Summary: Artificial intelligence (AI) heralds societal changes that could rival those associated with past transformational general-purpose technologies, such as metallurgy, the steam engine, electricity, and the internet. As with such technologies, AI offers the opportunity for tremendous increases in human well-being while also threatening to destabilize social, governance, economic, and critical infrastructure systems and disempower many humans.
These conference proceedings describe a workshop that explored a novel combination of two methods—backcasting and the use of history—as one means of addressing the challenges associated with advancements in AI. The workshop, held in Washington, D.C., in June 2024, was the result of a collaborative effort among the Long Run Institute, RAND Social and Economic Well-Being, and the RAND Frederick S. Pardee Center for Longer Range Global Policy and the Future Human Condition. The workshop aimed to serve as a pilot of a methodology that we hope can be employed more extensively with policymakers and other decisionmakers.
Backcasting, an explicitly normative futures methodology aimed at exploring hopeful outcomes, invites participants to place themselves in a desired future and describe pathways that could lead to that destination. Backcasting is valuable because it can help participants overcome the constraints that make it difficult to imagine a very different future and to consider near-term actions that might bring that future into being. While people will disagree on the ideal AI-enabled utopia, envisioning such utopias can prove useful if for no other reason than that the process can help identify near-term policies that are compatible with these many envisioned futures.
While backcasting is a powerful method, the societal transformations that AI makes possible could be so substantial that one’s imagination might have trouble encompassing them. Thus, this workshop aimed to improve on backcasting by adding an explicit and deep historical perspective. History can help inject realism into imagination by subjecting speculation about the future to lessons from the past. History can also help liberate the imagination. While we can only speculate about the future, history confronts us with the reality that, in many cases, the present is very different than the past, which makes more concrete any suggestion that the future may be similarly different than the present.
Workshop participants—12 individuals from diverse backgrounds, including business leaders, policymakers, and technologists—were presented with two scenarios, each depicting a different future of AI-enabled human flourishing in the year 2045. The first, called Rising Choir, was adapted from a scenario developed by the group Existential Hope. Rising Choir envisions a highly decentralized society in which a sophisticated AI tool called Voice for Open-Source Information and Community Engagement (V.O.I.C.E) empowers a prosperous economy with an emphasis on AI-enabled local production and direct democracy across all levels of society, from local communities to the international community. To balance this scenario, the RAND team developed a second, less decentralized scenario called The Singularis, in which AI-augmented humans and unaugmented humans live together in harmony and mutual benefit.
Workshop participants were also presented with three historical case studies focusing on the societal impacts of general-purpose technologies in the 19th and early 20th centuries. The first case study, Brave New Worlds, examined the impact of labor-displacing technologies. The second case study, Boom and Bust, examined the impact of such general-purpose technologies as electricity on productivity, innovation, and economic growth. The third case study, Monopoly’s Moment and Markets of the Mind, examined how monopolies and trusts in the late 19th and early 20th centuries exerted a profound influence on markets, industries, and government policy.
Workshop participants discussed several potential areas for policy action, including developing more-sophisticated and more-nuanced AI governance models that strike a balance between encouraging innovation and ensuring the public good. Workshop discussions considered the potential for monopolistic tendencies in the AI sector, the need to enhance the government’s ability to implement a more flexible regulatory system with more-effective collaboration among government and industry, and the possibility of pursuing multilateral international cooperation to avoid the fragmentation of regulation and markets thereby supporting a system in which technological advancements benefit a broader variety of stakeholders.
Our experience with this workshop suggests that engaging history during futures exercises can expand the imagination of those engaged in backcasting and ground their discussion in historical realism. The workshop also offers some lessons for future efforts. These include beginning with scenarios generated by participants instead of developing scenarios among workshop organizers, employing more-detailed and data-supported scenarios, engaging a wider variety of stakeholders, and formally designing the workshops to evaluate the benefits of including history in backcasting exercises. While this workshop provides only an initial exploration of the potential of historically informed visioning and backcasting, the exercise does suggest that the combination of these two approaches could prove important in addressing not only AI governance but a variety of areas in which technology might substantially transform society, including genetic engineering, bioengineering, nanotechnology, robotics, and climate change.
Cite as: Robert J. Lempert, Jonathan W. Welburn, Laurence B. Mussio, Michael Aldous, ‘Applying History to Inform Anticipatory AI Governance: Using Foresight and Hindsight to Inform Policymaking’
and 
by