Our team developed an algorithm that works well but has clear bias in how it treats different demographic groups, and we're split on whether we have an obligation to fix it before launch when the client hasn't asked for that. Some people say bias is just reflecting reality in data, others say we're actively perpetuating harm if we don't address it.
More people experience this than they realize.
A group creating something together must decide what obligations they have to prevent their work from perpetuating systemic inequities.
“Where Are You with Rights — their origin and their limits?”
If this sounds familiar, the Library can help you find the bigger picture.