AI: Diversity in will equal fairness out
The United Nations International Day of Women and Girls in Science (11 February) throws a spotlight on achieving full and equal access and participation for women and girls in science, citing the importance of this goal in global development. The UN has highlighted that over the past decades the global community has made great strides in inspiring and engaging women and girls in science, yet there is still much work to be done.
This is the case in financial services as much as many other sectors. One critical area is artificial intelligence (AI) and how it affects financial decisioning.
AI is becoming pervasive
There’s no contesting the far-reaching growth of AI. From loan applications to fraud prevention, it and machine learning are entrenched in our lives and has a say in the important decisions we make as well as those that are made for us. To make fair and accurate assessments, AI software needs to be reflective of the people it scrutinizes and the best way to achieve this is to have a diverse team at work.
Of course, gone are the days of gender discrimination in financial decisions – it is mandated that risk cannot be measured based on gender. But to achieve the equality that is expected of financial services providers, it is crucial to make it easier for girls and women to enter the sector and further their careers, because one of the real challenges in AI is fighting the bias that can be coded into the models themselves.
All AI models are trained on datasets, and these datasets frequently have coded into them a level of bias. In fact, FICO Chief Analytics Officer Scott Zoldi says, “All data is biased.” It’s up to the data scientists to correct for this, and that is why it is so important to achieve more diverse teams building AI.
Recognizing that we need diversity in innovation and teams is the first step. In many cases, AI learns from data generated by human actions. Left unchecked by data scientists, algorithms can mimic our biases, conscious or not. However, we can mitigate those biases by including people across race, gender, sexual orientation, age, and economic conditions to challenge our own views. By bringing in people with different thoughts and approaches to our own, analytics teams will see a quick improvement in their code.