The AI Bill of Rights: Should People Have the Right to Challenge Automated Decisions?

Earlier this month, the White House Office of Science and Technology Policy published the Blueprint for an AI Bill of Rights. It sets out five principles for automated systems that affect people’s lives.
Safe and effective systems. People should be protected from systems that are unsafe or do not work as intended.
Algorithmic discrimination protections. Systems should be designed and used in an equitable way.
Data privacy. People should have agency over how their data is used.
Notice and explanation. People should know when an automated system is being used and understand how it affects them.
Human alternatives, consideration and fallback. People should be able to opt out where appropriate and reach a human who can consider and remedy problems.
The document is not legally binding. Critics on one side call it toothless. Critics on the other call it a first step toward heavy regulation. Both miss what I think is most interesting about it.
A shift in perspective
Most AI governance so far has been written from the perspective of the organization deploying the system. What risks does it create for the company? What controls should the company have?
The Blueprint is written from the perspective of the person affected. It asks what rights an individual should have when an automated system decides something about their job, their loan, their benefits or their medical care.
That is a significant change, and Europe is already further down the same road. GDPR’s Article 22 gives people the right not to be subject to decisions based solely on automated processing with significant effects, and to obtain human intervention and contest the decision. The proposed EU AI Act builds on similar ideas.
Why the last principle is the hardest
For businesses, the most demanding principle is the last one: a human alternative and a way to remedy problems.
It sounds simple. In practice it requires an organization to maintain people who understand the automated process well enough to review it, the authority to override it and the capacity to handle appeals at scale. Many companies automated precisely to remove that cost.
There is a real tension here. The economic case for automation often depends on removing human review. The ethical and increasingly legal case depends on keeping it, at least where decisions matter.
Accuracy is not the whole answer
The question the Blueprint poses to business is whether a technically accurate algorithm is sufficient. My answer is no, for a practical rather than philosophical reason. Even a very accurate system will be wrong for some people. If those people have no way to understand or challenge the decision, each error becomes a potential complaint, a regulatory case or a news story.
The ability to appeal is not only a right for the individual. It is an error correction mechanism for the organization. Companies that design it well learn faster where their systems fail.
Leadership question: Should individuals have the right to understand, question and appeal decisions made by artificial intelligence?


