R/GA Tech Leaders Discuss How Diversity in Tech Can Help Eliminate Bias in AI
Photo Credit: R/GA

R/GA Tech Leaders Discuss How Diversity in Tech Can Help Eliminate Bias in AI

An in-depth interview with three leaders within the company on steps to breaking down bias in tech.

Artificial Intelligence plays an increasingly significant role in our daily lives and is sure to make an even bigger impact in the future. Yet, the reputation for diversity and inclusion in the field is nothing short of abysmal. In April of this year, AI Now Institute released a damning report on the state of diversity in AI. Among its findings were that:

Only 18 percent of authors at leading AI conferences are women.
More than 80 percent of AI professors are men.
There is no public data on transgender workers or other gender minorities.

AI has even shown signs of discrimination against persons of color and women when it comes to algorithms, further creating a bias and disparity in the field.

R/GA is a leader in the field of marketing and advertising. With an impressive roster of clients like The Cosmopolitan of Las Vegas, Nike and Samsung, the company uses emerging technologies like AI to connect clients with consumers in innovative ways. It’s great to see leaders within cutting-edge technology companies come together to speak about the diversity practices that work and those that don’t in hopes of having the field reflect the world in which we live.

R/GA

Carl Desir, global executive director of diversity and inclusion at R/GA, sat down with Lamar Hines and John Tubert, both vice presidents of technology at R/GA, to talk about how the diversity crisis in artificial intelligence is affecting Black people both personally and professionally.

Carl Desir: AI dictates much of our lives from existing tech like voice assistance to emerging tech like self-driving cars. And the lack of diversity is so pervasive in research, tech companies, sample audiences and even in professions like math teachers and professors. How do we ensure that tech of tomorrow represents the increasing majority-minority population?

Lamar Hines: One critical piece is continuing to support movements like AfroTech, which promotes Black professionals in technology. The other crucial components relate to mentoring, sponsorship, and institutional change. We must first support youth coming into this field by creating awareness of the opportunities. After this, we must make significant changes as an industry and within individual companies to ensure diverse talent receives the necessary support to facilitate professional growth. 

John Tubert: I agree with Lamar, mentoring is super important. Start early. There are a lot of high schools that offer mentoring programs, like the Academy of Software Engineering in New York, with which R/GA has collaborated. The other thing is awareness. Companies still don’t fully understand the impact of not hiring diverse teams. I think a good example of that is self-driving cars, right? Research found that the cameras on self-driving cars didn’t recognize people with darker skin. Imagine the impact of having self-driving cars that cannot recognize people. This wouldn’t have happened if minorities and people with darker skin were engineers and testers.

Carl Desir: As an extension of that, once companies have more diverse teams what do you think can be done to retain them? What has to happen for the workplace to become more inclusive and equitable for everybody?

R/GA

Lamar Hines: In terms of improving the work culture, I think the first thing is to have both top-down and bottom-up Diversity and Inclusion strategies. To drive inclusion, you need to have strong support at the executive level to ensure you can implement organizational change. Part of that is executive sponsorship with business resource groups, such as BEN, the Black Employee Network, at R/GA. 

The first step is executive engagement. There needs to be a certain level of investment, effort, and structure to support the professional development of minority talent. Organizations and leaders have to look at policies and create systems to ensure that we can get out of this cycle of failing our diverse talent. Unfortunately, there are many instances where diverse talent is marginalized implicitly and/or explicitly in this industry and more broadly in this country. This executive effort is the foundation that allows business resource groups to flourish. When these things come together, we can take a step towards inclusion and equity in the workplace.

John Tubert: I think the other big part of it is, and this is especially true for big companies, working on diversity at a macro level instead of looking at individual teams. I think it’s important to start small, concentrate on particular teams and start growing diversity from there. That will influence other teams and continue to grow culture within their own team. Over time, put a lot of emphasis on working very closely with the recruiting team on hiring more diverse candidates, more women, and more women of color.

Carl Desir: What are the important stages of AI development in this case? Where do you see everything we’ve been talking about diversity, inclusion, and equity, overlaying with AI development to prevent and eliminate bias, before it gets to the finished product that cannot detect darker skin tones?

Lamar Hines: It starts before the technology becomes a part of the process. We are moving into a new world where Artificial Intelligence and Machine Learning are going to play a significant role in society. We must look at refactoring our education system to continue the path towards equity. Also, we must start to explore how ethics and equity are priorities as we produce new technologies. 

Ultimately, the goal is having diverse perspectives involved every step of the way—from preparing data and training the model, to product development and validation—because AI will change the world beyond modern conveniences like voice assistants or self-driving vacuums.

AI has a serious intersection with bias—it can be an incredible tool that helps to address systemic bias, or it can be a tool that is used to scale and accelerate bias. One example of this can be seen in the criminal justice system in the U.S. There are risk assessments that are being integrated into the court system to predict recidivism. These algorithms make decisions about freedom or incarceration, this impacts decisions from pre-trial, to sentencing. 

In many of these cases, these algorithms are classified as intellectual property and developed by private companies with limited transparency or accountability. This is where the intersection between A.I. and bias becomes real, and can create a world where something like mass incarceration and the disproportionate population of Black and Latinx in criminal justice can be further perpetuated.

John Tubert: I think that there is always bias in data. So you need to specifically look for it. Thankfully, there are more and more companies that are becoming aware of this issue. There are consulting companies that are specifically auditing algorithms, and big tech companies like IBM, Microsoft and Facebook are creating tools to detect bias in their algorithms.

Lamar Hines: I completely agree. The new corporate responsibility in this space is really going to be centered around transparency and companies embracing independent audits from leaders that are coming in the space of ethics and Artificial Intelligence. This will also be critical for activism; a good example of this is Data for Black Lives.

Carl Desir: That’s great. What are you doing to ensure a more diverse, equitable, and inclusive environment and then one piece of advice you would give to people that they can do?

R/GA

John Tubert: Sure. One thing that I’m doing, I touched on it a little bit before, is hiring more diverse candidates, I put a lot of work into that. And creating inclusive spaces so when they join, they feel more comfortable. They are surrounded by people that share the same values. I think it’s super important. I think what people can do is look for diversity companies and boot camps. For example, we’re working with one here, Coding Dojo, which is a boot camp that specializes in working with diverse populations. That’s the first place we look when we are hiring somebody.

Lamar Hines: I think it really comes down to us, as leaders, to ensure that we devote time,  money and resources to promote inclusion in the industry. I’ve worked with groups like All Star Code to help young Black and Latinx students get into computer science. We’ve done a lot of work connecting with and supporting historically black colleges as well. 

Non-diverse candidates are more likely to have connections through formal and informal mentorship and sponsorship. We need to have structures in place to ensure that those same activities and the same level of support happen for our diverse talent coming into and actively working in the field. We have a long way to go, but with dedication and focus, we can drive change.  

Tying it all together

Discussions about diversity in AI tend to be hyper-focused on women. However, as AI Now Institute concluded, the diversity crisis is more than just about gender, it’s about how AI companies work, what products get built, who they are designed to serve, and who benefits from their development. The future of AI in the tech industry can be setup for more measurable results once the the workforce starts to reflect the problems that need solving. We can create all-inclusive results by providing opportunities for those who are generally excluded.

This piece is brought to you in partnership with R/GA and has been revised since its original publishing date of October 16, 2019.