Velocity by Booz Allen

EQUITY

4 PILLARS FOR FOSTERING A DIVERSE, EQUITABLE, AND INCLUSIVE AI-DRIVEN FUTURE John Larson and Ramon Hill AI for Everyone A rtificial intelligence will have profound implications for every aspect of society, from healthcare and transportation to finance and national security. Already academia, policymakers, technologists, pundits, enterprises, economists, and average citizens are contemplating and debating the many ways AI could or should be used. To be sure, it’s essential to do all we can to understand and address the real-world impact of AI deployment—beyond recommending your next streaming video or predicting your shopping preferences. AI is already assisting doctors in making better diagnoses, streamlining access to critical services, helping remediate climate challenges, and much more as described elsewhere in this issue of Velocity . However, a critical question of our time is how to harness AI’s power for society’s broader good, to make lives better. There is necessary public discussion about AI’s capacity to exclude marginalized populations—but there is an equally resounding opportunity to use AI in fostering diversity, equity, and inclusion (DEI). Here are four spotlight areas where AI and DEI intersect to help promote the equitable treatment and full participation of people from across society.

“In a world where AI is rapidly reshaping every facet of our lives, it’s imperative that its development and application are rooted in diversity, equity, and inclusion. We believe that the future of responsible AI lies in equipping every student, regardless of their background, with the knowledge and tools to navigate, influence, and innovate in this transformative landscape. Cultivating diverse talent isn’t just about filling seats—it’s about ensuring that the AI systems of tomorrow are built with the collective wisdom of our entire society , reflecting the richness of our shared experiences and values.”

Alex Kotran Co-Founder & CEO, aiEDU

4 Understand and Mitigate Technical Bias AI algorithms directly reflect the data they are informed with and their parameters for learning. When models are fed biased data, such as erroneous, unrepresentative, or discriminatory information, those models can amplify or propagate the bias. With AI increasingly driving decisions that impact individuals’ lives, algorithmic bias against subgroups of our society can reinforce and perpetuate discrimination and imbalanced power structures. The mortgage lending industry exemplifies how AI bias can impact lives and livelihoods. In the past, minorities were not able to get mortgages at the same rate as nonminorities due to discriminatory lending practices. If that historical data is used to train an AI model, that model would likewise approve mortgages at a higher rate for nonminority applicants than for minority applicants—perpetuating and amplifying this historic inequity. Although it’s difficult to know if a data set is biased until certain groups are affected, AI leaders must train talent to understand those types of systemic biases in data to minimize them early and often throughout the modeling process and guard against biased outcomes. For example, employing quality control strategies helps find biases during the development cycle, response bias sampling corrects issues in the dataset by oversampling specific populations, and generating synthetic data helps represent diverse populations when actual data doesn’t exist.

communities could spark an interest in AI as a career and expand the pool of future leaders and practitioners. As a community, we must be intentional about bringing these technologies into educational settings and equitably unlocking student potential. For example, the AI Education Project (aiEDU) is a nonprofit organization that seeks to do just that, by developing AI literacy and training content that educators, schools, and school districts can use in classrooms. 3 Diversify AI Development Teams AI leaders are responsible for ensuring that those who work in AI reflect the rich tapestry of our society. It’s no secret that teams of individuals with diverse perspectives and demographic backgrounds bring forward richer, more innovative solutions and better business outcomes. We believe that the same is true with AI. Consider the development of facial recognition algorithms around a decade ago. At that time, those working on the problem didn’t consider that if the model was trained on historical data, it would perform best for lighter-skinned males and poorest for darker-skinned females—which is exactly what happened. However, when AI models and use cases are designed from the outset by a diverse talent base that brings a broad range of inputs and perspectives to the table, they are much more likely to be built for the benefit of all people—not just a subset of society.

Principles of DEI in the Age of AI

1 Foster AI Literacy for All AI is arguably the single most transformative technology of a generation since the invention of electricity. Just as the Rural Electrification Act of 1936 facilitated and accelerated the provisioning of electricity and, ultimately, its economic benefits to all corners of the nation, we need to ensure that everyone in the U.S. has high-level AI literacy and thus can be part of the emerging economy. AI is already touching virtually everyone daily as they search for jobs, apply for loans, access healthcare, vote, travel, select clothes, and more. Having a basic understanding of what AI is and the skills to use AI-driven tools and technologies is becoming just as essential as knowing how to read, write, and use a computer. Fostering AI literacy also involves educating people about the ethical implications of AI, including those related to bias, privacy, and transparency. As a nation, if we don’t succeed in broad AI literacy that reaches people in underserved and underrepresented communities—or those in jobs that are traditional and nontechnical in nature—we risk increasing disparity and leaving people behind, unable to enjoy the economic benefits that AI

will open for the country. That, in turn, would create incredible individual and community hardship, put the nation at a competitive disadvantage, and inhibit the creation of AI that represents our democratic values. 2 Empower the Future Leaders of AI Our ability to build a diverse AI workforce and to harness the power of AI for societal good is predicated on developing the technical talent of the next generation without leaving anyone behind. Concerningly, much of the focus on students now is on how to prevent them from using AI tools to cheat on homework assignments. position. It’s clear that a broader talent pool is vital: On the graduate level, the number of American students in AI programs hasn’t increased since 1990. While graduate programs are just one of many pathways to AI leadership, we know that more upcoming talent needs to be trained in AI. Providing early access to technical AI education for young people from diverse or historically underserved The next generation of talent is critical to national security and to the U.S. maintaining a leadership

Optimism for the Future AI is going to be transformative for our society, for our country, and for the world. It will help us solve the most challenging problems we face—but only if we ensure that everyone has an equitable opportunity to be part of the AI revolution. This is just a snapshot of an important topic. Tune in to Booz Allen’s Unstoppable Together podcast to learn more about the intersection of AI and DEI, and explore other DEI topics and challenges facing today’s workforce.