rationale:

The highest hurdle for any of these outcomes is the strict resolution criteria: a general-purpose AI must originate the central idea and main result, the field must widely consider the dispute settled, and researchers must credit the AI analysis as the decisive contribution. Even decisive human work rarely produces consensus in the social sciences, where disputes are often entangled with ideological differences, subjective methodological preferences, and data access constraints. Consequently, "None by 2035" is the most likely outcome at 46%. Social science consensus is notoriously slow, and researchers may simply absorb AI as a routine tool rather than explicitly crediting an AI system with a decisive settlement.

While AI capabilities have progressed at a breakneck pace in mathematics, empirical social science presents a different challenge. Recent milestones demonstrate impressive scale but highlight reliability issues: an automated audit of 4,452 economics replication packages flagged thousands of discrepancies and highly sensitive estimates Reproducing published economics (economics) - BootLoops An LLM Workflow That Reproduces, Improves, and ..., and Project APE generated 1,000 papers end-to-end Project APE: Can policy evaluation be automated? Or is .... However, none of the APE papers were peer-reviewed, the system was deemed unreliable with many confirmed error flags https://ape.socialcatalystlab.org/research-notes/evaluating-the-first-1000-ape-papers.pdf Project APE: Can policy evaluation be automated? Or is ..., and the massive NBER audit explicitly does not aim to settle substantive disputes An LLM Workflow That Reproduces, Improves, and .... Technical reproduction and fragility analysis do not establish causal validity, meaning current AI is better at unsettling fragile literature than definitively closing long-running debates.

If an AI does settle a dispute by 2035, it will likely be an unlisted one. "Another" holds a 26% probability because it captures a broad long tail of narrower behavioral, cognitive, educational, or theoretical controversies. AI systems are already running automated theory-revision loops with human participants auto-psych: Automating the science of mind using agent ... Closing the Loop to Discover Psychological Theories with .... A narrow experimental psychology or survey-methods dispute is much more likely to be decisively resolved by massive AI-driven online experiments mapping boundary conditions than any of the macro-level named options.

Among the listed disputes, ego depletion and social priming (8%) is the strongest candidate because discriminating studies can be conducted via scalable online experiments. However, large multi-lab human replications have already dealt severe blows to these theories The Collapse of Ego Depletion - by Michael Inzlicht A Multisite Preregistered Paradigmatic Test of the Ego- ... Evaluating the Replicability of Social Priming Studies, leaving little room for an AI to earn decisive credit. Minimum-wage employment effects (7%) and social media and adolescent mental health (7%) are prominent but heavily constrained by data access, differing interpretations, and political bias. Productivity-slowdown mismeasurement (3%) is largely considered settled by human economists Challenges to Mismeasurement Explanations for the US ..., and immigration's wage effects (3%) remains deeply tied up in contested identification strategies Does Immigration Affect Native Wages? A Meta-Analysis.

The primary risk to the "None by 2035" plurality is the extraordinary trajectory of agentic capabilities. Nobel-caliber AI capabilities will likely emerge in the early 2030s , supporting a substantial chance of a decisive contribution before 2035, particularly on an unlisted dispute, even though forcing widespread consensus on a long-running debate remains a uniquely difficult sociological challenge. With the time horizon for autonomous research agents doubling every few months and systems already solving open mathematical problems, it is entirely plausible that by 2030–2034, AI labs could launch massive pre-registered multi-analyst studies or seamlessly integrate vast administrative datasets. If an AI system can overcome the causal identification challenges that currently bottleneck human research, it could successfully force consensus and claim the necessary credit before the deadline.