Complexity — David Krakauer, Part 4

On: Rethinking Our Assumptions During the COVID-19 Crisis

Transmission Series Ep. 4

Episode: 30

Date: April 28, 2020

Evolutionary biologist.

Review of 5 Transmission Essays.

Key Takeaways

  • Decision making.
    • Improve predictions through trial and error, test and update.
    • Requires good models:
      • Priors: in order to interpret (noisy) data we need to have an expectation.
      • Simulation: in order to have an expectation we need rigorous simulation.
      • Limiting novelty: use what you already know to gain insight about what you don’t.

Key Concepts

David Wolpert on statistical tools for making pandemic predictions

  • Good decision making: test and update.
    • Producing estimates from noisy, changing data.
      • Constant need to go back to the data to make better estimates.
    • Conditional probability.
      • Bayes’ Rule: probability of A given B.
        • B = the background statistic, the prior.
      • Example:
        • A = probability you have a disease if you have tested positive.
        • B = incidence of disease in the population.
        • Reliability of the test varies widely with changing values for B.
      • How to estimate the value of a prior (B)?
        • Requires testing and constant updating as well.
        • Hopefully converge on a value close to accurate situation.

Sidney Redner on quantitative ways to consider the economic impact of COVID-19

  • Good model making: use what you know to gain insight into what you don’t know.
    • Models.
      • Segregate what you know (semi-permanent), from what you don’t.
      • Model’s framework and assumptions are known (minimize novelty, use verifiable assumptions).
      • Model’s outcome provides insight into something new (something harder to verify).
    • Example:
      • Model using simple verifiable math: excess death rate, ICU costs, etc.
      • Use model for broad estimate of COVID-19 costs.
      • Hard to verify novelty: COVID-19 costs are in trillions / billions.
      • Use insight to compare virus costs to preventative costs (likely much lower).
      • Simple model provides benchmark to think about value and costs of preventative measures.

Stephanie Forrest on privacy concerns that arise with a pandemic

  • Immune system and surveillance: generate random responses, filter out the ones that hurt the system.
    • Immune response: generate many random antibodies (somatic hypermutation).
    • Any new element that attacks the body is tagged and deleted.
    • What is left is negative compliment of the “self”, hopefully sufficient to defeat new pathogens.
    • Computer virus: similarly, generate random responses to potential attacks, remove the ones that hurt the program itself.
    • Same approach for contact tracing to ensure that surveillance doesn’t result in privacy issues.

Carrie Cowan on the future of education

  • COVID-19 has accelerated the adoption of different types of learning …
    • Digital: digitally distributed education and massively open online courses (MOOCs).
    • Flipped classroom: prep in advance, use social context to ask questions.
  • … and the need for cheaper, broader education platforms.
    • Costs of (university) education are too high.
    • Degrees are too narrow and specialized (need to make connections, etc.).
    • See also the End of Average on suggested changes to higher education.
  • What we can learn from videogames.
    • Community based, experiential, curiosity-driven free, playful environments in which people can apply their mind.
    • Freedom.
      • Move away from authoritarian, top down approach of knowledge transfer.
    • Diversity.
      • Different time-scales of learning, different ways of learning.
    • Collaboration.
    • Construct and contribute.
      • Construct something, contribute and have it scrutinized.

Anthony Eagan on Federalism in a time of pandemic

  • Constitutions are regulatory networks.
    • Legal operating system of a country.
      • Manages the tension between centralized and de-centralized power.
    • Works well under normal conditions …
      • General consensus on allocation of power.
    • … in time of crisis, things change.
      • A need for and a willingness to accept more centralization may emerge.
    • Construct a dynamic constitution.
      • Different rule systems according to different societal conditions.
  • What we can learn from genomes, our biological constitution.
    • Genomes are the regulatory networks of the body.
      • Allow various cell types to have greater or lesser freedoms in pursuing their function.
    • More complex phenotype -> more “democratic” regulatory network.
      • Complex organisms: everything is connected, less hierarchy.
      • Simple organisms: sparse connections, top-down control.
  • The interplay of simplicity, specificity, constraints.
    • The “shorter” regulatory networks are, the more room for interpretation.
    • The “longer” they are, the more they specify how to act in certain situations.
    • Vary size according to societal needs.
  • Ultimately, constitutions may become more algorithmic and dynamic.

 

Leave a Reply