Browse Prior Art Database

Numeric Estimation of Partial Utility Models for Automated Preference Interviewing Disclosure Number: IPCOM000015921D
Original Publication Date: 2002-Oct-16
Included in the Prior Art Database: 2003-Jun-21
Document File: 3 page(s) / 62K

Publishing Venue



1. Introduction

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 32% of the total text.

Page 1 of 3

  Numeric Estimation of Partial Utility Models for Automated Preference Interviewing

1. Introduction

    Many online systems today attempt to personalize their interaction with users. For example, electronic shops may want to provide decision support to users by recommending products, and portals may try to customize their content in a unique fashion for each user. Personalization requires a model of a user's preferences, and in many cases, preferences are represented in the form of a utility function. Because it is difficult to elicit the information necessary to specify completely a user's utility function, many online systems that employ user modeling must reason with partial utility models; for instance, an online recommender system might have to use incomplete information about a user's preferences when choosing a product to suggest. Although a ranking of alternatives based on partial utility models cannot be certain, such a ranking can sometimes be beneficial. In that case, one requires an easily interpretable numeric estimate of the utilities of alternatives. The proposed defensive estimation technique has the advantages that it is guaranteed to improve in accuracy over time and that its estimates can be clearly understood by the user to represent the minimum utility of an alternative.In the next sections it is first explained a method for defensive utility estimation in more detail, then some limitations and difficulties are discussed, and finally possible extensions are pointed out.

2. Defensive Utility Estimation

    The idea behind defensive utility estimation is straightforward. It is wanted to track the lower bound of the user's utility for each alternative, at least with respect to the model. That is, one tracks the greatest lower bound that can be known with certainty. This can be done for utility functions of any form, whether they are additive or a more general form of a multilinear function, or whether they are too complicated to be decomposed. The focus is on multilinear, and particularly additive, utility functions for which the subutility functions are specified.

    Because of the form of multilinear functions, it is possible to track the lower bound of the utilities for alternatives if one tracks the lower bound of each attribute weight. To do this, it is needed to specify both a way of interpreting unknown weights and a way of adjusting these weights, especially if weights, or the ratings used to compute them, can be changed by an assessment method not just once, but rather can be altered incrementally.

    If an interview uses questions that set weights directly, then to maintain a defensive estimate, one restriction in order to ensure a consistently improving estimate is needed. This restriction is that, if a weight is changed more than once, it is adjusted only in an upwards direction. This is quite easy if the questions asked are of the style, described in Section 3.2, that ask the user to specify p directly, since each...