AI systems may have multiple stakeholders, depending on what the system is required to do and where it will be operating. JSP 936 Part 2 Appendix B has a Stakeholder Identification Tool that can assist in working out what is appropriate to consider for your system using the following headings:
Creators
Operators
Executors
Decision subjects
Data subjects
Examiners
Creators
Creators represent those agents that create the system. This category is broader than just coders and includes a wide variety of aspects of the system, including documentation, training and maintenance. It also includes the owners of the systems and those responsible for procurement.
- Example creators: Data scientists, data engineers, procurement managers, intellectual property (IP) owners, data architects, training development coordinators
Operators
Operators are the agents that interact directly with the system by providing inputs and receiving outputs. They may be able to interact with creators. These are the core users of the system.
- Example operators: Intelligence analysts, UXV operators, logisticians.
Executors
Executors are agents make decisions informed by the AI system. This means they are not always distinct from the operator.
- Example executors: Intelligence analysts, operational commanders, tactical operators.
Decision-subjects
Decision subjects are agents who are affected by decisions made by executors. The ability to engage with these stakeholders will vary based on the specific system. For instance, decision-subjects for an AI-enhanced human resources (HR) system used to allocate career postings would be (relatively) straightforward to engage given they are MOD employees. However it may be impossible to directly engage with decision-subjects of an AI-enabled intelligence, surveillance and reconnaissance (ISR) system, where they are part of a military adversary.
- Example decision subjects: Adversary combatants, civilians living in the area of operations, MoD personnel.
Data-subjects
Data-subjects are agents whose data have been used to train the AI system (in the case of machine learning). Where personal data is collected, stored and processed, more stringent ethical and legal requirements apply to protect individuals; particularly, compliance with the UK General Data Protection Regulation.
- Example data subjects: Adversary combatants, individuals posting on social media, individuals whose images are included within open source databases.
Examiners
Examiners are agents who audit, test or investigate an AI system. This may involve them also being responsible for other roles in the system. It is not unusual, for instance, for the creator to also undertake (or support) the testing, evaluation, validation and verification of a system.
- Example examiners: Data scientists, lawyers, policy advisors.
JSP 936 Part 2 Appendix B also has a number of useful suggestions for how to engage with these different stakeholders, depending upon what outcomes you are hoping to achieve and answers you are hoping to generate.
4
4
4
4