AI governance system

Organisations often consolidate their ethical initiatives under the guidance of a chief AI ethics officer (CAIEO).

The role of CAIEO is to:

• provide advice to CEOs and boards about the undesirable risks of AI usage and develop accountability frameworks

• integrate corporate values and ethical principles related to AI across the organisation's divisions

• communicate with internal and external stakeholders about AI governance and promote changes and progress in AI ethics

• ensure the responsible use and development of AI technology

• assist developers in developing AI technology with the right tools, education and training so that the values and principles underlying the technology can be appropriately embedded.

(Minevich et al., 2021); (Stirrup, 2022)

CAIEOs should have a multidisciplinary background in AI, social science, technology law, and business to help technology departments and business leaders mitigate the risks of AI by aligning the organisation's business strategy with ethical practices. CAIEOs in organisations can centralise decision-making regarding ethical considerations and principles around AI development and use.

This allows organisations to quickly develop policies around ethics while ensuring accountability for each decision. Furthermore, CAIEOs can communicate to the public how their company engages with AI and other technologies responsibly and ethically.

The ethics office is responsible for providing advice and assistance to the executives and staff members on matters of ethics and conduct. It consists of a team of members from different levels of the organisation, with a common goal of ensuring that ethical principles are followed and enforced throughout the company. Ethics offices can be independent or part of a broader risk, compliance, or legal team. Organisations can establish ethics offices without having a leading role, but their CAIEOs or CEOs (chief ethics officers) can also take on that role. In any case, an ethics office should:

• perform its functions impartially

• maintain a strict code of confidentiality

• provide service to all employees, regardless of their position, role, or contract status.

Ethics committees and advisory boards can be established for responsible AI to develop standard decision-making processes and approve and monitor AI projects. The main difference between ethics offices and ethics committees is that ethics committees are usually formed by people from various fields of expertise, such as ethics, law, and technology. The ethics committee develops and implements policies, standards, and best practices for responsible AI in collaboration with other groups within the organisation.

The benefits of having an ethics committee are:

• Internal governance is enforced: The ethics committee ensures that responsible AI governance is established at the organisational level.

• Feedback and guidance without bias: Project teams can receive feedback and guidance on how to further mitigate risks or maximise benefits using AI.

• Diverse expertise and versatile risk management: Diversity of expertise and perspectives in the committee can help the organisation identify and address complex issues related to AI technology and ethics.

• Enhance public and internal trust in AI products and services: Ethics committees can increase transparency in how AI is used in the organisation and demonstrate the organisation's values and proactive approach to advancing AI governance.

(CSIRO, n.d.)

In a centralised AI governance model, an organisation's governance, standards, strategy team, data scientists and architects are brought in one central team. Other business units of the organisation can then pull the people they need from the team and extract knowledge to initiate their own innovation. 'Centralisation' in this context indicates the degree to which the coordination, oversight and regulation of a set of AI policy issues or technologies are housed under a single institution. (Cihon et al., 2020)

Centralising AI governance is known to have the following benefits:

• Consistent governance practices: All personnel with the necessary skills and knowledge for AI development and governance are in one team, which enables strong central control over AI governance policies and processes. This ensures that AI governance practices are shared and consistent across other business units of the organisation.

• Efficiency and participation:  The specialised governance team oversees and enforces governance across an organisation. This means all departments within the business follow the same policies and procedures, which is more efficient than having each department develop its own AI governance. A centralised team ensures every department participates in, and complies with, the security and technical requirements for good AI governance.

• Knowledge sharing: Governance teams work closely together, and their AI governance solutions and processes are compatible. Therefore, centralised governance makes knowledge sharing convenient and efficient.

(Cihon et al., 2020)

In the decentralised model, every business unit within an organisation is empowered with the skills they need to innovate rapidly in their own silos. Decentralised teams delegate or distribute their activities away from a central, authoritative location or group, particularly those related to planning and decision-making. Therefore, decentralisation is a practical approach when different departments or individuals in an organisation have different AI needs, responsibilities, and strategies. Decentralisation can bring the following benefits to organisations:

• Adaptability and speed: Since decentralised business units are not subject to the bureaucratic requirements of centralised governance, they can adapt and speed up as new developments occur in both technical and regulatory spaces.

• Creativity: Decentralised business units are free to develop approaches that reflect their own needs and environments without worrying about how responsibilities and governance activities are distributed.

The hybrid model is a combination of the centralised and decentralised models. A team (with personnel essential for setting AI standards, strategies and governance) develops principles to drive innovation. This 'governance' team works with other business units, such as marketing, sales and finance. The technical capability team, which consists of experts such as data scientists, works within each business unit to help them realise innovations while following standards and strategies. A community develops around this structure for further governance activities and innovations.

Organisations can benefit from using the hybrid model through:

• Knowledge sharing: By having a community around the governance structure, departments can easily share knowledge and learnings. This is also sufficiently supported by the common platform shared across the business units, which ensures that everyone follows and maintains the central standards consistently.

• Flexibility and adaptability: In the rapidly changing environment of AI, hybrid models provide better flexibility and adaptability. Having the right balance between the 2 models of governance can better accommodate new initiatives and can also evolve into the model that is most suitable for the organisation's needs and initiatives.