At Absolutely Interdisciplinary 2024, speakers from a wide range of fields discussed current challenges and complexities of aligning artificial intelligence systems with human values. Photo: Johnny Guatto.

➦ Watch video recordings of all three days of Absolutely Interdisciplinary 2024 on the SRI YouTube channel.

How can we design artificial intelligence (AI) systems that are aligned with human values? And how are these technologies impacting our world today?

At the Schwartz Reisman Institute for Technology and Society’s (SRI) annual academic conference, Absolutely Interdisciplinary, leading researchers from diverse fields came together to tackle the complexities of AI alignment and how to better understand the social impacts of data-driven technologies. Held at the University of Toronto’s newly-opened Schwartz Reisman Innovation Campus, the conference spanned three days of discussions that fostered innovation and collaboration, as 28 distinguished speakers presented new approaches and ideas to over 200 participants.

The conference kicked off May 6th with "Interdisciplinary Dialogues on AI," a workshop convened by SRI’s graduate fellows, where panelists engaged in conversations on the ethical dimensions of AI applications across medicine, geopolitics, social media, and smart cities. The workshop’s sessions explored how real-world use cases can benefit from interdisciplinary approaches and collaboration, and how AI can offer improved services while generating unique challenges that require new frameworks to understand.

In a session on AI in healthcare, Dr. Mamatha Bhat of UHN explored how AI can increase equity of access for transplant procedures, while CAMH’s Daniel Buchman discussed the potential for AI to perpetuate “epistemic injustice” in mental health contexts, despite the best intentions of practitioners. Muhammad Mamdani, director of the Temerty Faculty of Medicine Centre for Artificial Intelligence Education and Research in Medicine (T-CAIREM) at U of T, succinctly captured this tension by observing his team has successfully deployed training algorithms across Unity Health Toronto to reduce unplanned patient mortality by 26 per cent, but questions regarding bias and explainability within these systems still remain unsolved. To make AI work, Mamdani observed, “You have to have courage to do something different that nobody else is doing, and oftentimes that’s the hardest stumbling block.”

Day one of the main conference, held on May 7th, delved into the importance of aligning AI systems with human values in a world in which artificial agents are integrated into society. Inaugural Schwartz Reisman Chair in Technology and Society Gillian Hadfield, who is also a professor of law and strategic management at U of T, engaged in a thought-provoking conversation with philosopher Peter Railton of the University of Michigan on developing AI agents that possess moral or normative competence, while Roger Grosse of the Department of Computer Science addressed the pressing need for assessing AI safety risks, introducing novel approaches to understand different levels of risk, and algorithmic advances for determining if a model is unsafe. In the afternoon session, Harvard Law School’s I. Glenn Cohen and Berkman Klein Centre for Internet & Society’s Huili Chen delved into intricacies of human-AI interaction across the domains of medical applications and educational tools, exploring principles from law, bioethics, engineering, and psychology.

The conference continued May 8th with explorations of AI’s impacts across society, including challenges of adoption and democratic values. The morning featured a discussion moderated by SRI Interim Director Kelly Lyons between Ray Perrault, co-chair of Stanford HAI’s AI Index Report, and SRI Associate Director Peter Loewen, director of the Munk School of Global Affairs & Public Policy. Presenting the 2024 AI Index, Perrault shared a wealth of data on technical advancements, public perceptions, and geopolitical dynamics surrounding AI development, noting that while many AI models are getting closer to human performance in many areas, adoption has lagged. Following this, Loewen unveiled findings from SRI’s Global Public Opinion on AI Report, a survey on public perceptions of and attitudes toward AI conducted across 21 countries in 2023. The report’s findings demonstrate that people in the Global South are more optimistic about AI, including what it means for their jobs and belief that it will make the future better. Loewen proposed this vision to be more pragmatic than predominant views in countries like the U.S., the U.K., and Canada, which are mostly pessimistic about AI’s potential. “One of these worlds is optimistic, realistic, and broadminded; the other is not,” Loewen ventured. “There’s a real question about which world we are in and which world we want to be in.”

In the next session, SRI Research Lead Beth Coleman engaged in a discussion with Harper Reed, CTO of the Obama 2012 campaign, on how algorithms have transformed political campaigns and the unintended consequences of leveraging social demographic data. Speaking about the future, Reed cautioned that we must carefully consider the economic and social impacts of automation: “This is a very large bell that we can’t un-ring… Do we choose a path towards Star Trek, or do we choose a path towards Mad Max?”

The conference concluded with a session led by SRI Research Lead and Rotman Chair in AI in Healthcare Avi Goldfarb on the challenges of integrating AI into businesses. As SRI Faculty Affiliate Kristina McElheran of UTSC’s Department of Management demonstrated through comprehensive data collected across the U.S., the reality of AI adoption is more limited than recent media hype suggests—only about 6 per cent of U.S. firms successfully use AI, and this statistic has been mostly flat over the past eight years.

To complement these findings, Christy Prada of Future Fertility unpacked challenges in building her company, which uses AI-driven insights to improve IVF treatments through analyzing oocyte and embryo images. Prada described her journey in developing scientific credibility and building new standards of care, workflows, and educational initiatives. The session concluded with an insightful discussion between participants on what paradigm shifts are needed to unlock AI’s transformative economic potential and what barriers continue to stand in the way.

The conference concluded with the 2024 Ian P. Sharp Lecture, presented by SRI in collaboration with U of T’s Faculty of Information, which was delivered by Beth Simone Noveck, director of the Burnes Center for Social Change at Northeastern University. Noveck explored the potential for AI to support participatory democracy and expand government capacities to engage citizens, citing examples from California, Brazil, and Italy in which AI tools have been used to summarize information into clearer and more accessible forms, improve public services by responding to inquiries with customized responses, and surface threats like wildfires more rapidly. Noveck noted these technologies offer vast new resources to public officials, and that because generative AI tools are driven by language, they represent more accessible forms of technology to develop and deploy. In the Q&A session that followed, some participants expressed uncertainties around the risks and limitations associated with AI systems, including potentials for bias and the implications of private ownership, but many felt Noveck’s optimistic message to be refreshingly resonant given the challenges of the present moment—including an environment influenced by climate change and political disagreements, and the urgent need to rebuild public infrastructure.

“If we are to reboot democracy, if we are to avert the crisis, we need to start thinking about how to use AI to upgrade our institutions,” contended Noveck. “Democracy hangs in the balance.”

As society moves rapidly towards an increasingly AI-driven world, it is clear stakeholders from across society must come together to help build technologies and systems that can unlock the benefits of AI in responsible and safe ways. As the conference made clear across all its sessions, this challenge requires developing new ways of thinking, and fostering new perspectives and collaborations. To ensure a future that reflects our collective values, we must be ready to build new bridges and imagine new solutions that can get us there.