2015 Short Course: Exploring Complexity in Social Systems and Economics

On August 25-27, 2015 I attended the SFI short course on complexity in social systems and economics. Information on this course can be found here…

http://tuvalu.santafe.edu/events/workshops/index.php/SFI_Short_Course_2015_-_Home_2015

Here’s the agenda and few highlights. The highlights, of course, are heavily biased by my personal interests.  Note that all speakers are affiliated in some way with the Santa Fe Institute.

David Krakauer (President, SFI): Welcome and Introduction to the Santa Fe Institute

  • We were free to wander around the site.  It’s an extremely nice campus on top of a hill overlooking Santa Fe.  Inside has offices around the edges, some shared some not, with large open spaces in the middle.  Lots of windows and light.  Libraries are always an attraction for me (thanks Mom!) their’s is no exception.  It’s comfortable but it’s the content that’s the real hook.SFI Library
  • SFI is an “idea collider” across disciplines.  They offer month long summer school for graduate students and post-docs in pretty much any field that makes sense for their mission.
  • An SFI project worth checking out is Complexity Explorer, which provides on-line courses and other materials related to complexity science.

Melanie Mitchell: Introduction to Complexity

[iframe style=”width:120px;height:240px;” marginwidth=”0″ marginheight=”0″ scrolling=”no” frameborder=”0″ align=right src=”//ws-na.amazon-adsystem.com/widgets/q?ServiceVersion=20070822&OneJS=1&Operation=GetAdHtml&MarketPlace=US&source=ss&ref=ss_til&ad_type=product_link&tracking_id=hawksprphotog-20&marketplace=amazon&region=US&placement=0199798109&asins=0199798109&linkId=AOUHAL5LAU4PSQZW&show_border=true&link_opens_in_new_window=true”]

  • Melanie has a very cool, very approachable book called Complexity: A Guided Tour.  I’ve read it and highly recommend it.
  • I was introduced to the NetLogo programming environment, which has a lot of fun simulations to play around with.
  • Three, no four, themes of complexity science:
    1. More is different
    2. Simple rules can produce complex behavior
    3. Key concepts are information, adaptation and inductive inference
    4. A lot of insight can come from simple models!

Aaron Clauset: Introduction to Networks

[iframe style=”width:120px;height:240px;” marginwidth=”0″ marginheight=”0″ scrolling=”no” frameborder=”0″ align=right src=”//ws-na.amazon-adsystem.com/widgets/q?ServiceVersion=20070822&OneJS=1&Operation=GetAdHtml&MarketPlace=US&source=ss&ref=ss_til&ad_type=product_link&tracking_id=hawksprphotog-20&marketplace=amazon&region=US&placement=0199206651&asins=0199206651&linkId=Y5OUZHGGBMQWKW6H&show_border=true&link_opens_in_new_window=true”]

  • A general introduction to networks and different ways of characterizing networks.  Networks are vertices or nodes (distinct objects) connected by edges (pairwise relationships). Aaron described four basic concepts relating to networks:
    • Degrees – The number of connections to any given node.
    • Paths – Number of “hops” between two nodes. Includes notions like clustering and small world networks.
    • Position – A measure of positional importance.
    • Community – A group of vertices that connect to other groups in similar ways.
  • Networks are a notion that fit between components viewed as discrete parts and systems viewed as a whole.  Viewing a system as a collection of nodes and edges can provide insights that aren’t available from either the component view or the whole system view.
  • Aaron described a particularly intriguing hard problem, intriguing to me at least: “How does network structure influence network function? robustness, resilience, adaptability, design.” Designing network based systems with specific forms of these characteristics is potentially very interesting when thinking about networks of things (IoT).  He asserted that most work in this area tends to be compartmentalized by discipline with very little general work being done.   This is worth some further research to find out what has been done so far.

Rob Axtell: Agent Computing as a Gateway to Complexity in the Social Sciences

[iframe style=”width:120px;height:240px;” marginwidth=”0″ marginheight=”0″ scrolling=”no” frameborder=”0″ align=right src=”//ws-na.amazon-adsystem.com/widgets/q?ServiceVersion=20070822&OneJS=1&Operation=GetAdHtml&MarketPlace=US&source=ss&ref=ss_til&ad_type=product_link&tracking_id=hawksprphotog-20&marketplace=amazon&region=US&placement=0060925876&asins=0060925876&linkId=6TQVVQCZLIZN6UB3&show_border=true&link_opens_in_new_window=true”]

  • A survey of agent based modeling applications ranging from populations of 1 to 10,000,000,000.
  • At 10^1 he explored Hotelling, who asks the question “Why is their a diamond district in New York City?”  Hotelling’s Law is the principle of minimum differentiation and there’s a great Netlogo simulation for playing around with this.
  • Rob asserted that there are three gross forces revolutionizing many of the social sciences:
    • Routinization of human behavioral experiments
    • Availability of ‘big data’/comprehensive data/ administratively-complete data/universe of data
    • Large-scale agent computing for explaining data

