Fair and Responsible AI: A Focus on the Ability to Contest

Contestation is a critical safeguard in AI decision-making: As AI systems increasingly make high-stakes decisions affecting people's lives (such as loan approvals or hiring), the ability to challenge these decisions is recognized as essential in AI ethics guidelines, yet remains underimplemented in practice.

How contestation is designed matters as much as whether it exists: The paper finds that the specific design of contestation processes significantly impacts how fair people perceive the system to be and their overall satisfaction with outcomes, meaning poorly designed appeal mechanisms can undermine trust even when they technically exist.

Transparency barriers complicate contestation: A major design challenge is that lack of transparency in AI systems makes it difficult for people to understand why a decision was made, which is a prerequisite for effectively contesting it.

One-size-fits-all approaches won't work: Organizations face numerous design choices when building contestation processes, and the paper argues these decisions should be guided by the specific needs of affected individuals and communities rather than generic templates.

Human-centered design is essential: Rather than treating contestation as a technical compliance checkbox, the authors advocate for involving actual decision subjects and affected communities in designing how people can challenge AI decisions.