This paper presents the preference ratios in multiattribute evaluation (PRIME) method which supports the analysis of incomplete information in multiattribute weighting models. In PRIME, preference elicitation and synthesis is based on 1) the conversion of possibly imprecise ratio judgments into an imprecisely specified preference model, 2) the use of dominance structures and decision rules in deriving decision recommendations, and 3) the sequencing of the elicitation process into a series of elicitation tasks. This process may be continued until the most preferred alternative is identified or, alternatively, stopped with a decision recommendation if the decision maker is prepared to accept the possibility that the value of some other alternative is higher. An extensive simulation study on the computational properties of PRIME is presented. The method is illustrated with a re-analysis of an earlier case study on international oil tanker negotiations.

Record linkage is the task of quickly and accurately identifying records corresponding to the same entity from one or more data sources. Record linkage is also known as data cleaning, entity reconciliation or identification and the merge/purge problem. This paper presents the "standard" probabilistic record linkage model and the associated algorithm. Recent work in information retrieval, federated database systems and data mining have proposed alternatives to key components of the standard algorithm. The impact of these alternatives on the standard approach are assessed. The key question is whether and how these new alternatives are better in terms of time, accuracy and degree of automation for a particular record linkage application.

Environmental decision making typically concerns several stakeholders with conflicting views. Multicriteria decision analysis provides transparent ways to elicit and communicate individual preferences. When the stakeholders clearly understand each other's views, a consensus can be reached more easily. Computer software provides a substantial enhancement to support participatory decision making processes, for example, in the preference elicitation and in the analysis of the results. In this paper, we describe the first web-based multicriteria decision support software called Web-HIPRE, and the use of it in participatory environmental modelling. The world wide web provides new possibilities to support the process, for example, by allowing distributed decision support. The stakeholders can be located in different geographical areas, especially in environmental problems. We illustrate the use of Web-HIPRE in a case dealing with the evaluation of regulation policies for the Lake Päijänne in Finland. Decision analysis interviews were carried out and preference models of some typical stakeholders were collaboratively analyzed in order to increase the understanding of other stakeholders' views. This proved to be an applicable approach in trying to reach consensus between the stakeholders. Web-HIPRE also provided a flexible way to allow complementary support via the web.

The thesis focuses on the biases appearing when decision makers are asked to give numerical preference statements on attribute weights in multiattribute value tree analysis (MAVT). The thesis focuses on two problems. First, different attribute weight elicitation methods yield different weights although they are based on the same theoretical principles. Second, the attribute weights change when the structure of a value tree is varied. We run two experiments to study the differences between five different weight elicitation methods (Analytic Hierarchy Process, direct weighting, Simple Multi Attribute Rating Technique, SWING weighting, and tradeoff weighting) and to study how verbal expressions are used in preference elicitation. The weighting methods do yield different weights. These differences originate from the way decision makers restrict their responses depending on the numbers that the methods explicitly or implicitly propose. With the biases related to the structural variation of value trees we first point out that earlier experiments are insufficient because they have drawn conclusions of individual behavior based on averages over large group of subjects. In an experiment we then show that the weights change when the structure of a value tree changes because the subjects again give numbers to describe their preferences so that they clearly favor some response scales. We propose, based on the results from our experiments and earlier observations, that two same origins for many different problems in attribute weighting are that the decision makers give numbers that reflect ordinal information on preferences only and that the weights are normalized to sum up to one. The decision makers' interpretation of the numbers that they use differs from the as...

Abstract. Today’s economy is a service economy, and an increasing number of services is electronic, i.e. can be ordered and provisioned online. Examples include Internet access, email and Voice over IP. Typically, e-services are offered as bundles consisting of more elementary services, offered by different suppliers, forming a network. This allows for best-of-breed solutions, in which the customer selects the best services from different suppliers to satisfy his need, and in which the supplier can focus on his core-competences. The research question is then how to compose such a multi-supplier service bundle. In this paper, we argue that first understanding of the context of the service is important. We propose a framework of ontologies, called e 3 family, which can be used to reason about the contextual socio-economical aspects of e-services. This framework can be used to elicit customer’s need, to compose service bundles satisfying such a need, and to reason about profitability of the found service provisioning network. We illustrate e 3 family by presenting two of its core-ontologies: e 3 value and e 3 service. 1

We propose a method, e 3 service, to reason about satisfying customer needs in the context of a wide choice of multi-supplier ICT service bundles. Our method represents customer needs, their ensuing consequences, and the services that realize those consequences in a service catalogue. This catalogue is then used by a reasoner, which elicits customer needs, computes their consequences, and automatically matches these consequences with services offered by suppliers. The e 3 service method has been implemented and tested in software to demonstrate its feasibility. 1

We investigate the existence of biases, in particular the so called splitting bias, in environmental decision analysis. The splitting bias suggests that presenting an attribute in more detail may increase the weight it receives. We test whether the splitting bias can be eliminated or reduced through instruction and training. Also the effect of the order of the attribute levels in a value tree is analyzed.

Methods for dealing with incomplete preference information in hierarchical weighting models have continued to attract attention in the literature on multi-criteria decision analysis (MCDA). In this paper, we give a structured overview of several such methods which (i) accommodate incomplete preference information, (ii) offer dominance concepts and decision rules for the generation of decision recommendations and (iii) support the iterative exploration of the decision maker's preferences. By doing so, we synthesize much of the relevant literature and provide an integrative perspective on these methods which are here subsumed under the term `preference programming'. We then demonstrate that these methods may outperform conventional decision analyses when the costs of preference elicitation are high and, moreover, provide guidelines for responsible uses of preference programming. We conclude by outlining topics for future research.

In value tree analysis, additive value functions are used to model alternatives' overall values. Difficulties in complete evaluation of alternatives' attribute-specific values (scores) and attributes' relative importance have spurred research on the modeling of incomplete preference information. This thesis shows how incomplete ordinal information, captured as ordinal information where a set of rankings is associated with a set of attributes or alternatives, can be used in preference modeling. The thesis develops a model for computation of non-dominated alternatives and decision recommendations in a setting where incomplete ordinal information about (i) attributes ' relative importance, and/or (ii) alternatives' values with regard to any subsets of attributes, is present. Computational challenges arising from possible non-convexity of the corresponding feasible region of attribute weights and scores are addressed by mixed integer linear programming formulation. Linearity allows also other forms of incomplete information to be used with incomplete ordinal information and thus extends applicability of other preference modeling methods. Incomplete ordinal information can be a useful extension to preference modeling possibilities, including for example settings where: (i) preliminary results or initial screening of alternatives

In this paper, we discuss why surveys are an interesting application of spoken dialog systems from both commercial and research perspectives. We then describe a prototype survey spoken dialog system, the Rate-A-Course system. We show how dialog epiphenomena, including the order in which questions are answered and the duration of respondents ’ answers, can be used to learn information beyond that covered explicitly by survey questions. Index Terms — Speech communication 1.