AI Standards: BS 8611 Guide

AI Standards and Regulation

• Emerging ethical standards address the ethical, legal, and societal implications of AI and robotics.
• Standards embody ethical principles, whether explicitly stated or implied.
• Existing standards are still evolving, with limited public information available.

BS 8611 Guide

• Possibly the earliest explicit ethical standard in robotics
• It is titled "Guide to the Ethical Design and Application of Robots and Robotic Systems".
• Serves as guidance for designers to identify potential ethical harm.
• Aids in conducting ethical risk assessments for robots or AI.
• Helps mitigate identified ethical risks.
• Based on 20 distinct ethical hazards and risks.
• Hazards and risks are grouped into societal, application, commercial & financial, and environmental categories.
• Includes advice on measures to mitigate risk impact and how to verify or validate such measures.
• Societal hazards examples: loss of trust, deception, privacy/confidentiality infringements, addiction, job loss.
• Ethical Risk Assessment considers foreseeable misuse, risks causing stress/fear (and their minimization).
• Considers control failure (with psychological effects), reconfiguration, responsibility changes, and application-specific hazards.
• Addresses robots that learn and the implications of robot enhancement.
• The ethical risk of robot use must not exceed that of the same activity by a human.
• Assumes physical hazards imply ethical hazards
• Defines ethical harm as affecting psychological, societal and environmental well-being.
• Physical and emotional hazards have to be balanced against expected user benefits
• Highlights the need to involve the public/stakeholders in robot development.
• Includes key design considerations:
• Robots should not be designed primarily to kill humans.
• Humans remain responsible agents.
• Determining robot responsibility must be possible.
• Robots should be safe and fit for purpose.
• Robots should be designed to not be deceptive.
• The precautionary principle should be followed.
• Privacy should be built into the design.
• Users should not face discrimination or forced robot use.
• Guidelines are provided for roboticists, specifically those conducting research
• Guidelines include engaging the public, considering public concerns, and working with experts from other fields.
• They also include correcting misinformation and providing clear instructions.
• Methods for assuring ethical robot use: user validation, software verification, involvement of other experts, economic and social assessment and legal implications assessment.
• Considers compliance testing against relevant standards.
• Other guidelines/ethical codes (e.g., medical/legal) should be considered where appropriate in design/operation.
• Military application does not absolve humans of responsibility and accountability.

IEEE Standards Association

• Launched a standard via its global initiative on the Ethics of Autonomous and Intelligent Systems.
• Positions 'human well-being' as a core principle.
• Seeks to reposition robotics/AI as technologies enhancing the human condition, not just economic growth.
• Aims to train and empower AI/robot stakeholders to prioritize ethical considerations for humanity's benefit.
• 14 IEEE standard working groups are drafting so-called 'human' standards with AI implications.

IEEE Human Standards

• P7000: Model Process for Addressing Ethical Concerns During System Design
• Aims to establish a process for ethically designing Autonomous and Intelligent Systems.
• P7001: Transparency of Autonomous Systems
• Aims to ensure the transparency of autonomous systems to a range of stakeholders, specifically will address:
• Users: build understanding and trust by ensuring users understand what the system does and why
• Validation & certification: ensure the system is subject to scrutiny
• Accidents: enable accident investigators to undertake investigation
• Lawyers/expert witnesses: ensure ability to give evidence after accidents
• Disruptive technology:enable the public to assess technology and build confidence
• P7002: Data Privacy Process
• Aims for ethical personal data use in software engineering.
• Develops privacy impact assessments (PIA) to identify need/effectiveness of privacy controls.
• Provides checklists for software developers using personal information.
• P7003: Algorithmic Bias Considerations
• Helps algorithm developers make how they sought to eliminate or minimise the risk of bias in their products explicit.
• Addresses overly subjective information use.
• Provides guidelines on communicating the boundaries for which the algorithm has been designed and validated.
• Provides strategies to avoid incorrect interpretation of system outputs by users.
• P7004: Standard for Child and Student Data Governance
• Specifically aimed at educational institutions.
• Provides guidance on accessing, collecting, storing, using, sharing and destroying child/student data.
• P7005: Standard for Transparent Employer Data Governance
• Similar to P7004, but aimed at employers.
• P7006: Standard for Personal Data Artificial Intelligence (AI) Agent
• Describes the technical elements required to create and grant access to personalised AIs.
• This will enable individuals to safely organise and share their personal information at a machine-readable level
• P7007: Ontological Standard for Ethically Driven Robotics and Automation Systems
• Standard brings together engineering and philosophy to ensure that user well-being is considered throughout the product life cycle.
• Identifies ways to maximise benefits and minimise negative impacts
• Considers the ways in which communication can be clear between diverse communities.
• P7008: Standard for Ethically Driven Nudging for Robotic, Intelligent and Autonomous Systems
• Draws on 'nudge theory' to delineate current/potential nudges by robots/autonomous systems.
• Strives to create effective methodologies for the development and implementation of robust, transparent and accountable fail-safe mechanisms.
• P7009: Standard for Fail-Safe Design of Autonomous and Semi-Autonomous Systems
• Strives to create effective methodologies for the development and implementation of robust, transparent and accountable fail-safe mechanisms.
• Will address methods for measuring and testing a system's ability to fail safely.
• P7010: Well-being Metrics Standard for Ethical Artificial Intelligence and Autonomous Systems
• Establishes baseline metrics to assess well-being factors affected by autonomous systems
• Establishes baseline metrics to assess well-being factors for how human well-being could proactively be improved.
• P7011: Standard for the Process of Identifying and Rating the Trustworthiness of News Sources
• Sets out to standardise the processes for assessing the factual accuracy of news stories
• P7012: Standard for Machine Readable Personal Privacy Terms
• To establish how privacy terms are presented and how they could be read and accepted by machines.
• P7013: Inclusion and Application Standards for Automated Facial Analysis Technology
• To provide guidelines on the data used in facial recognition, the requirements for diversity, and benchmarking of applications and situations in which facial recognition should not be used.