US Senators propose ‘Algorithmic Accountability Act’ that requires companies to “self-test” their AI systems for accuracy, fairness, and bias

Yesterday, US senators Ron Wyden, Cory Booker, and Yvette D. Clarke proposed a bill named “Algorithmic Accountability Act”. This bill calls for large companies to assess and fix their algorithms if they show any type of unfairness, inaccuracy, or bias.

While automated decision systems are being used to make some of the most important decisions for us, in some use cases they fail to give the right result because of algorithmic bias. The machine learning algorithms behind these systems are often trained using data that has an underrepresentation of a particular race or sex, which leads to their racist, sexist and non-inclusive behavior.

Sen. Ron Wyden said in an interview with The Associated Press, “Computers are increasingly involved in the most important decisions affecting Americans’ lives — whether or not someone can buy a home, get a job or even go to jail. But instead of eliminating bias, too often these algorithms depend on biased assumptions or data that can actually reinforce discrimination against women and people of color.”

This bill comes at a time when we are witnessing several controversies surrounding AI systems. The list of such controversies just goes on and usually includes the big tech. For instance, Facebook facing a charge of housing discrimination by the Department of Housing and Urban Development and Amazon shutting down its automated recruitment tool after it was found favoring men in industries where they already dominate. Earlier, we have seen facial recognition systems misidentifying black women. Computerized money lending tools often charge higher interest rates to Latino and black borrowers.

What is Algorithmic Accountability Act?

This 15-page bill lists all the “impact assessments” companies are required to perform on their automated decision systems. These assessments will include evaluating the development process, system design, and training data for their accuracy, fairness, bias, discrimination, privacy, and security.

Companies will be required to evaluate a broad range of algorithms, including the systems that attempt to predict and analyze customer’s behavior, involve large amounts of sensitive data, or systematically monitors a large, publicly accessible physical place. If the assessment report shows any risks of discrimination, privacy problem, or other issues, companies will need to address them within a timely manner.

The act targets companies who make over $50 million per year and has access to personal information of at least 1 million customers or devices. It will also be applicable to “data brokers” who, as a major part of their business, collect, assemble, or maintain personal information of individuals.

Once passed, the bill would let the Federal Trade Commission to create rules for companies to check for the concerns in their automated systems and correct them if any problems are found. It also gives FTC the authority to regularly monitor how these automated systems are performing.

Many Twitter users welcomed this much-needed move to fight against tech bias:

You can read the proposed bill here: Algorithmic Accountability Act.

Read Next

Facebook will ban white nationalism, and separatism content in addition to white supremacy content

Facebook takes an initiative against discriminative ads on its platform

The FTC issues orders to 7 broadband companies to analyze ISP privacy practices given they are also ad-support content platforms