Local utility elicitation in GAI models

Proceedings of the Thirteenth Conference on …, 1997

Decision theory has become widely accepted in the AI community as a useful framework for planning and decision making. Applying the framework typically requires elicitation of some form of probability and utility in-formation. While much work in AI has fo-cused on providing representations ...

1998

Abstract We investigate the application of classification techniques to utility elicitation. In a decision problem, two sets of parameters must generally be elicited: the probabilities and the utilities. While the prior and conditional probabilities in the model do not change from user to user, the utility models do. Thus it is necessary to elicit a utility model separately for each new user. Elicitation is long and tedious, particularly if the outcome space is large and not decomposable. There are two common approaches to utility function elicitation.

Journal of Risk and Uncertainty

This paper is about behaviour under ambiguity-that is, a situation in which probabilities either do not exist or are not known. Our objective is to find the most empirically valid of the increasingly large number of theories attempting to explain such behaviour. We use experimentally-generated data to compare and contrast the theories. The incentivised experimental task we employed was that of allocation: in a series of problems we gave the subjects an amount of money and asked them to allocate the money over three accounts, the payoffs to them being contingent on a 'state of the world' with the occurrence of the states being ambiguous. We reproduced ambiguity in the laboratory using a Bingo Blower. We fitted the most popular and apparently empirically valid preference functionals [Subjective Expected Utility (SEU), MaxMin Expected Utility (MEU) and α-MEU], as well as Mean-Variance (MV) and a heuristic rule, Safety First (SF). We found that SEU fits better than MV and SF and only slightly worse than MEU and α-MEU.

2010

Abstract Most frameworks for utility elicitation assume a predefined set of features over which user preferences are expressed. We consider utility elicitation in the presence of subjective or user-defined features, whose definitions are not known in advance. We treat the problem of learning a user's feature definition as one of concept learning, but whose goal is to learn only enough about the concept definition to enable a good decision to be made. This is complicated by the fact that user utility is unknown.

Proceedings of the Twenty-third Conference on …, 2007

We describe the semantic foundations for elicitation of generalized additively independent (GAI) utilities using the minimax regret criterion, and propose several new query types and strategies for this purpose. Computational feasibility is obtained by exploiting the local GAI structure in the model. Our results provide a practical approach for implementing preference-based constrained configuration optimization as well as effective search in multiattribute product databases.

Conjoint Measurement, 2001

Conjoint Preference Elicitation Methods in the Broader Context of Random Utility Theory Preference Elicitation Methods Hensher, Louviere & Swait 2 but some are not, some methods permit one to combine experimental with actual marketplace choice data but many do not, etc. Indeed, there are many ways to understand and model preferences and choices, some of which bear scant relation to one another and others that are incompatible, both theoretically and analytically. RUT offers a way to unify many seemingly disparate approaches to understand and model preference formation and choice. Figure represents an overview of the general problems covered by RUT, which can assist our understanding of the role of conjoint analysis methods within RUT. Figure should be regarded as a pedagogical vehicle to help explain why a more general view is required; it is not a theory per se.

2021

In this paper we propose efficient methods for elicitation of complexly structured preferences and utilize these in problems of decision making under (severe) uncertainty. Based on the general framework introduced in Jansen, Schollmeyer & Augustin (2018, Int. J. Approx. Reason), we now design elicitation procedures and algorithms that enable decision makers to reveal their underlying preference system (i.e. two relations, one encoding the ordinal, the other the cardinal part of the preferences) while having to answer as few as possible simple ranking questions. Here, two different approaches are followed. The first approach directly utilizes the collected ranking data for obtaining the ordinal part of the preferences, while their cardinal part is constructed implicitly by measuring meta data on the decision maker’s consideration times. In contrast, the second approach explicitly elicits also the cardinal part of the decision maker’s preference system, however, only an approximate ve...

Scalable Uncertainty Management, 2019

Preference elicitation is a key element of any multi-criteria decision analysis (MCDA) problem, and more generally of individual user preference learning. Existing efficient elicitation procedures in the literature mostly use either robust or Bayesian approaches. In this paper, we are interested in extending the former ones by allowing the user to express uncertainty in addition of her preferential information and by modelling it through belief functions. We show that doing this, we preserve the strong guarantees of robust approaches, while overcoming some of their drawbacks. In particular, our approach allows the user to contradict herself, therefore allowing us to detect inconsistencies or ill-chosen model, something that is impossible with more classical robust methods.

