Overview

Teams should develop diversity and inclusion policies and procedures addressing key roles, responsibilities, and processes within the organisations that are adopting AI. Bias risk management policies should specify how risks of bias will be mapped and measured, and according to what standards. 

AI risk practice and associated checks and balances should be embedded and ingrained throughout all levels of the relevant stakeholder organisations. This may require a cultural shift while the AI system is evolving, and an acceptance that neither all questions will be answered necessarily, nor all problems well understood.

AI risk mitigation should not be framed as a quantitative balance but rather as an interdisciplinary qualitative judgement. Stakeholders should consider vetoing a deployment if they foresee potential unintended consequences.

The iterative and continuous AI risk assessment process should be adopted and understood as contributing to an organisationalcultural shift. Risk management should be viewed as being beneficial to the organization. The use of methods not necessarily common in the computer science field such as story-telling could be explored.

Regular risk assessment for diversity and inclusion in an AI system should assess the following points:

• which practices have emerged to date

• which practices seem no longer relevant

• if the initial vision for diversity and inclusivity has been achieved or on-track

• if the use case and user group have been equitably defined or require refining

• whether a proactive approach to inclusivity has been adequately prioritised throughout the development process

• whose perspectives are or were over-represented or under-represented

• if any unforeseen challenges or activities arose and if so, what measures were taken in response

• how the development process has further informed the understanding of any protected or at-risk groups.