Are We Building AI for People, or Just for Buzzwords?
Everyone’s talking about generative AI – the dazzling, sometimes terrifying, ability of machines to create text, images, and code. But while Silicon Valley chases the next viral chatbot, a crucial question gets lost in the noise: are we actually building technology that helps people, or just technology that generates hype? The real story here isn't the latest LLM breakthrough—it's the quiet, vital work being done by researchers like Jaylin Herskovitz, a newly appointed assistant professor at Tufts University, who is demonstrating how deeply understanding human needs can unlock genuinely useful AI applications.
Original reporting: tuftsdaily.com.
Herskovitz’s journey into computer science wasn't a straight line from algorithm design to venture capital pitches. It began in a human-computer interaction lab at the University of Michigan, where she realized that the most compelling aspect of the field wasn't the technical wizardry, but the opportunity to work directly with people. This realization steered her towards accessibility research, specifically focusing on how blind and visually impaired (BVI) individuals utilize AI technology. It’s a perspective that challenges the prevailing Silicon Valley narrative, where innovation often prioritizes novelty over utility.
What Herskovitz discovered is that BVI users aren't casually asking AI for knock-knock jokes. They’re leveraging it to tackle complex, high-stakes tasks. “They are essentially extreme early adopters of this technology and actually using it in much more practical ways than the average person,” she explained. This “hacking, switching, and combining” of different AI assistive programs to navigate daily life offers a powerful lens for understanding the technology’s potential—and its limitations. Herskovitz’s work with ProgramAlly and AllyExtensions, apps that allow users to program AI to identify specific visual information and create custom shortcuts between accessibility tools, are direct results of observing these user behaviors.
The distinction between Herskovitz’s approach and much of mainstream AI development lies in her deliberate integration of social science concepts. Herskovitz, while not formally trained in the social sciences, actively seeks to broaden her interdisciplinary perspective, drawing on fields like anthropology, psychology, and cognitive science. She’s even been gently corrected by other HCI researchers who pointed out she was already using social science theories, just without explicitly recognizing them. This isn’t about adding a “diversity and inclusion” checkbox to a project; it’s about fundamentally reshaping the design process to center the lived experiences of the people who will actually use the technology.
Herskovitz’s emphasis on interdependence within disability communities is particularly insightful. The common narrative in accessibility often frames independence as the ultimate goal – a world where individuals can “do everything on their own.” Herskovitz argues that this overlooks the reality that many people with disabilities thrive in community and value collaborative solutions. “A lot of people with disabilities actually like living in community with other people,” she said, a sentiment that directly challenges the often-isolating ethos of individualistic tech solutions.
Her current work at Tufts, building the AI, Design and Accessibility Lab, is focused on tackling even thornier issues, like AI hallucinations and how to help BVI users reason about errors in AI systems. One student is currently exploring how to help users understand the source of errors when they can’t see the input to an AI system—a critical challenge for ensuring trust and usability. This focus on error handling, rather than simply chasing the next flashy feature, is a refreshing departure from the typical AI development cycle.
The broader implication here is that true innovation isn’t about building the most powerful algorithm; it’s about understanding the human context in which that algorithm will operate. Herskovitz’s work serves as a powerful reminder that the most impactful technology isn’t always the most visible, but the technology that quietly empowers individuals to navigate their lives with greater agency and connection.
Here’s what to watch for: In six months, I predict we’ll see a surge in university-led research projects explicitly incorporating social science methodologies into AI development, spurred by the growing recognition that technical prowess alone isn’t enough. Will venture capitalists follow suit, shifting funding away from purely speculative AI ventures and towards projects with demonstrable social impact? That’s the question that will truly determine whether Herskovitz’s approach becomes a mainstream paradigm shift, or remains a valuable, but isolated, example of responsible innovation.
