Doing the right thing often brings business benefits further down the track, says Xero co-founder and Executive Director Rod Drury.
For example, the software company’s efforts to build a diverse engineering team over the last decade have put it in a strong technical position to develop artificial intelligence that avoids the pitfalls of unconscious bias, Drury said.
“As we get into AI and machine learning, we are starting to see a lot of issues come up around unconscious bias,” Drury said during a panel at the company’s Xerocon event in Brisbane last week.
“So the work we’ve done on diversity in our business over the last five to ten years now really pays off. With that female engineering talent and other [types of] diversity inside our engineering teams, we think they are pretty well set up that we are doing AI and machine learning in the right way.”
In the last 12 months, 32 per cent of Xero’s technical hires have been women, and women make up 27 per cent of Xero’s technical workforce. Anna Curzon, Xero Chief Product Officer, says those numbers are far above the industry average.
“It’s a constant focus for us and we work really hard to make sure that we are getting wonderful female developers but also developers from different countries,” Curzon told Which-50.
“One of the great things about Xero is a lot of our development teams are in Australia and New Zealand so we get a lot of immigrants coming to Australia and New Zealand with these amazing skills and wanting to come and work for us. So we literally have, almost like a United Nations of developers — which is really important to us.”
As well as creating diverse teams to work on technology, Xero also follows principles that govern the use of data across the organisation. That includes anonymisation and aggregation of data, as well as ensuring not just the accuracy, but the fairness of the models they use.
Ahead of legislation
Also speaking on the panel, Dr Catriona Wallace, Founder and Executive Director of Flamingo Ai, said companies like Xero will be ahead of the game when legislation is passed that forces businesses to explain how their AI makes its decisions.
Wallace is working with the Human Rights Commissioner to develop Australia’s first AI ethics framework [pdf], which is in front of cabinet at the moment. The framework focuses on transparency, accountability and contestability, where technology makers need to explain to end users how a decision was made, and users in turn can challenge the machine’s decision.
“That’s all coming. There’s nothing legislated yet, but it’s coming,” Wallace said.
Explaining the connection between unconscious bias and technology teams, Wallace said 90 per cent of the coders who are coding artificial intelligence or emerging tech are male.
“We’ve got less than ten per cent female engineers globally who are working on this emerging tech, which is a real problem because it does build conscious or unconscious bias into the tech,” Wallace said.
She cited the example of Amazon’s recruiting tool, which used a model trained mostly on men’s resumes and penalised any application that contained the word “women’s” — such as women’s sports teams or colleges.
“We are seeing huge problems already, and to a degree the train has already left the station,” Wallace said.
“It will be a necessary thing to have diverse teams — not just women and men, but minority groups — to be building this tech to make sure the bias is not built-in.”