Why women must have a voice at the AI table
Equality by design is the right approach, for women and other marginalized groups.
Katya Moskalenko, product marketing manager at London-based Measure Protocol, is sounding the alarm that artificial intelligence might have a negative impact on the existing gender gap in the tech industry. Although it creates a world of exciting possibilities, it also creates — almost inherently — risks of bias and exclusion.
Avoiding those traps, she told us, will require “a whole chain of decisions and strategies such as ensuring diverse datasets, ensuring diverse teams, ensuring ethical considerations in any AI-powered solutions.”
There’s a long way to go. Moskalenko quotes World Economic Forum data from 2021 suggesting that only 26% of data and AI positions are held by women; there’s little reason to suppose that that statistic has dramatically improved.
It’s going to be a long journey
Katya admits there’s no magic fix for this. “We can’t solve everything at once,” she said. “It’s going to be a long path in order to ensure that AI is serving all individuals. By having a more balanced representation, we can ensure that all new technologies, including AI, are inclusive and unbiased by design.”
But the time to start is now. It sometimes seems like every company, big and small, has a team developing new AI-powered solutions — from Adobe and AirBnB to small start-ups, she said.
“So the first thing is, let’s diversify teams with more women, with more people of color, with more minority groups to ensure a diverse and inclusive perspective,” she continued. “Also, let’s be sure that we have datasets that are inclusive and diverse to be fed into systems, to be fed into large language models.”
The concern that AI models, especially those trained on the web, will develop built-in biases is far from new. And there’s no easy solution, although Moskalenko acknowledges that some of the big players in the space are making an effort to address the issue.
“I’m quite impressed that OpenAI has good guidelines; instructions on how to be cautious and how to be responsible,” she said. “Also, it’s incredibly important to share best practices, and also share some of the weaknesses and vulnerabilities we find. Mitigating unintentional harm is important.”
She asks for thoughtful cooperation; we need to acknowledge that problems exist and be working together to solve them. “With all the economic and business rivalry happening, we should also think of society as a whole, humanity as a whole. We will be living in the society and with the humanity we have built. Effective cooperation is the key.”
Getting those datasets right
Data evangelist and MarTech contributor Theresa Kushner is likewise heavily invested in diversity and inclusion when it comes to women in AI. We asked her about the relationship between diversity in teams and diverse, unbiased datasets.
“Ensuring gender equality starts long before you get to the phase of designing AI algorithms,” she emphasized. “It starts with what data you collect. Interestingly enough most companies don’t always believe that capturing gender is an imperative. Therefore, they really can’t tell you if their data is biased or not. I’ve worked with some companies who have had to infer gender from information they have collected such as name, college affiliation, extracurricular activities. This is not a good way to secure gender, but is often all they have.”
She agrees that it’s necessary to have a diverse team evaluating the AI algorithms. “As data is viewed more and more as a product, AI teams have to start thinking like product developers considering their users and their markets. I once heard an engineer talk about the team that created a Fitbit-like product. Here’s a tool that should track all your bodily functions, but the designers, who were all men, left out of the design the one thing that every woman tracks — her period. Without diverse groups creating data products, we have similar situations.”
Dig deeper: Why we care about AI in marketing
Hire, but also retain
In taking steps towards diversifying the teams working in this space, hiring for diversity is not enough. “There should be efforts that promote not just hiring but also retention — because unfortunately it’s not the end of the mission to hire a woman, it’s also important to ensure that she is upskilled and reskilled and that she has everything she needs to thrive.”
If one thing is clear, it’s that an emphasis on prioritizing the reskilling of male workers in generative AI — and in AI and data generally — will only make a poor situation worse.
Kushner agreed and expanded the point. “Of course, you should train on AI equally, but it’s not the tools that we should be worried about. Access to tools should be easy for any woman in the IT field to get. We should, however, ensure that women are also involved in the governance of the tools and the AI-created algorithms. Remember, that diversity is not just gender and ethnicity, it’s also diversity of thought and approach. Including women often gives you that kind of diversity as well.”
Women in Tech: A global movement
Moskalenko has been involved is these issues, not just concerning AI, through her participation in the Women in Tech movement. She explained why.
“I’ve been through an interesting path from very traditional marketing fields, from very traditional media companies, to the fast-paced reality of technology start-ups,” she told us (Measure Protocol offers software to track competitive intelligence and consumer behavior). “I felt it was important to me to share my experience and to help other women to explore this incredibly interesting and vibrant industry — and make this biased bubble a little bit more diverse and inclusive.”
More on Women in Tech’s worldwide mission can be found here.
Get MarTech! Daily. Free. In your inbox.