Towards Ethically Driven Social Service Robots for Older Adults Care

Social service robots (SSRs) deployed in older-adult care face a practical ethical bottleneck: they must make online decisions under partial observability, and those decisions must be auditable by caregivers and institutions. Purely data-driven approaches can be brittle or opaque in safety-critical interactions, while static if–then policies do not scale across contexts. This paper presents a deployable ethical assessment service that operationalizes ethical constraints as machine-checkable, ontology-grounded commitments aligned with the IEEE 7007-2021 ontological view of ethically driven robotics. Beyond rule formalization, the paper main contribution is a runtime architecture that makes ethical reasoning usable in a real robot: (i) an enrichment workflow for under-specified intents that returns the semantic elements required to form a valid ethical query; (ii) a data-property validator that prevents inference on malformed or incomplete context; and (iii) an ethical engine that builds request-specific staging ontologies and performs OWL/SWRL inference (Pellet) without contaminating the base knowledge. The service runs on a virtual machine and is accessed over a LAN via a REST API, enabling lightweight robot integration and centralized rule governance. A restrictive deny-by-default policy turns uncertainty into a safety signal, while structured responses (missing context, inferred elements, and rule-trigger explanations) support transparency and debugging. The results demonstrate the approach implemented on the AMIGA platform with rules for conversation initiation, yielding ethical, unethical, or explicitly inconclusive outcomes depending on contextual completeness.