Scientific Progress and Political Feedback Loops (with Michael Nielsen)
Spencer Greenberg speaks with Michael Nielsen about the rate of scientific progress, the pros and cons of being a para-academic researcher, and the state of feedback loops in American politics. They discuss potential reforms for scientific funding and institutional structures.
Deep Dive Analysis
13 Topic Outline
Analyzing the Rate of Scientific Progress
Measuring Scientific Discovery Importance: The Nobel Prize Tournament
Findings on Scientific Progress and Diminishing Returns
Models Explaining Scientific Stagnation and New Field Emergence
Institutional Sclerosis in Science and Academia
Innovative Approaches to Science Funding and Design
Skepticism Towards Economic Measures of Scientific Progress
Understanding Para-Academia: Independent Research
The Value of Rapid, Iterative Research Approaches
Challenges in Academic Hypothesis Generation and Rigor
Stability of Political Approval Ratings and Media Polarization
Tightening Feedback Loops in Democratic Politics
Improving Societal Metrics and Public Awareness
6 Key Concepts
Diminishing Returns in Science
This concept suggests that as scientific research progresses, it becomes increasingly difficult and resource-intensive to make discoveries of comparable importance to earlier ones, leading to reduced productivity per unit of effort or funding. This can manifest as the quality of top discoveries not improving despite a massive increase in overall resources.
Sclerosis in Scientific Institutions
This refers to a state where academic and research institutions become resistant to change, new ideas, and innovative approaches. It's often attributed to a lack of rapid feedback loops, unlike the corporate world, which prevents new, better-run institutions from quickly overtaking older, established ones.
Randomized Research Grants
An experimental funding approach where, after an initial basic assessment to ensure a proposal isn't 'completely insane,' research grants are awarded via a lottery system. The goal is to increase the variance of funded projects, avoid the biases of expert committees, and prevent a concentration of research on 'hot' topics.
Para-academia
This describes a growing community of semi-independent researchers who operate outside traditional academic institutions. These individuals often have PhDs or academic backgrounds but pursue their own research projects, facilitated by modern communication technology that reduces the need for traditional university resources.
High-Throughput Research Approach
A method of scientific inquiry, particularly in exploratory phases, that involves rapidly testing many different ideas or running numerous small experiments in quick succession. The aim is to learn quickly from each iteration, informing subsequent studies, rather than committing to a single, large-scale, slow research project.
Political Feedback Loops
These are the mechanisms in a democracy by which the actions of politicians are influenced by the reactions, opinions, and needs of their constituents. Effective feedback loops are crucial for error correction in governance, ensuring that politicians are incentivized to act in ways that benefit the population.
7 Questions Answered
Based on a study analyzing Nobel Prize discoveries, the quality of the most important scientific advancements appears to be experiencing strongly diminishing returns, suggesting a slowdown in per-scientist productivity despite a massive increase in resources.
New fields often arise as offshoots of older, more developed fields, effectively creating new 'low-hanging fruit' for discovery, as seen with the development of modern computing from mathematical logic.
Unlike the corporate world with its rapid market feedback loops, academic institutions lack similar mechanisms that couple the quality of organizational ideas to their rate of growth, leading to institutional sclerosis and slow adaptation.
Para-academics gain the freedom to choose their research topics, pursue unconventional ideas, and engage in rapid exploratory work without the typical academic constraints of publishing in top journals or securing specific grants.
The process of developing hypotheses is complex and highly individualized; strategies that work for one student may not work for another, making it challenging to teach in a standardized, repeatable way.
Possible reasons include the president's personal characteristics, a changing and increasingly polarized media environment where people consume news that supports their worldview, or a 'denial of service attack' effect where constant new controversies prevent any single issue from gaining enough momentum for political error correction.
Improving political feedback loops could involve developing and widely publicizing better societal metrics (like a 'score sheet' for national well-being) to increase public awareness, and providing politicians with constant, high-quality data from random sampling polls on a wide range of issues.
