TL;DR
This paper introduces a novel way to model group decision-making by defining a "majority preference relation" based on individual decision-makers' cone preference relations. This approach allows for aggregating diverse individual preferences into a collective choice, particularly useful in multi-criteria scenarios, and offers a method to refine group decisions by incorporating "quanta" of information about individual preferences.
What Problem Does This Paper Address?
In real-world scenarios, group decisions often involve multiple conflicting criteria and diverse individual preferences. Aggregating these individual preferences into a coherent group decision is a fundamental challenge in decision theory. Traditional approaches like social welfare functions or Arrow's axioms have limitations. This paper tackles the problem of multicriteria group choice by proposing a model where individual preferences are represented by cone relations. The core challenge is to define a consistent and reasonable "majority preference relation" that reflects the collective will, especially when individual preferences are not perfectly aligned. The paper also addresses how to refine group choices by incorporating specific types of information, termed "quanta," which quantify the compromise an individual is willing to make.
Methods
The paper models individual preferences as cone relations. A cone relation is defined by a convex cone in a vector space, where a preference $y^{(1)} > y^{(2)}$ is equivalent to $y^{(1)} - y^{(2)}$ belonging to this cone. The authors assume these individual cones contain the nonnegative orthant and do not contain the origin, reflecting a desire for improvement across criteria.
The central contribution is the definition of a majority preference relation. This relation is established by considering the preferences of at least half of the decision-makers. Specifically, it's defined as a cone relation where the cone is formed by the union of intersections of subsets of individual preference cones. The paper then investigates the properties of this majority preference relation, particularly its convexity (which implies transitivity).
A key aspect is the introduction of "quanta" of information. These are defined as specific vector conditions that reveal how much compromise an individual DM is willing to make between criteria or groups of criteria. The paper shows how to use these quanta to construct transitive parts of the majority preference cone, thereby refining the group choice and reducing the set of potential solutions. The analysis focuses on the case of three decision-makers and two criteria, exploring different scenarios of how "quanta" information is provided.
Datasets
Not discussed in the paper. The paper focuses on theoretical modeling and uses hypothetical examples rather than real-world datasets.
Key Insights and Ideas
- Cone Relations as a Flexible Preference Model: Representing preferences as cone relations provides a mathematically tractable way to model complex individual preferences in multicriteria settings.
- Majority Preference as a Union of Intersections: The definition of the majority preference relation as a union of intersections of individual preference cones offers a principled way to aggregate preferences, ensuring that a decision is supported by a significant portion of the group.
- Convexity Implies Transitivity: The paper highlights that the convexity of the majority preference cone is directly linked to the transitivity of the majority preference relation, a crucial property for a "reasonable" group choice.
- "Quanta" of Information as a Refinement Tool: The concept of "quanta" is a novel way to quantify individual compromise. Incorporating this information allows for the construction of transitive preference relations that are more specific than the general Pareto relation, leading to a more refined group choice.
- The Majority Cone May Not Be Convex: A significant finding is that the majority preference cone is not always convex, meaning the majority preference relation may not be transitive. This necessitates finding transitive parts of the cone to achieve a well-defined group choice.
- Group Choice as a Reduced Pareto Set: The ultimate goal is to define a group choice that is a subset of the Pareto set of the original problem. By using "quanta" of information, the paper shows how to construct new criteria that lead to a smaller, more refined Pareto set representing the group's collective decision.
Key Results
The paper establishes that the majority preference relation, defined as a cone relation, is irreflexive, invariant under linear positive transformations, and its cone contains the nonnegative orthant. It is shown that the majority preference cone is not necessarily convex, implying the relation may not be transitive.
For the specific case of 3 DMs and 2 criteria, the paper analyzes how "quanta" of information influence the group choice. It demonstrates that:
- When "quanta" information leads to a convex majority cone, the group choice is a unique Pareto set derived from a modified vector criterion.
- In situations where the majority cone is not convex, the group choice might not be unique, and the paper provides methods to identify potential transitive parts of the cone and the corresponding group choices.
- The paper presents theorems detailing how the group choice (represented by a Pareto set of a transformed problem) is determined based on the consistency and nature of the "quanta" of information provided by the DMs.
Limitations and Future Work
- Focus on 3 DMs and 2 Criteria: The detailed analysis is primarily for the specific case of 3 DMs and 2 criteria. Generalizing these results to an arbitrary number of DMs and criteria would be a natural extension.
- Convexity of Individual Cones: The paper assumes individual preference cones are convex and pointed. Investigating scenarios with non-convex or non-pointed individual cones could be a future direction.
- Uniqueness of Transitive Parts: For non-convex majority cones, the paper notes that a maximum transitive part might not be unique. Further research could explore methods for selecting a preferred transitive part when multiple exist.
- Practical Implementation: While the theoretical framework is laid out, practical implementation and testing with real-world decision-making data would be valuable.
- "Quanta" Definition: The paper defines "quanta" in a specific way. Exploring alternative or more general definitions of individual information and compromise could broaden the applicability of the model.
Summary updated