T49. Transparency, explainability and contestability expectation canvassing

Navigation: AIGA AI Governance Lifecycle > E. Transparency, explainability, contestability > T49. Transparency, explainability and contestability expectation canvassing (this page)

Task description

AI systems will be subject to varying stakeholder transparency, explainability, and contestability requirements.

Some of the expectations may be regulatory in origin. GDPR, for example, imposes a right to explanation for automated decisions.

Internal and external stakeholders also impose TEC requirements on the system. For example, recommendation engine users may require information on the system logic to be able to assess the accuracy and appropriateness of the recommendations.

To prevent investment slippage, the AI System Owner should ensure that the organization conducts a TEC requirement canvassing at the outset of the development process. In the canvassing, the organization should identify relevant TEC stakeholders and TEC expectations.

Identifying regulatory transparency and explainability expectations typically requires input from legal experts and is a part of the compliance process. Identifying non-regulatory TEC stakeholder expectations may benefit from stakeholder consultations or co-creation efforts.