UNIVERSITY PARK, Pa. — Artificial intelligence (AI) systems affect many parts of daily life, including healthcare, education and public policy, but the public has had few meaningful opportunities to participate in the development, governance or modifications of AI systems, according to Tanusree Sharma, assistant professor in the Penn State College of Information Sciences and Technology. As a result, AI systems may not align with the needs of diverse communities.
To help address this, Sharma — and collaborators from OpenAI, the University of Illinois at Urbana-Champaign, the University of California, Berkeley, and Stanford University — proposed InclusiveAI, a system inspired by democratic decision-making that uses decentralized voting and discussion tools to give people with different perspectives a voice in how AI is used. The work, funded by a grant from the OpenAI Democratic Inputs to AI Initiative, was recently published in Scientific Reports.
“Current efforts to align AI with human values or governing AI models are largely top-down, controlled by big tech corporations and institutions,” Sharma said. “Design and improvement decisions may be based on surveying or collecting opinions from users, but these methods don’t allow people to discuss ideas together, learn from one another or reach shared decisions — things that are needed to make AI tools fairer and more inclusive.”
According to Sharma, people from different cultures and backgrounds may have different views about what counts as fair and accurate in AI. But users are often presented with AI outputs built on system specifications that do not represent diverse societal values and perspectives.
“Imagine asking AI to generate an image of a CEO or a nurse, and the system produces gender-biased images that reflect embedded stereotypes,” Sharma said. “What if there were an option to challenge AI’s response and make suggestions to improve it?”
Working with 177 participants from around the world, the researchers tested different ways of sharing decision-making power, seeking a fairer, democratic method to give more people an equal opportunity to shape how AI behaves.
In place of traditional survey methods, they proposed using a Decentralized Autonomous Organization (DAO)-based transparent governance/decision platform to collect people’s preferences and AI behavior to guide AI development and improvements. DAOs are blockchain-based groups that use smart contracts to support decentralized decision-making and new ways of governing AI collectively.
“Rather than forcing everyone to agree, the DAO process encourages discussion, compromise and understanding of different perspectives,” Sharma said. “This kind of community-driven decision-making can help create AI systems that are more inclusive, balanced and representative of society.”
In practice, a DAO system would allow users to enter a deliberation space — a forum where they could discuss alternative viewpoints with other users and developers and use their participation power to vote on proposed changes, ensuring transparency and immutability. The approach could also be used in areas such as medicine, law enforcement and community safety by involving experts, policymakers and regulators in design and improvement discussions.
“The aim is to balance the tension between majority rule and minority protection,” Sharma said. “A decentralized participatory — digital democracy — system would help reflect broad public opinion but still protect minority viewpoints on sensitive issues.”
Penn State is shaping the future of higher education in the age of artificial intelligence. Our focus is on human-centered, ethical AI innovation that delivers meaningful impacts for Penn State and the broader community. Through visionary planning, strategic partnerships, targeted hiring and strategic investments, we will equip every Penn State student, staff and faculty member with the AI-related knowledge, experience and confidence they need to succeed in the AI-powered future. Learn more at Penn State's AI website.