Preference-Based Optimisation in Group Decision-Making

TL;DR

This paper introduces a preference-based optimisation framework for group decision-making that ensures a single, optimal solution aligned with human preferences, overcoming limitations of traditional multi-objective methods.

Contribution

It develops the ODESYS methodology with the IMAP solver, operationalising preference aggregation to produce unique, group-optimal decisions in complex design problems.

Findings

Successfully maps system behaviour into a unified preference domain.

Achieves a single best-fit solution even in highly constrained scenarios.

Demonstrates applicability through two real-world case studies.

Abstract

Conventional multi-objective optimisation approaches (e.g., MOO-CP or MIP) fail in group decision-making by aggregating heterogeneous objectives without a valid preference foundation, producing Pareto sets instead of a unique actionable decision. As only humans define objectives, preferences constitute the legitimate basis for decision-making. Accordingly, four conditions for complex design-decision systems are established: (1) Preference-Key - all objectives, constraints, and trade-offs are evaluated within a unified preference domain using valid preference function modelling (PFM); (2) Integration - feasible system performance (object capability) and acceptable actor preferences (subject desirability) coexist within a single design-decision space; (3) Association - actors freely specify individual preferences and weights, enabling consistent aggregation towards group-optimal decision-making; and (4) Uniqueness - the solver identifies a single best-fit solution with maximum aggregated preference. The ODESYS methodology, employing the IMAP solver, enables integrated multi-objective design optimisation and multi-criteria decision-making. Its extension within the ODESYS/FIVES formulation broadens applicability while achieving elegant simplicity, explicitly operationalising affine preference aggregation and preserving equivalence with validated ODESYS 1.0 results. By mapping system behaviour into a unified preference-performance domain, ODESYS/FIVES delivers a single best-fit solution, even for highly constrained problems, guaranteeing feasible and acceptable outcomes. Two applications demonstrate transformation of multi-objective optimisation into pure group decision-making, achieving a best-fit-for-common-purpose within socio-physical reach.