Can Artificial Intelligence Discriminate?
- February 1, 2021
- Kimberly Srivastava
- Comments Off on Can Artificial Intelligence Discriminate?
The content on this page is no longer being updated here. For news and updated content you can find it on the Ross & McBride News page.
Can Artificial Intelligence Discriminate? In short, the answer is yes.
Dr. Jutta Treviranus, Director and Founder of the Inclusive Design Research Centre and the Inclusive Design Institute outlines what’s at stake. In this slide show (link), she identifies three areas of discrimination. The first arises from data gaps. Often persons in minority groups are not represented in the data used to program AI technologies. Secondly, the algorithms used to program such technologies may have inherent biases built into them. This could occur from human intervention for example, in how data are labelled. The final source of discrimination is how persons in small minorities are treated in statistical methods. For example, the methods do not account for the discrimination faced by minority groups. Dr. Treviranus indicates that persons with disabilities are the “most vulnerable to data abuse and misuse”.
Dr. Treviranus provides an example of these limitations. She explains that when stress testing self-driving car programs, she programmed a scenario where a person in a wheelchair was driving backwards though the intersection. The self-driving car chose to run over the person despite numerous simulations. Each time, the program anticipated that the person the wheelchair would drive it forward and not backwards. As Dr. Treviranus says, the AI programs “are unable to handle diversity and complexity… they replicate our own inadequacies and automate and amplify them”. Dr. Treviranus stresses that AI technologies need to focus on those who are different. Only then will they be more inclusive and work to decrease discrimination rather than further it.
As AI gets used for more decisions and products, it is imperative that the needs of minority groups are represented. Bakerlaw is currently thinking through ways in which this can be challenged in court to provide greater protections for minority groups most at risk of AI-based discrimination. If you think you have been discriminated against by AI such as in a hiring or promotion decision, contact us (link) to see if we can help.
If you want to learn more about Dr. Treviranus’ important work, you can check out this link where she discusses AI and Inclusive Hiring (link), or check out her profile here (link).