Marginal Revolution Saturday Links: Anthropic on AI Herd Behavior, Modi Measured Against Synthetic Controls, and More
Key Takeaways
- •The roundup included a Nature-linked item on speculative Chinese claims about Alzheimer’s disease.
- •Anthropic published research arguing that multi-agent systems can exhibit more herd behavior and collusion than human systems.
- •A working paper on SSRN presents cross-cultural data on masculinity, while another project evaluates Narendra Modi’s performance using synthetic controls.
- •A tracker documents which U.S. federal datasets have been terminated, raising practical concerns for users of government statistics.
- •The New York State Fair is paying $40,000 to keep its iconic $1 potato booth operating with a new vendor.
Marginal Revolution, the economics blog co-authored by Tyler Cowen and Alex Tabarrok, published its recurring Saturday assorted links roundup on August 15, 2026. The post surfaces six items spanning biomedical research, artificial intelligence, social science, Indian economic policy, U.S. federal data, and state-level procurement economics.
1. Speculative Chinese claims about Alzheimer's. The item links to a Nature news article covering speculative claims from Chinese researchers concerning Alzheimer's disease, the neurodegenerative disorder that is the most common cause of dementia worldwide.
2. Anthropic on multi-agent systems. A research piece from Anthropic, the AI company behind the Claude model family, examines why multi-agentic systems appear to show more herd behavior and collusion than human systems. And a comment from Rune Kvist on X: https://x.com/RuneKvist/status/2087916055369003507
3. Cross-cultural data on masculinity. An SSRN working paper presents cross-cultural data on masculinity.
4. The Griers on Modi's performance. The Griers assess the performance of Indian Prime Minister Narendra Modi relative to synthetic controls. The synthetic control method, widely used in economics, builds a counterfactual from a weighted combination of untreated comparison units to estimate outcomes absent the policy or leader under study. That makes this kind of comparison useful for readers trying to separate a leader’s measured record from broader trends in the data.
5. Which federal datasets have been terminated? A terminations tracker documents which U.S. federal datasets have been terminated, highlighting a practical issue for anyone who relies on government statistics for research, reporting, or public analysis.
6. The economics of NY state potato procurement. Coverage from New York Upstate reports that the New York State Fair is paying $40,000 to keep its iconic $1 potato booth in operation with a new vendor, a small example of how preserving a familiar public-facing tradition can require a larger budget commitment than the sticker price suggests.
