Accountable System Design Architecture for Embodied AI: A Focus on Physical Human Support Robots
This academic paper proposes an accountable design architecture for learning-based embodied AI systems, focusing on physical human support robots like standing assistive devices. It argues that because machine learning components act as untrustworthy black boxes, systems must describe their entire structure using SysML and transcribe relevant information tailored to different stakeholders and interfaces. The authors demonstrate this architecture through a developed support robot that provides transparent operational states, user state estimations, and investigation interfaces.
Summary and content fields are drawn from the document text and are pending editorial review.
Who issued it
Who it governs
Engineers, roboticists, and researchers seeking a concrete technical methodology to design transparent and accountable learning-based assistive robots.
What the document contains
The document addresses design architecture, system modeling using SysML, stakeholder-information relationships, interface design media selection, and prototype implementation for learning-based physical support robots.
It does not provide formal legal standards or binding regulatory enforcement mechanisms for robot accidents, nor does it address broad commercial software applications outside of embodied physical robotic devices.
- Learning-based AI robot systems function as opaque black boxes, creating anxiety and trust issues for human users during physical interaction.
- Describing the entire system architecture using Systems Modeling Language (SysML) helps clarify internal processes even when machine learning algorithms remain opaque.
- Accountability is achieved by transcribing the described system into appropriate information tailored to specific stakeholders such as users, engineers, and accident investigators.
- Interface design should align stakeholder needs with information ontology, utilizing visual representations for spatial information and sound for temporal or event-driven feedback.
- Implementation on a physical standing assistive robot confirms that accountable systems can be successfully designed and investigated using the proposed architecture.
· Introduction · Related Works · System Design Concept · Describing Whole System · Representation of Information · Implementation · Conclusion
How it scores
Each score is backed by a quote verified against the document text. Editorial review has not yet confirmed them.
Accountability
4/5Clear responsibility for AI outcomes
Describing the entire system is a necessary first step, and transcribing the described system for each stakeholder based on several principles is effective for achieving accountability.
1. Introduction
screening score · 2026-10-24
Fairness
0/5Non-discrimination and equitable treatment
N/A
screening score · 2026-10-24
Human Dignity
0/5Respect for the inherent worth of persons
N/A
screening score · 2026-10-24
Human Oversight
2/5Humans remain in control of consequential decisions
Clinical image recognition systems detect diagnostic signs based on deep learning, and report to a doctor. Thus, humans can understand classification results and decision making, while conducting the action remains the humans role.
3.1 AI Transparency
screening score · 2026-10-24
Privacy
0/5Protection of personal data and private life
N/A
screening score · 2026-10-24
Transparency
4/5Openness about how AI systems work and are used
In this study, we propose a new design method, focused on accountability and transparency, for learning-based robot systems.
1. Introduction
screening score · 2026-10-24
Enforcement
0/5Consequences and compliance mechanisms with real force
N/A
screening score · 2026-10-24
Impact Assessment
1/5Required assessment of an AI system's risks/impacts before or during use
Physical human-robot interaction comes with safety risks, and therefore humans become anxious when they do not understand them.
1. Introduction
screening score · 2026-10-24