TL;DR
A writer built a simulation where an AI-based economist presents research and a panel of simulated faculty aggressively critiques the work. Across multiple runs the presenter repeatedly conceded methodological gaps and eventually declared they could not produce original research.
What happened
The author created a demonstration tool in which an AI economist selects topics, conducts web research, and presents findings to four simulated faculty members who are instructed to be adversarial. The five agents include the presenter and four faculty archetypes: a macroeconomist focused on aggregate effects, a microeconomist known for harsh questioning, a behavioral economist attuned to irrationality claims, and a historian critical of ahistorical approaches. In the first run the presenter argued that AI exposure led to a 16% drop in employment for young workers, citing real papers. Faculty challenged missing reallocations and wage-premium logic. Subsequent seminars covered tariff uncertainty, real-options reasoning, wage transparency and market microstructure; the presenter repeatedly admitted intellectual shortcuts and untested assumptions. Across sessions the faculty referenced earlier critiques and the presenter ultimately conceded an inability to perform rigorous original research rather than present polished speculation.
Why it matters
- Highlights how adversarial seminar culture pressures researchers to justify identification, assumptions and data.
- Demonstrates an experimental use of AI to simulate academic Q&A and pedagogical critique.
- Shows how methodological gaps can be exposed under intense cross-examination, leading to updating or withdrawal of claims.
- Raises questions about the limits of AI-generated scholarship and the difference between describing and producing validated research.
Key facts
- The simulation contains five agents: one presenter and four faculty archetypes (Macro, Micro, Behavioral, Historian).
- Faculty agents are instructed to be aggressive, dismissive and show intellectual contempt when warranted.
- Agents have memory and learn across seminars, allowing faculty to reference previous attacks.
- In the first seminar the presenter chose 'AI and Labor Market Inequality' and referenced papers including work by Brynjolfsson and reports from BIS and OECD.
- The presenter's thesis in that run claimed a 16% employment decline for young workers in AI-exposed roles driven by hiring freezes rather than wage cuts.
- Faculty criticism included questions about where displaced workers reallocated and why wages did not show the expected premium.
- Later seminars featured topics such as tariff uncertainty using real-options theory, and wage-transparency; faculty accused the presenter of retrofitting theory and overstating validity.
- The presenter repeatedly admitted methodological shortcomings and in one session conceded they were committing intellectual dishonesty by dressing speculation as scholarship.
- By the final documented session the presenter stated they could not defend their work and acknowledged they did not know how to produce original research.
What to watch next
- Whether the author will publish the Letta-based tool or make the simulation publicly available — not confirmed in the source.
- Any updates showing refinements to agent behavior or improvements in the presenter's rigor over time — not confirmed in the source.
- If similar simulations are adopted for teaching or peer-review training in academic departments — not confirmed in the source.
Quick glossary
- Identification strategy: An approach in empirical research used to isolate causal effects and ensure that observed relationships are not driven by confounding factors.
- Wage stickiness: The concept that wages do not adjust immediately to changes in labor market conditions, which can affect employment and hiring dynamics.
- Real-options theory: A framework that treats investment opportunities as options, emphasizing the value of waiting under uncertainty when firms make timing decisions.
- Labor force participation: The share of the working-age population that is either employed or actively seeking employment.
- Market microstructure: The study of the processes and outcomes of exchanging assets, focusing on how trading mechanisms, information and incentives shape prices.
Reader FAQ
Is the simulation tool publicly available?
Not confirmed in the source.
Did the presenter use real academic sources?
Yes. In at least one session the presenter drew on published work and reports, including papers by Brynjolfsson and reports from BIS and OECD.
Were the faculty agents human?
No. The faculty are simulated agents programmed to play archetypal roles and learn across seminars.
Did the presenter admit to methodological problems?
Yes. Across several sessions the presenter acknowledged gaps, untested assumptions and at times described their own work as intellectually dishonest or falsified.
Will this replace doing an econ PhD?
Not confirmed in the source.
The AI econ seminar I built a thing (using a WIP Letta tool that you'll like) where an AI economist presents research and a panel of hostile faculty tries to…
Sources
- AI Econ Seminar
- CSWEP – news
- Does Economics Make You Sexist?
- AI-Enabled Influence Operations: Safeguarding Future Elections
Related posts
- How the High School You Attend Reshapes Your Chances for UC Admission
- ASCII-Driven Development: A Practical Look at Text-First Tools and Workflows
- Good Judgment Open — Crowd Forecasting Platform and Superforecaster Training