25 Actionable Insights
1. Develop Meta-Methods for Science
To overcome diminishing returns in scientific progress, focus on developing “meta-methods” for generating scientific progress, such as dramatically increasing the rate at which new scientific fields are produced or inventing new intellectual tools. This addresses the underlying causes of stagnation and can lead to scientific acceleration.
2. Diversify Science Funding Approaches
Implement a wide diversity of funding strategies for scientific research, experimenting with different timescales, amounts, and recipient criteria. Allocate small amounts to many varied experiments to gather data and build a theory of effective resource allocation in science.
3. Implement High-Throughput Research
Adopt a “high-throughput” or rapid exploratory approach to research, running many fast studies in quick succession. This allows for iterative learning, quickly discarding non-working ideas, and refining understanding of a phenomenon before designing definitive studies.
4. Plan for Real-World Research Impact
When conducting research, explicitly consider the “mechanism of action” for real-world impact beyond mere publication. Actively identify and engage communities or stakeholders who are ready to consume, use, and apply the research to ensure it translates into tangible change.
5. Prioritize Accelerating Science Funding
Allocate funding to “accelerating science” – research focused on developing tools, statistical methods, or other meta-scientific advancements that improve the speed and quality of other scientific endeavors. This indirectly boosts overall scientific progress by making other science better.
6. Couple Ideas to Institutional Growth
To improve scientific institutions, foster a strong coupling between the quality of organizational ideas and the institution’s growth rate. This allows better models to scale and potentially overtake older, less effective ones, similar to dynamics in the corporate world.
7. Publicize Key Societal Metrics
To improve democratic error correction, tighten political feedback loops by developing and widely publicizing a “score sheet” of key societal metrics (e.g., 30 different variables) annually. This would increase public awareness of national performance and create a shared reality, similar to how water levels in a dam influenced public behavior during a drought.
8. Systematically Monitor Public Sentiment
For effective governance, establish constant, ongoing, random sampling polls that ask a wide range of open-ended questions to systematically monitor public sentiment and identify people’s biggest problems. This provides politicians with robust data to inform decisions and improve citizens’ lives, moving beyond anecdotal feedback.
9. Randomize Research Grant Allocation
Implement randomized grant allocation for research proposals deemed “not completely insane” to increase variance in funded projects. This approach aims to foster a wider range of research and potentially uncover huge breakthroughs that might be overlooked by conventional committee-based selection.
10. Provide Small Exploratory Grants
Introduce grants that are easy to obtain, provide small amounts of money, and are explicitly for exploratory research without the immediate expectation of a published paper. These grants should support initial exploration and idea development to inform larger, more rigorous studies.
11. Empower Young Scientists with Grants
Reform grant systems to make it significantly easier for younger scientists (e.g., 20-year-olds) to become Principal Investigators and receive their first grants. This counters the current trend where older scientists are disproportionately funded, potentially stifling new ideas.
12. Invest in Fundamental Science
Allocate funding to “fundamental science” that explores basic phenomena with broad implications, such as electromagnetism or number theory. While not immediately applicable, these foundational discoveries can lead to unexpected and significant breakthroughs over time.
13. Fund Useful Problem-Solving Science
Allocate funding to “useful science” that directly aims to solve specific, real-world problems, such as curing diseases. This ensures research addresses immediate societal needs and challenges.
14. Bypass Consensus for Breakthrough Funding
Be wary of relying solely on consensus-driven committees for funding decisions, as they tend to suppress high-variance, potentially transformative ideas. Instead, seek mechanisms that allow funding for projects that are controversial or not universally agreed upon, akin to venture capital’s approach to identifying unique opportunities.
15. Explore Para-Academic Research
Consider pursuing research as a “para-academic” outside traditional academic institutions, leveraging modern communication technology. This offers the freedom to pursue unconventional research questions and approaches that might not be recognized or funded within conventional academia.
16. Combine Breadth and Depth Research
Employ both breadth-first (trying many ideas in parallel) and depth-first (iteratively focusing on a phenomenon with successive “flashlights”) strategies in research. This allows for both broad exploration of possibilities and deep understanding of specific phenomena.