European Journal of Operational Research, 2015

This work addresses the early phases of the elicitation of multiattribute value functions proposing a practical method for assessing interactions and monotonicity. We exploit the link between multiattribute value functions and the theory of high dimensional model representations. The resulting elicitation method does not state any a-priori assumption on an individual's preference structure. We test the approach via an experiment in a riskless context in which subjects are asked to evaluate mobile phone packages that differ on three attributes.

RePEc: Research Papers in Economics, 2013

This work addresses the early phases of the elicitation of multiattribute value functions proposing a practical method for assessing interactions and monotonicity. We exploit the link between multiattribute value functions and the theory of high dimensional model representations. The resulting elicitation method does not state any a-priori assumption on an individual's preference structure. We test the approach via an experiment in a riskless context in which subjects are asked to evaluate mobile phone packages that differ on three attributes.

Journal of Choice Modelling, 2008

We show how to combine statistically efficient ways to design discrete choice experiments based on random utility theory with new ways of collecting additional information that can be used to expand the amount of available choice information for modeling the choices of individual decision makers. Here we limit ourselves to problems involving generic choice options and linear and additive indirect utility functions, but the approach potentially can be extended to include choice problems with non-additive utility functions and non-generic/labeled options/attributes. The paper provides several simulated examples, a small empirical example to demonstrate proof of concept, and a larger empirical example based on many experimental conditions and large samples that demonstrates that the individual models capture virtually all the variance in aggregate first choices traditionally modeled in discrete choice experiments.

European Journal of Operational Research, 2014

Many methods to elicit preference models in multi-attribute decision making rely on evaluations of a set of sample alternatives by decision makers. Using orthogonal design methods to create this set of alternatives might require respondents to evaluate unrealistic alternatives. In this paper, we perform an empirical study to analyze whether the presence of such implausible alternatives has an effect on the quality of utility elicitation. Using a new approach to measure consistency, we find that implausible alternatives in fact, have a positive effect on consistency of intra-attribute preference information and consistency with dominance, but do not affect inter-attribute preference information.

In the context of eliciting preferences for decision making under risk, we analyse the features of four different elicitation methods-pairwise choice, willingnessto-pay, willingness-to-accept, and the Becker-DeGroot-Marschak mechanism-and estimate noise, bias and risk attitudes for two different preference functionals, Expected Utility and Rank-Dependent Expected Utility. It is well-known that methods differ in terms of the bias in the elicitation; it is rather less well-known that methods differ in terms of their noisiness. It has also been reported that risk attitudes are not stable across different elicitation methods. Our results suggest that elicited preferences should only be used in the context in which they were elicited, and the bias in the certainty-equivalent methods should be kept in mind when making predictions based on the elicited preferences. Moreover, conclusions should be moderated to take into account the various methods' noise, which is generally lowest in the case of pairwise choice.

2003

When searching for multi-attribute services or products, understanding and representing user’s preferences is a crucial task. However, many computer tools do not afford users to adequately focus on fundamental decision objectives, reveal hidden preferences, revise conflicting preferences, or explicitly reason about tradeoffs with competing decision goals. As a result, users often fail to find the best solution. From building decision support systems for various application domains, we have observed some common areas of design pitfalls, which could lead to undesirable user behaviors and ineffective use of decision systems. By incorporating findings from behavior decision theory, we have identified and accumulated a set of principles for avoiding these design pitfalls: 1) provide a flexible order and choice in preference elicitation so that users can focus on fundamental objectives, 2) include appropriate information in a decision context to guide users in revealing hidden preferences...

2021

Identifying the preferences of a given user through elicitation is a central part of multi-criteria decision aid (MCDA) or preference learning tasks. Two classical ways to perform this elicitation is to use either a robust or a Bayesian approach. However, both have their shortcoming: the robust approach has strong guarantees through very strong hypotheses, but cannot integrate uncertain information. While the Bayesian approach can integrate uncertainties, but sacrifices the previous guarantees and asks for stronger model assumptions. In this paper, we propose and test a method based on possibility theory, which keeps the guarantees of the robust approach without needing its strong hypotheses. Among other things, we show that it can detect user errors as well as model misspecification.