Research conducted by Dr Joanna Bryson in collaboration with Aylin Caliskan and Arvind Narayanan from Princeton University has shown that applying machine learning to text taken from the Internet results in AI with human-like semantic biases.

These biases can be morally neutral as toward insects or flowers, problematic as toward race or gender, or even simply veridical, reflecting the status quo distribution of gender with respect to careers or first names.

Their finding indicates that standard written and spoken words contain recoverable and accurate imprints of our historical biases. It also holds the promise for identifying and addressing sources of bias in our culture and technology.

In her talk, Dr Bryson will present her research on machine bias, and discuss what this reveals about the origins of human biases, stereotypes, and prejudices. Joanna will also explore how implicit and explicit human bias accounts for bias in AI, and discuss when and how we may be able to use AI to address prejudice.