## The Human Edge ## Source details **Canonical URL:** [The Human Edge](https://www.imf.org/en/publications/fandd/issues/2025/12/the-human-edge-pablo-pena) ## Other formats - [Markdown version](/en/publications/fandd/issues/2025/12/the-human-edge-pablo-pena/index.md) - [Structured JSON version](/en/publications/fandd/issues/2025/12/the-human-edge-pablo-pena/index.json) - [Bundle manifest](/en/publications/fandd/issues/2025/12/the-human-edge-pablo-pena/bundle-manifest.json) ## Bibliographic details - Authors: PABLO PENA - Published: November 17, 2025 --- ### Overview - Publication: F&D Magazine; PABLO PENA; December 2025. - Core thesis: Three components of human capital—curiosity, critical thinking, and self-regulation—are crucial complements to artificial intelligence (AI) and help determine whether AI augments or substitutes human capabilities. - Key implication: Continued investment in and cultivation of these human-capital attributes is necessary to push the knowledge frontier, preserve innovation, and realize the benefits of AI. ### Curiosity - Arguments and examples: - Historical thought: Alfred Marshall and Michel Foucault emphasized the centrality of human capital to rising living standards. - Thought experiment: Feeding all human-recorded data up to the year 1939 into a large language model (LLM) would not allow the LLM to produce The Beatles’ “Yesterday” because: - There would be insufficient information about future creators’ life experiences. - A vague prompt (adjectives like melancholic, timeless, elegant, lyrical, intimate) would not narrow possibilities sufficiently. - Extension beyond art: LLMs summarize existing studies (e.g., on reducing gun violence in Chicago) but do not empirically design, fund, execute, and evaluate new interventions—human curiosity drives that new research. - Peak data concept: We may reach a point when all available information has been incorporated into LLMs (“peak data”); without humans generating new information, LLM outputs will stagnate. - Financial-market analogy: Parallels Sanford Grossman and Joseph Stiglitz’s refinement of the efficient market hypothesis—if no one gathers information, prices/information cannot reflect new knowledge; similarly, AI requires humans to produce new knowledge for continued improvement. - Finding: - Curiosity and AI function as complements; AI improves only if humans keep developing new ideas and pushing the knowledge frontier. ### Critical thinking - Distinction between domains: - In hard sciences, LLMs can give definitive answers that do not require lay judgment. - In social sciences and humanities, LLM outputs remix historical arguments and rarely yield definitive answers; human judgment is necessary. - Examples of socially meaningful questions where judgment matters: “How do I know someone is in love with me? Is there a God? Should I have children? Who should I vote for in the presidential election?” - Risks: - Campbell’s law: “The more any quantitative social indicator is used for decision-making, the more subject it will be to corruption pressures.” - Data poisoning: Widespread reliance on LLMs creates incentives for bad actors to contaminate training data with disinformation, making vigilance and critical thinking essential. - Finding: - Critical thinking is essential to interpret, vet, and apply AI-generated information, especially in social-science contexts vulnerable to manipulation. ### Self-regulation - Argument: - AI can produce personalized plans and recommendations (e.g., workout plans, retirement savings advice, alcohol consumption guidance), but outcomes depend on human discipline and follow-through. - Historical grounding: - Adam Smith in The Theory of Moral Sentiments emphasized superior reason and “self-command” to forego present pleasure for greater future benefits. - Household production and the weakest-link problem: - Household production transforms purchases into consumed value through time, effort, and skill; Michael Kremer’s O-ring theory illustrates that other inputs cannot substitute for individual effort when the weakest link fails. - As AI’s informational quality improves, the limiting factor will increasingly be human ability to implement known best actions. - Finding: - The benefits of self-regulation will grow as AI becomes better at supplying relevant information, making cultivation of self-discipline a higher-return investment. ### Human capital (integration and future scenarios) - Development: - Curiosity, critical thinking, and self-regulation are forms of human capital that grow when environments (schools, workplaces) encourage questioning, reflection, and autonomy; environments that discourage these traits erode them. - Singularity scenarios (two speculative paths): - Adversarial scenario: Humans and machines are adversaries (e.g., The Matrix); each generation’s human capital is the means of resistance, prioritizing its accumulation. - Coexistence scenario: AI and humans peacefully coexist; interactions might resemble humans working for large organizations—AI aggregates greater brainpower but compensates humans for their knowledge and skills. Some people may prefer AI-free, low-tech communities relying on member human capital. - Historical stages of technological change: - Mechanization of agriculture, automation of manufacturing, and “algorithmization” of services each freed and demanded human capital in different areas. - Displaced human capital in one stage supported the next stage’s boom (e.g., mechanized agriculture enabled manufacturing boom; automated manufacturing enabled services boom); AI is poised to enable the next boom. - Contemporary evidence: - Firms (example cited: Meta) recruiting human talent with large compensation packages to develop more powerful AI technologies indicate the ongoing importance of human capital. - Finding: - Investing in human capital remains a strong policy priority across scenarios; human capital will remain relevant in new and hard-to-foresee ways and will create value in future sectors. ### Podcast and author note - Podcast summary: - Pablo Peña and journalist Rhoda Metcalfe discuss the article in the December issue of Finance & Development; Peña argues that critical thinking and curiosity will keep humans “in the driver’s seat” despite AI’s performance. - Author: - PABLO PENA is an associate instructional professor in the University of Chicago’s economics department and author of Human Capital for Humans: An Accessible Introduction to the Economic Science of People. *Source: The Human Edge, PABLO PENA; December 2025.* --- ## Content in this bundle - **The Human Edge** - [The Human Edge (Markdown version)](/-/media/files/publications/fandd/article/2025/12/pea.pdf.md){rel="alternate" type="text/markdown"} - [The Human Edge (PDF)](/-/media/files/publications/fandd/article/2025/12/pea.pdf){rel="external" type="application/pdf"} --- _Source: https://www.imf.org/en/publications/fandd/issues/2025/12/the-human-edge-pablo-pena_