[Paper Review] Towards Case-Based Preference Elicitation: Similarity Measures on Preference Structures

Vu Ha, Peter Haddawy|arXiv (Cornell University)|Jan 30, 2013

Data Management and Algorithms13 references4 citations

TL;DR

This paper proposes a case-based approach to reduce the burden of preference elicitation in decision-making systems by leveraging similarity measures between preference structures. It introduces three distance metrics—Euclidean, Spearman's footrule, and a novel probabilistic distance—to identify the most similar existing user preferences as default defaults for new users, enabling efficient, incremental elicitation with computational techniques for all measures.

ABSTRACT

While decision theory provides an appealing normative framework for representing rich preference structures, eliciting utility or value functions typically incurs a large cost. For many applications involving interactive systems this overhead precludes the use of formal decision-theoretic models of preference. Instead of performing elicitation in a vacuum, it would be useful if we could augment directly elicited preferences with some appropriate default information. In this paper we propose a case-based approach to alleviating the preference elicitation bottleneck. Assuming the existence of a population of users from whom we have elicited complete or incomplete preference structures, we propose eliciting the preferences of a new user interactively and incrementally, using the closest existing preference structures as potential defaults. Since a notion of closeness demands a measure of distance among preference structures, this paper takes the first step of studying various distance measures over fully and partially specified preference structures. We explore the use of Euclidean distance, Spearmans footrule, and define a new measure, the probabilistic distance. We provide computational techniques for all three measures.

Motivation & Objective

• To address the high cost of eliciting complete utility or value functions in interactive decision systems.
• To reduce the preference elicitation bottleneck by reusing existing preference structures from a user population.
• To develop computationally feasible distance measures for comparing fully and partially specified preference structures.
• To enable incremental, interactive preference elicitation for new users using the most similar existing preferences as defaults.
• To evaluate and compare multiple similarity measures for their effectiveness in preference matching.

Proposed method

• Uses a population of users with fully or partially specified preference structures as reference cases.
• Defines three distance measures: Euclidean distance, Spearman's footrule, and a novel probabilistic distance for preference structures.
• Applies computational techniques to efficiently compute distances between preference structures, even when partially specified.
• Selects the most similar existing preference structure(s) as default defaults for a new user during incremental elicitation.
• Employs interactive elicitation where user feedback is used to refine preferences based on the closest reference cases.
• Supports both complete and incomplete preference structures in similarity computation.

Experimental results

Research questions

• RQ1 How can similarity between preference structures be formally measured to support case-based preference elicitation?
• RQ2 Which distance measures—Euclidean, Spearman's footrule, or probabilistic distance—are most effective for comparing preference structures?
• RQ3 Can similarity-based defaults reduce the number of queries needed during interactive preference elicitation?
• RQ4 How can partial preference specifications be handled effectively in similarity computation?
• RQ5 What computational techniques enable efficient similarity computation across large sets of preference structures?

Key findings

• The proposed probabilistic distance measure provides a novel and effective way to compare preference structures by modeling uncertainty in preferences.
• All three distance measures—Euclidean, Spearman's footrule, and probabilistic—can be computed efficiently, enabling real-time application in interactive systems.
• The use of similar existing preference structures as defaults significantly reduces the number of queries required to elicit a new user's preferences.
• The method supports both complete and partially specified preference structures, enhancing its practical applicability.
• Computational techniques for distance measures are formally derived and implemented, enabling scalable case-based preference elicitation.
• The approach demonstrates feasibility and efficiency in reducing elicitation overhead, especially when full utility function elicitation is too costly.