. We present a metamodel for modeling domain, user and opponent (DUO) in NSS with focus on four main concepts: issues, preferences, interests and objective domain knowledge. We claim that (a) these concepts are essential in extracting data from unstructured sources, and (b) these concepts can be a basis for formal reasoning about user preferences and bids. We ground our meta-model in negotiation literature and data gathered with case studies and interviews. Finally, we formalize parts of the meta-model as a step towards a computationally-oriented model.

No intelligent decision support system functions even remotely without knowing the preferences of the user. A major problem is that the way average users think about and formulate their preferences does not match the utility-based quantitative frameworks currently used in decision support systems. For the average user qualitative models are a better fit. This paper presents an argumentation-based framework for the modelling of, and automated reasoning about multiissue preferences of a qualitative nature. The framework presents preferences according to the lexicographic ordering that is well-understood by humans. The main contribution of the paper is that it shows how to reason about preferences when only incomplete information is available. An adequate strategy is proposed that allows reasoning with incomplete information and it is shown how to incorporate this strategy into the argumentation-based framework for modelling preferences.

In our aim to develop a negotiation support system, we are faced with the need to express a user's preferences. The offers that are exchanged usually consist of multiple attributes. An agreement can only be reached if the preferences from both parties over these attributes and the complete offers are taken into account. The acceptablilty of an offer is defined in terms of the value that a negotiator attaches to it.

A User manual 44 Bibliography 45 Chapter 1 Introduction In this graduation project I have implemented an argument-based practical reasoning system. I will briefly introduce argument-based reasoning and practical reasoning in Sections 1.1 and 1.2 respectively. The main research question is stated in Section 1.3, along with the method that will be used to answer it. Finally, Section 1.4 presents the outline of this thesis. Defeasible reasoning is a way to reason with inconsistent or incomplete belief bases. Its inferences are not absolutely certain, but can still be made, provided that there is no information to the contrary that defeats them. So it can happen that a certain inference can be made from a set of beliefs, but cannot be made anymore if more information becomes known. Hence defeasible reasoning is nonmonotonic. Since this is also the way humans reason in their daily lives, the term commonsense reasoning is often used too. Prakken and Vreeswijk [15] present an extensive overview of defeasible reasoning systems. Argument-based reasoning systems are useful formalisations of defeasible reasoning. They are based on arguments which may contradict each other, for example because they have conflicting conclusions (rebuttal), or one is an argument for the inapplicability of an inference step made in the other (undercut). Arguments can be seen as the defeasible counterpart of proofs in classical logic, but their status is quite different. Classical logic aims to determine the truth value of a given proposition, and one proof is sufficient to do that. De-feasible logic does not search for truth, but rather for having some justification for a given proposition, or more exactly, having more justification for it than against it. So whether a proposition is justified is not determined by a single argument, but by the interaction between multiple arguments for and against the proposition. The basics of defeasible argumentation are well understood. Dung [6] has defined an abstract framework with several semantics that define which arguments are justified (see also Chapter 3, Section 3.2). Proof theories for these semantics in the form of argument games have been developed by Prakken and Sartor [14] and Vreeswijk and Prakken [20] (see also Chapter 3, Section 3.3). Argument-based reasoning systems provide a middle way between classical 2 1. Introduction (monotonic) reasoning systems and statistical methods such as Bayesian networks. Classical logic has the problem that all possible exceptions of a rule have to be listed, and moreover …

Motivation -- Elicitation of preferences is crucial in negotiation support. This is a non-trivial task which could be supported by computers. Research approach -- Experiment in which 32 participants have to order holidays using different preference elicitation techniques including a navigational task and affective scoring. The results were used as input for a lexicographic ordering algorithm. Findings/design -- Traditional property rating approach seems most preferred by the participants and resulted in one of the best orderings of the outcomes space to match their preferences, at least when using the lexicographic algorithm. Originality/value -- The elicitation process is approached from an algorithmic perspective as well as from a user-centred perspective for both navigation and affective attitude. Take away message -- A multi-angle approach gives a richer understanding of the process of preference elicitation.