Last Spring in Chicago, we hosted an interdisciplinary conference focused on examining the possibilities, limitations, and broader implications of AI technologies for human learning, with particular attention to considerations for racial equity and justice in educational contexts. Amid a flurry of AI events and conferences over the past year, Towards New Horizons of AI, Learning, and Equity in Education contributed to the thinking in this space with an explicit focus on equity and a deliberate approach to fostering interdisciplinary dialogue across experts in technology, policy, philanthropy, and research communities. With an overarching aim of fostering cutting-edge research, the two-day meeting engaged participants from a range of disciplines and methodological traditions, including educational scholars, AI scientists and developers, educators and school and systems leaders, and policy experts. In designing the convening, we benefited tremendously from our collaboration with Dr. Alondra Nelson, prominent scholar of science and tech policy and the former director of the Office of Science and Tech Policy at the White House. We organized the conference around a set of fundamental questions residing at the intersection of learning, equity, and technology policy:
1. What practices, research, and/or policies are already happening at the intersections of AI, learning, and equity, and what can we learn from these efforts?
2. What are some of the key equity and ethics issues that we should be attentive to as AI technologies are developed and used? How do we weigh the equity costs and benefits thoughtfully? What safeguards need to be in place?
3. What new research do we need to expand our understanding of how AI technologies can be tools to advance equity in learning?
4. What transformative policies and practices can we imagine for AI education that could have deep impacts on educational systems for equity?
5. What kinds of new systems, practices, technologies, policies, and outcomes might we imagine if we weren’t bound by current conditions in policy, practice, and our fields?
Coming into the convening, we kept a few things in mind. First, we were keenly aware that this moment is not the first time that new technologies in education have been intended to drive equitable outcomes– and there is a long history of unfulfilled promises. Second, with respect to AI technologies specifically, many harms have already been documented. These include algorithmic discrimination and data and privacy violations, issues that have a disproportionate impact on underserved communities, as well as the energy and environmental costs of computationally demanding AI systems.
While acknowledging these very real threats and limitations, we were also keen to take a serious look at emerging research suggesting that AI not only may have the potential to improve student learning outcomes, but to do so in ways that address longstanding inequities in access to instructional resources and learning outcomes. We don’t just see this as an opportunity, but rather our responsibility to soberly consider the potential positive impact AI may have on marginalized communities and learners.
And finally, it was important for us to recognize that the discourse on AI in education is occurring against a backdrop of (and in some cases in parallel with) other high-stakes conversations. These include the effects of social media on adolescent mental health as well as its influence on global politics, including in the current high-stakes presidential elections. In our weekly design meetings preparing for the conference, Professor Nelson would frequently remind us that there is a unique moment here to “get it right” by learning the hard lessons from recent tech policy decisions in other domains. That charge and responsibility —to get it right when it comes to AI and education—and the role of educational research in this endeavor, grounded our work together over the two days in Chicago.
In a series of memos and moderated panels and discussions, participants identified several key openings and opportunities for AI to move the needle on educational equity. First, and perhaps unsurprisingly, participants highlighted the potential of AI to democratize access to high-quality resources and provide personalized learning experiences tailored to individual needs. A second major theme was the potential application of AI to create culturally relevant and culturally sustaining curriculum and assessments, ensuring that educational materials and evaluations resonate with diverse student backgrounds. Further, there was also a sense of cautious excitement regarding the possibilities of leveraging AI for justice-oriented projects, including decolonization, redistribution, and educational repair for marginalized communities. The use of data-driven analytics for diagnostics and targeted interventions—such as supporting students with disabilities or those experiencing homelessness—was also recognized as a promising area for AI to impact educational equity positively.
Despite the promising opportunities identified, the memos also revealed significant challenges and risks associated with welcoming AI into educational contexts. To start, one prevalent theme across the memos was the significant —and many pointed out—distinct ethical challenges unique to AI technologies. For instance, many of the memos drew attention to the undeniable concentration of power and privilege in the largely unregulated AI industry, expressing that AI might ultimately consolidate existing inequities rather than address them. The question of adverse environmental impact of resource-intensive AI tools also emerged as an urgent issue and concern across several memos, as well as within conversations during the conference. Even while some AI applications promise to model and otherwise address climate change, the sheer compute power necessary to train large language models has led some to project that the AI industry will be one of the largest contributors to carbon emissions.
