Multiobjective Optimisation for Others: How Anchoring Effects Change Based on Who Guides the Interaction

It is common in many real‐world scenarios for decisions to be made on behalf of others. However, Multicriteria Decision Aiding methods often implicitly assume that the person whose preferences operate the methods, known as the decision‐maker, is the source of those preferences. This assumption could influence interactive Multiobjective Optimisation, where the Pareto‐optimal solutions are not known in advance and their discovery depends on continual feedback from the decision‐maker. Due to this iterative nature, anchoring—defined as initial information influencing subsequent choices—could cause the decision‐maker to stop interacting before they find the best solution according to the preference information, the Most Preferred Solution. Surprisingly, there are very few behavioural studies of anchoring and Decision‐Making‐for‐Others in Multicriteria Decision Aiding and even fewer that have considered interactive Multiobjective Optimisation. We address this gap by studying whether anchoring in interactive Multiobjective Optimisation differs when the decision‐maker uses their own or someone else's preferences. We conducted a behavioural experiment with 128 participants that acted as decision‐makers using an interactive goal programming task that does not restrict the participant's choices to solve a two‐objective problem. We found that anchoring arises in interactive Multiobjective Optimisation; it causes decision‐makers to miss the Most Preferred Solution, and that it is stronger when the decision‐maker uses someone else's preferences. These findings indicate that interactive methods should account for preference source and initialisation when designing elicitation and search methods.