Constructive Preference Elicitation

Abstract

When faced with large or complex decision problems, human decision makers (DM) can make costly mistakes, due to inherent limitations of their memory, attention, and knowledge. Preference elicitation tools assist the decision maker in overcoming these limitations. They do so by interactively learning the DM's preferences through appropriately chosen queries and suggesting high-quality outcomes based on the preference estimates. Most state-of-the-art techniques, however, fail in constructive settings, where the goal is to synthesize a custom or entirely novel configuration rather than choosing the best option among a given set of candidates. Many widespread problems are constructive in nature: customizing composite goods such as cars and computers, bundling products, recommending touristic travel plans, designing apartments, buildings, or urban layouts, etc. In these settings, the full set of outcomes is humongous and can not be explicitly enumerated, and the solution must be synthesized via constrained optimization. In this article, we describe recent approaches especially designed for constructive problems, outlining the underlying ideas and their differences as well as their limitations. In presenting them, we especially focus on novel issues that the constructive setting brings forth, such as how to deal with sparsity of the DM's preferences, how to properly frame the interaction, and how to achieve efficient synthesis of custom instances.

On the equivalence of optimal recommendation sets and myopically optimal query sets

“…(As mentioned earlier, a pairwise comparison is equivalent to a choice query of size k = 2.) Augmenting our analyses to study the regret and/or query complexity of our heuristic elicitation schemes is of tremendous interest; we refer to Dragone et al [46] for excellent progress in that direction.…”

To Ask or Not to Ask: A User Annoyance Aware Preference Elicitation Framework for Social Robots

“…However, these techniques cannot actively learn the unobserved preferences. On the other hand, interactive preference elicitation models can gather full preference information by interacting with the user, but without taking into account user annoyance during the process [8], [9], [10]. Notable exceptions are the work of [5], [13], that integrate user annoyance costs into the elicitation process.…”

“…To overcome this issue, interactive preference elicitation has been proposed as a technique to speed up the learning process (i.e., to proactively ask the user to provide information about their preferences) [7], [8], [9], [10]. This is typically done through some sort of human-robot interaction or feedback framework.…”

Automating Layout Synthesis with Constructive Preference Elicitation

“…In this section we frame custom layout synthesis as a constructive preference elicitation task. In constructive preference elicitation [7], the candidate objects are complex configurations composed of multiple components and subject to feasibility constraints. Choosing a recommendation amounts to synthesizing a novel configuration that suits the user's preferences and satisfies all the feasibility constraints.…”