Artificial intelligence (AI) is used so widely that most of us just assume it’s accurate. But the algorithms that inform AI are created by humans, who by their very nature are biased. This means AI could create damaging outcomes for insurance customers.

In 2019, the University of California, Berkley, published a study that found a widely used software program helping to guide care for millions of patients was flawed by unintentional racial bias.

Biased software passed over blacks

The program used artificial intelligence (AI) to generate commercial health risk scores for around 70 million patients.

Higher scores indicated the likelihood of complex and intensive healthcare needs in the future and recommended those patients be enrolled in a care management program that would provide them with greater attention from healthcare specialists.

But the researchers found that the software had been prioritising white patients for the care management program and passing over black patients who were less healthy.

Premium

You need to login to access this

Login

What are CIP Points? About ANZIIF Membership