Large-Scale Group Decision-Making with Linear Uncertain Preference Relations Using Improved Density Peak Clustering and Two-Stage Feedback
This paper proposes a large-scale group decision-making model for linear uncertain preference relations that integrates dynamic trust updating, improved Density Peak Clustering for subgroup identification, and a two-stage feedback mechanism to achieve high consensus rates with minimal preference adjustments.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine a room filled with dozens of people, each holding a unique perspective on how to solve a complex problem. They are not strangers; they know and trust one another to varying degrees, forming a web of relationships that shifts as they talk. The goal is not for everyone to agree instantly, but to find a single, shared path forward that feels acceptable to the whole group. This is the essence of large-scale group decision-making, a challenge that arises whenever organizations, communities, or governments must coordinate the views of many individuals. The difficulty lies in the fact that people's opinions are rarely black and white; they often exist in a gray area of uncertainty, and the trust between them is not static but evolves as they listen to one another. When a group is large enough, simply asking everyone to compromise often fails, because the sheer number of voices creates noise, and rigid methods of forcing agreement can ignore the subtle, changing nature of human relationships.
Researchers have long sought a way to navigate this complexity, particularly when people express their preferences using ranges of uncertainty rather than fixed numbers. In a recent study, a team of scholars tackled this problem by designing a system that treats decision-making as a living process. They focused on a scenario where twenty individuals, acting as decision-makers, needed to prioritize critical success factors for a specific type of project delivery. Instead of treating these twenty people as a static block, the researchers built a model that watches how their trust in one another changes as their opinions shift. They recognized that if two people start to see eye-to-eye, their trust in each other naturally grows, and if they drift apart, that trust wanes. By tracking these fluctuations over time, the model creates a dynamic map of who trusts whom, rather than relying on a snapshot taken at the beginning.
The core of their approach involves grouping these individuals into smaller, more manageable clusters based on how much they trust each other and how similar their views are. To do this, they used an improved version of a clustering technique that looks for "peaks" of density in the data—essentially finding the people who are most central to a group of like-minded individuals. A key innovation in their method is that it does not just look at the current moment; it remembers the history of how trust and opinions have changed in previous rounds. This historical memory helps the system avoid making hasty decisions based on a single, perhaps temporary, shift in opinion. Once the groups are formed, the researchers assigned weights to each person and each group, giving more influence to those who are consistent in their thinking, who have a strong track record of trust, and who belong to a tightly knit cluster.
When the group as a whole fails to reach a high enough level of agreement, the system activates a two-stage feedback process. First, it identifies which smaller groups are holding back the consensus. Instead of demanding that everyone in a struggling group change their mind, the model looks for a way to compensate. If some groups are already in strong agreement, they can effectively "carry" the weight of the groups that are struggling, allowing the threshold for agreement to be adjusted fairly. Second, for the individuals who still need to adjust their views, the system calculates the cost of that change. It considers how much a person has already compromised in the past and how much they trust the people they are being asked to listen to. The goal is to find the path of least resistance, asking only those who can adjust with the least friction to do so, and guiding them toward a specific, trusted reference point rather than a generic average.
The researchers tested this method using a real-world dataset involving twenty decision-makers and five different project dimensions. In the first round of discussion, the group's overall agreement fell just short of the target. The system then identified specific individuals who needed to adjust their views and guided them through a process of negotiation. After just a few rounds of this dynamic clustering and targeted feedback, the group reached a consensus that satisfied the pre-set requirements. The study also ran thousands of computer simulations to see how the model would hold up under different conditions. These tests showed that the method is highly effective under conventional and moderately high consensus requirements, but it faces significant challenges when high consensus requirements are combined with very large group sizes, as reaching agreement becomes increasingly difficult under those specific conditions.
What makes this work significant is not just that it finds a solution, but how it respects the complexity of human interaction. By allowing trust to evolve and by using history to smooth out temporary disagreements, the model avoids the pitfalls of rigid, one-size-fits-all approaches. It acknowledges that in a large group, not everyone needs to change their mind to reach a solution; sometimes, the right people just need to adjust slightly, and the rest of the group can follow. The study concludes that while reaching perfect agreement in very large groups with extremely high standards is difficult, this approach offers a reliable way to achieve a strong, workable consensus with minimal disruption to the individuals involved. It provides a clear, structured way for diverse groups to move from a collection of uncertain opinions to a unified decision, proving that even in a world of shifting trust and vague preferences, a shared path can be found.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.