Equitable Approaches to AI Are Sorely Needed
Waves of new technologies, especially algorithms that make decisions in notoriously biased ways, have long impacted the lives of the millions of Latinos we represent as UnidosUS, the largest Hispanic civil rights and advocacy organization in the U.S. Across critical areas like housing, access to credit, employment, and health, tech-driven decisions can trap communities of color in longstanding patterns of discrimination, invisibility and exclusion.
By Janet Murguía, UnidosUS President and CEO
Last month, I was one of only a handful of civil rights and civil society leaders invited to speak at the U.S. Senate AI Insight Forum, alongside the largest technology firms in the world. I brought the message that technological advances have, for too long, left out and left behind communities of color, and that equal access to the benefits of technology is a civil rights issue.
We have witnessed incredible changes in technology over the past few decades. Yet too few of these innovations addressed real needs in communities of color, and some increased inequality by driving gains in wealth that are not widely shared. That is why, with AI, our approach must be fundamentally different than it has been in our recent past.
While AI offers potentially world-changing opportunities for Latinos and other communities, it could also merely entrench discrimination and threaten civil liberties. Our success in achieving a brighter future depends not on technology—but on whether we can increase diversity in the workplace and tech sector, and whether our values are harnessed to ensure the transparent, accountable, and equitable use of this powerful new set of tools.
Some 35% of Latino households currently lack consistent access to home internet, resulting in a lack of access to devices, services, and digital readiness. At several leading tech companies, the share of U.S. technical employees—such as coders—who are Black or Latino increased by less than a percentage point between 2014 and 2019. While these companies innovated over those five years in many ways, diversity in the workforce was not one of them. Hispanic workers are grossly underrepresented in STEM jobs—we are 8% of STEM workers but 17% of workers overall. Yet Hispanics were 51% of America’s population growth between 2010 and 2020.
In fact, Latinos will account for a stunning 78% of new U.S. workers between 2020 to 2030, and Latinas are the leading demographic group starting new businesses. But because Latinos over-index in the digital divide, 57% ages 16 to 64 have “low or no” digital skills. This makes our community particularly vulnerable to job displacement from technologies. Without meaningful action, AI could become another driver of income inequality for workers and entrepreneurs.
Furthermore, the data to build and train AI tools—including algorithmic models—come from our deeply inequitable world. Disinformation, scams and predatory schemes—too often neglected by platforms and authorities, especially when in Spanish—proliferate and include attempts at voter suppression. And a lack of basic democratic guardrails about the use of data leads to government overreach, sweeping up both immigrant families and millions of U.S. citizens.
At the same time, AI tools hold breathtaking promise for communicating across language barriers, and for health and education by enabling nuanced research or creating tailored approaches to learning. Tools like chatbots are relatively simple to use and could—if well designed—increase retention for newer workers and cut training times. Substantial funding for creating more Latino and Latina digital economy workers, through partnerships with culturally competent community-based organizations, as well as skills-based hiring that connects workers to an AI economy, will be needed for workers who today are excluded from upwardly mobile, future-focused employment. The dividend for our national economy would be astonishing.
We will know we are innovating in the right way only when our tools achieve and advance transparency, access, safety, and equity, among other values, and when the process is participatory and inclusive. But we cannot address bias in the outcomes, manage the disruptive downsides of AI and its destabilizing potential, or effectively balance law enforcement needs against our valued freedoms, unless we center our vision on solutions in the near term while we create stable guardrails around emerging uses that pose a future risk. It is therefore essential that Congress inject shared values into the design and development of AI and create common sense rules of the road.
Lawmakers should also require that the people who are—or will be—most deeply impacted by transformative technologies, including workers and communities of color, have a powerful seat at the table concerning how, or even whether, specific tools will be built and used. While AI could accelerate advances across nearly every domain of our lives, we must proactively address the factors that could allow fundamental inequality to persist even through such rapid change. An ambition to do that is the true spirit of both democracy and innovation.