17. Rachet Up Research Rigor
Employ a “ratcheting rigor” approach in research, starting with highly exploratory, even “crazy” ideas and progressively increasing the level of rigor and confirmation. This iterative process allows ideas to evolve from speculative to well-established.
18. Teach Hypothesis Generation
Integrate explicit teaching on how to develop hypotheses into PhD programs and scientific training. Rather than treating it as a “magical” process, provide structured guidance and diverse strategies to help students generate useful and testable hypotheses.
19. Publicize Key Impactful Metrics
Identify and consistently publicize single, highly impactful metrics (e.g., life expectancy, water dam levels) to foster a strong shared cultural awareness. This can create a collective sense of responsibility and influence individual behavior in response to critical societal issues.
20. Quantify Societal Problems
Develop and utilize categorization systems to quantify the prevalence of people’s biggest life problems (e.g., loneliness, financial insecurity). This data can provide a clearer picture of societal challenges and guide efforts to improve well-being.
21. Program Alongside Symbolic Math
When doing symbolic mathematics, simultaneously write and run code that calculates the numerical representation of your symbols. This parallel programming and symbolic math approach dramatically reduces error rates by allowing real-time verification of transformations.
22. Leverage Integer Sequence Databases
When simplifying mathematical equations, calculate values at several points and input them into an integer sequence database (e.g., OEIS) to find matching sequences. This can probabilistically predict subsequent values and aid in identifying underlying patterns.
23. Fund Late-Career New Scientists
Create funding mechanisms that make it easier for individuals in their later careers (e.g., 60s) with no prior scientific background to become Principal Investigators. This is an experimental approach to see what different contributions such individuals might make.
24. Challenge Narratives with Data
Actively seek out and understand fundamental facts and statistics (e.g., gun death causes, gun ownership rates) that may contradict common narratives. This critical awareness helps challenge assumptions and fosters a more nuanced understanding of complex issues.
25. Assess Scientific Discovery Quality
To assess the quality of scientific discoveries over time, ask a large number of scientists to compare the relative importance of different Nobel Prize-winning discoveries in a pairwise “tournament” style. This provides a proxy measure for the quality of top discoveries.
6 Key Quotes
Harvard is a hedge fund, which happens to have a small educational institution attached to that.
Michael Nielsen
The best way of knowing that he, he, he, he really needs companies that, uh, kind of mucking this up, but that the time to buy is not when everybody is saying it now is a great time to buy. You actually, you need to know something that is not common knowledge.
Michael Nielsen
There's kind of this process that you need to go through where you actually need to try lots of different things and you can't necessarily expect that there'll be a silver bullet at any given time.
Michael Nielsen
If you don't have that feedback loop, there's no mechanism for error correction.
Michael Nielsen
To really have a scandal, it actually takes quite a bit of time where that's the only thing in the spotlight. And actually generating kind of distraction is maybe something which the mediator is poorly equipped to deal with.
Michael Nielsen
Most of the stuff that seems useless actually is useless. It's actually just these rare cases that we point to when something ends up being useful.
Spencer Greenberg
2 Protocols
Spencer's Rapid Iterative Research Approach (Social Science)
Spencer Greenberg- Identify a phenomenon of interest to understand.
- Run a series of quick, exploratory studies (like 'shining flashlights') to probe the phenomenon and learn about its nature.
- Learn from each study, often discarding initial studies as necessary steps to figure out how to design the right subsequent study.
- Use insights from exploratory studies to design a robust confirmatory study.
- Conduct the confirmatory study with high confidence in the expected outcome, having thoroughly explored the phenomenon beforehand.
Michael Nielsen's Math Research Approach (with parallel programming)
Michael Nielsen- Work on a mathematical problem, generating and evaluating many ideas per minute.
- Simultaneously write code to compute the values represented by the mathematical symbols and equations.
- After making a symbolic transformation or change, run the code to numerically verify that the result remains consistent.
- Utilize numerical calculations to identify and debug errors (e.g., sign errors, oversights) that might be missed when working solely with symbols.