Other challenges identified were pedagogical questions of defining where AI should fit within the curriculum and identifying core competencies. Questions here included debates around AI literacy, Computer Science education, and AI as a tool for civic engagement. Next, there is a significant and troubling disconnect between current theories of learning and those that underpin most AI technologies. In other words, the majority of currently available AI technologies actually are rooted in outdated, and in some cases invalidated, theories of human cognition and learning (e.g., some versions of personalized learning AI tutors completely sidesteps the last 30 years of research demonstrating the fundamentally cultural and social-interactional nature of learning). And from this perspective, AI tools in their current instantiations are not adequate tools for teaching and learning.
Policy concerns were also prominent, with many expressing the nuanced dilemma that there is a risk of either moving too quickly and failing to implement adequate protections for students or moving too slowly and missing out on the potential benefits AI could offer. And yet, despite these myriad challenges and concerns, the procurement of AI-powered educational technologies continues unabated, often taking place behind closed doors, and crucially, without the benefit of learning experts or community voices and values.
Given these opportunities and challenges, participants were asked, what new research do we need to expand our understanding of how AI technologies can be tools to advance equity in learning? Several key themes emerged from this prompt. Many participants highlighted the need for educational and human development research focused on understanding how AI technologies can enhance or potentially undermine the quality of teaching and learning.
Additionally, there were calls for technical and design research aimed at creating AI solutions that are more attuned to the cultural contexts of teaching and learning in diverse school environments. Policy research was also highlighted, particularly concerning AI use, regulation, adoption, and procurement within schools and districts. Several pointed to the importance of a global lens on questions of AI policy and schools. What might we learn from other nations? What might we offer? What might global cooperation mean for the future of AI and schooling?
So what do you imagine happens when you invite leading voices in academia, policy, tech industry, and the foundation world to spend nearly two days, inperson together to discuss the topic of AI, education, and equity? Yes, there was a healthy dose of tension and disagreement in the room. The diverse range of perspectives and opinions on AI, equity, and education led to real and sometimes challenging discussions. Ultimately, however, we found the tensions to be highly generative and managed skillfully by all participants, buttressed by numerous opportunities to socialize and build relationships throughout the conference. We had to learn how to talk to each other across the particular nomenclatures of our respective domains. This is a subtle but crucial point: the extent to which AI will play a beneficial and equitable role in the future of education depends in large part on the ability and willingness of diverse stakeholders to understand one another, work with one another, and ultimately design interventions in teaching, learning, and policy that are guided by ethics as well as evidence from multiple disciplinary perspectives and methodological traditions.
In this sense, the convening served as a proof of concept, demonstrating the value of synthesizing not just interdisciplinary perspectives, but across the layered and complex AI and education ecosystem that is comprised of individuals and groups from industry, government, schools/districts, non-profits and community organizations, and universities. As we know, the debates on AI are multifaceted and polarized, with some critics focusing on existential threats while others highlight immediate issues like bias, discrimination, and the environmental impact of AI technologies. Within educational research, conversations about AI are fragmented, ranging from outright opposition to enthusiastic support, with varying degrees of cautious optimism in between. The convening underscored the necessity and benefit of uniting individuals and organizations with diverse levels of concern and enthusiasm, all focused on influencing the future of AI in education in ways that advance equity and support marginalized communities.
Our task was (and remains) to simultaneously resist the unexamined grandiosity characteristic amongst AI boosters, while remaining open to the possibilities for learning enabled by the advent of a technology that some experts say could be as transformative as the Internet. This is precisely where academic research plays a vital role. What is clear is that in the coming years AI will alter teaching and learning processes within education systems. But how this change will occur, and who the beneficiaries will ultimately be, has yet to be determined, and should be proactively shaped by government and the public, as well as the research community. From an equity lens, some argue that mere refusal of AI in education, given its associated risks and the tech industry’s influence, is the most prudent and ethical stance. Our position (the one that shaped our initial approach to the convening and was significantly refined based on what we learned from the convening) is that sitting on the sidelines is not an option and will likely inadvertently further entrench the power of tech industries to impose their limited visions of teaching and learning on vulnerable communities.
Instead, now is the moment for education research to lean in and conduct rigorous studies that can inform the way AI will shape the future of education and, indeed, the future of learning. Funding research that addresses both the potential of AI and the associated concerns around equity, power, and justice can empower teachers, students, and communities and curb the undue influence of the tech industry into educational systems and processes. We believe deeply in the urgency of educational research to remain central to the conversation about AI and its role in schooling, and in advocating for a more equitable and ethical future for AI in the life of children, families, communities, and the planet.