Preference Elicitation for Group Decisions Using the Borda Voting Rule
- PDF / 870,533 Bytes
- 19 Pages / 439.37 x 666.142 pts Page_size
- 67 Downloads / 216 Views
Preference Elicitation for Group Decisions Using the Borda Voting Rule Lihi Naamani-Dery · Inon Golan · Meir Kalech · Lior Rokach
© Springer Science+Business Media Dordrecht 2015
Abstract This paper addresses the issue of preference elicitation for group decision making using voting rules. We propose a general, domain-free framework for preference management, where the goal is to minimize the communication cost with the users. We introduce novel heuristics and show how they can operate under a ranking voting protocol, specifically under the Borda protocol. We suggest an interactive incremental framework where at each step one user is queried for her ranking order of two items. We propose two approaches for heuristics that determine what query to select next (i.e., whom to query regarding what item or items). One heuristic computes the information gain of each potential query. The other heuristic uses the probability distribution of the voters’ preferences to select the candidate most likely to win and the voter that is expected to maximize the score of that item. Both heuristics rely on probabilistic rating distributions. We show how these distributions can be estimated. The rating distributions are updated iteratively, allowing their accuracy to increase over time. We demonstrate the effectiveness of our framework by evaluating the different heuristics on two real-world datasets. Keywords
Preference elicitation · Social choice · Decision support systems
L. Naamani-Dery (B) Industrial Engineering and Management Department, Ariel University, Ariel, Israel e-mail: [email protected] I. Golan · M. Kalech · L. Rokach Information Systems Engineering Department, Ben Gurion University, Beer-Sheva, Israel e-mail: [email protected] M. Kalech e-mail: [email protected] L. Rokach e-mail: [email protected]
123
L. Naamani-Dery et al.
1 Introduction Decision systems can assist group members to reach a joint decision. An example of such a task is to assist a group of friends who wish to find a restaurant for a dinner party. The process is easy when the group members state all of their personal preferences for the items in question (restaurants in this case). In this case a decision can be reached by picking the most preferred item. However, a requirement to state complete preferences for all items in question may be viewed as disruptive by the users as this can be difficult or time consuming. Fortunately, not all user preferences are needed in order to reach a “winning item” i.e., an item that certainly suits the group’s preferences and can be regarded as a group decision (Konczak and Lang 2005). We propose an incremental elicitation process that can be employed in order to reduce the number of questions that the users are asked in order to find a winning item. In this process, the users state only some of their preferences, and only when explicitly queried for them. Studies have shown it is easier for users to state opinions when the queries are pairwise (Balakrishnan and Chopra 2012). Consider, for example an application tr
Data Loading...
