Multi-User Personalisation in Human-Robot Interaction: Resolving Preference Conflicts Using Gradual Argumentation

TL;DR

The paper tackles the challenge of multi-user personalization in HRI by introducing MUP-QBAF, a gradual-argumentation framework that models both positive and negative user preferences and robot-observed context. It extends Quantitative Bipolar Argumentation Frameworks to handle multiple stakeholders, delivering a transparent, adaptable decision process without retraining. The authors formalize the framework, present a multi-user preference selector algorithm, and demonstrate its application to frailty assessment tasks, including dynamic adaptation and sensitivity analyses. The work contributes a principled, explainable alternative to data-driven methods for resolving conflicts in real-world, multi-user robotic environments.

Abstract

While personalisation in Human-Robot Interaction (HRI) has advanced significantly, most existing approaches focus on single-user adaptation, overlooking scenarios involving multiple stakeholders with potentially conflicting preferences. To address this, we propose the Multi-User Preferences Quantitative Bipolar Argumentation Framework (MUP-QBAF), a novel multi-user personalisation framework based on Quantitative Bipolar Argumentation Frameworks (QBAFs) that explicitly models and resolves multi-user preference conflicts. Unlike prior work in Argumentation Frameworks, which typically assumes static inputs, our approach is tailored to robotics: it incorporates both users' arguments and the robot's dynamic observations of the environment, allowing the system to adapt over time and respond to changing contexts. Preferences, both positive and negative, are represented as arguments whose strength is recalculated iteratively based on new information. The framework's properties and capabilities are presented and validated through a realistic case study, where an assistive robot mediates between the conflicting preferences of a caregiver and a care recipient during a frailty assessment task. This evaluation further includes a sensitivity analysis of argument base scores, demonstrating how preference outcomes can be shaped by user input and contextual observations. By offering a transparent, structured, and context-sensitive approach to resolving competing user preferences, this work advances the field of multi-user HRI. It provides a principled alternative to data-driven methods, enabling robots to navigate conflicts in real-world environments.

Figures (6)

• Figure 1: Iterative closed-loop multi-user preference selection system. The robot processes observations (Robot Observations Module), such as environmental states and user inputs to generate tasks and users' arguments, and the possible decisions (their preferences), which are stored in the Arguments Module. These are structured in the Preference Selector Module, where the arguments are combined into support/attack relations. The MUP-QBAF (Argumentation Framework) outputs a selected preference that directs robot action. Importantly, the environmental feedback from this action generates new observations, iteratively updating the argumentation framework.
• Figure 2: Argumentation Framework Examples. In a), A is attacking B and B is attacking C; therefore, A is defending C. In b), A is attacking B, and C is supporting B.
• Figure 3: Example of a simple QBAF. The nodes contain the argument name (A, B, C), the base score on top, and the final strength in bold at the bottom. Continuous lines are attacks, and the dashed line is a support.
• Figure 4: An example of the initial AF for determining whether to repeat a test in frailty assessments is shown on the left, along with the associated arguments on the right. The squared arguments highlighted in red represent the Option Arguments (to repeat or not to repeat the test). The solid lines indicate attacks on the arguments, while the dashed lines represent supports. In this context, $T$ denotes task arguments, $CR$ refers to care recipient arguments, and $CG$ signifies caregiver arguments. The numbers following each argument serve to enumerate them. The arguments positioned above the Option Arguments belong to the risk of falling observations, while those below are associated with the users.
• Figure 5: Representation of Example \ref{['example:frailty_adding_arguments']}. The caregiver and care recipient give their arguments to the robot, and those are activated (in blue and bold) in the Argumentation Framework. R and $\neg$R are the Option Arguments for repeating or not a test. Initially, the robot decides to repeat the tests, indicated by a rectangular argument in green. However, after activating arguments T1 and T3 due to an imbalance detected in the standing balance test, the robot changes its decision and decides not to repeat the tests.

Theorems & Definitions (14)

• Definition 1
• Definition 2
• Definition 3
• Definition 4
• Definition 5
• Definition 6
• Definition 7
• Definition 8
• Proposition 1
• proof