AI companies have faced public scrutiny in recent times for developing machine learning algorithms that exhibit bias against historically marginalized groups. In response to these concerns, many companies have made commitments to ensure fairness, transparency, and accountability in their products. However, critics often dismiss these promises as mere “ethics washing.”

To investigate the veracity of these criticisms, researchers from Stanford University conducted interviews with AI ethics professionals employed by some of the leading companies in the field.

The study revealed that implementing ethics initiatives and interventions in the tech industry’s institutional framework was a challenging endeavor. The research found that these ethics teams faced significant resource constraints and lacked sufficient support from leadership. Moreover, they encountered obstacles in exerting authority to address the problems they identified.

Doing better

Various things emerged that point to a better approach, however, including the incentivization of product teams to ensure that ethical considerations are factored in from the start of the product development process.

What’s more, ethics teams should have full bureaucratic support and authority to ensure that they can implement any ethical fixes required before the product is launched onto the market.

“It’s unlikely that these companies are going to change their priority of frequently releasing new products,” the authors explain. “But at least they could provide incentives so that ethics can be part of that conversation early on.”

Certain companies have adopted a formal ethics review process to address potential ethical implications during the early stages of new product development. Product teams are required to conduct an impact assessment, subject to scrutiny by the ethics team.

For products associated with sensitive applications like bail or sentencing, joint collaboration between the ethics and product teams takes place to evaluate the possibility of treating specific demographic groups unfairly. Should such potential concerns arise, the ethics team assumes an integral role in overseeing the entire product development process, from inception to completion.

Tech companies can bridge the gap between their promises of fairness and the actual implementation by instituting a measure of bureaucracy, empowering ethics teams, and offering incentives to encourage collaboration with these teams among other employees. This approach ensures that fairness is no longer detached from practical application in day-to-day operations.