Brian Arthur: Complexity and the Economy

[iframe style=”width:120px;height:240px;” marginwidth=”0″ marginheight=”0″ scrolling=”no” frameborder=”0″ align=right src=”//ws-na.amazon-adsystem.com/widgets/q?ServiceVersion=20070822&OneJS=1&Operation=GetAdHtml&MarketPlace=US&source=ss&ref=ss_til&ad_type=product_link&tracking_id=hawksprphotog-20&marketplace=amazon&region=US&placement=0199334293&asins=0199334293&linkId=SYJYIWAWVSQISCII&show_border=true&link_opens_in_new_window=true”]

  • All the sciences are shifting from order, formalism, determinacy, deductive logic, stasis to organicism, contingency, indeterminacy, inductive reasoning, evolutionary openness.
  • Non-equilibrium is the natural state of the economy due to fundamental uncertainty and changing technologies.  Lots of temporary phenomena such as bubbles and crashes.
  • A great quote from Max Planck, ““Science advances one funeral at a time.” Given the somewhat controversial nature of complexity science this quote was repeated at least once or twice more during the class.

Deborah Strumsky: Technological Change as an Evolutionary Process: A Complex Systems Perspective

(Still waiting on a copy of the slides.  This is from memory and notes.  I’ll update when slides are available.)

  • Schumpeter has very strict definitions for invention, innovation, and diffusion.  Basically an innovation is an invention that’s been commercialized.  A diffused innovation is a successful invention.
  • A way to think about technology evolution is searching for an optimum on an NK landscape.  The downside of this approach is that it assumes that you can jump to any arbitrary location on the landscape.  It doesn’t take into account the usually required progressive reframing from trying and learning.
  • Really interesting work on the U.S. patent database looking at both the classification of inventions and the life-cycles of various technologies.

Mirta Galesic: How Interaction of Minds and Social Systems Shapes our Judgments

  • Regarding how well we know other people: Biases are related to homophily.  People know their immediate social circles well, and don’t know much about other social environments such as the general population.
  • Regarding group judgement: Condorcet’s Jury Theorem which seems to indicate that groups of smart people are smarter and groups of dumb people are dumber.  Group size matters but so does the nature of the problem.  For some problems larger groups make worse decisions.
  • Regarding why humans are so cooperative: Possibly critical factors in human evolution. Factors include:
    • A harsh, but not too harsh, environment.
    • Profitable foraging in small groups.
    • Scarce population and fission-fusion dynamics.

Rajiv Sethi: Agent-Based Computational Economics

[iframe style=”width:120px;height:240px;” marginwidth=”0″ marginheight=”0″ scrolling=”no” frameborder=”0″ align=right src=”//ws-na.amazon-adsystem.com/widgets/q?ServiceVersion=20070822&OneJS=1&Operation=GetAdHtml&MarketPlace=US&source=ss&ref=ss_til&ad_type=product_link&tracking_id=hawksprphotog-20&marketplace=amazon&region=US&placement=0393329461&asins=0393329461&linkId=XDRKEVJ2DPUJYHES&show_border=true&link_opens_in_new_window=true”]

  • Agent-based computational economics (ACE) is the computational modeling of economic processes… as open-ended dynamic systems of interacting agents[1. Leigh Tesfatsion]. Agent-based models instead make disequlibrium dynamics explicit.  Conway’s Game of Life is an early example.
  • Associations arise from decentralized, uncoordinated choices.  It’s very hard to reach integrated equilibria and very easy to reach segregated equilibria.  See the Schelling Segregation Model[1. Schelling, Thomas C. (1978). Micromotives and Macrobehavior, Chapter 4].
  • Simulated the interaction of different high-frequency trading behaviors and explored whether or not these interactions could lead to extreme market events such as flash crashes.
    • Liquidity traders place orders based on consumption/saving needs.
    • Information traders buy/sell mispriced securities.
    • Algorithmic market makers post prices to profit from spread

Scott Page: Confronting Complexity With Many Models

[iframe style=”width:120px;height:240px;” marginwidth=”0″ marginheight=”0″ scrolling=”no” frameborder=”0″ align=right src=”//ws-na.amazon-adsystem.com/widgets/q?ServiceVersion=20070822&OneJS=1&Operation=GetAdHtml&MarketPlace=US&source=ss&ref=ss_til&ad_type=product_link&tracking_id=hawksprphotog-20&marketplace=amazon&region=US&placement=067443000X&asins=067443000X&linkId=5RNKPZPKDIX4XCTG&show_border=true&link_opens_in_new_window=true”]

  • Entropy (Information Entropy) is the average amount of information gained after a realization.  It’s a measure of uncertainty contrary to variance, which is a measure of dispersion. It is a measure of information that could be present. Complexity is the entropy in the system minus the randomness in the system.  Excess entropy is predictive information.
  • “Truth is the intersection of independent lies.” – Richard Levin[1. The Strategy of Model Building in Population Biology].
  • Models tend to make better predictions than people.  People combined with models tend to make better predictions than models alone.

Geoffrey West: Is a Quantitative, Predictive Science of Cities and Companies Conceivable?

(Still waiting on a copy of the slides.  This is from memory and notes.  I’ll update when slides are available.)

  • In 2014 the U.S. will be 80% urbanized.  Every 12 hours there’s another Santa Fe on the planet. The Fate of the Planet = The Fate of our Cities.
  • “What is a city but the people?” – Shakespeare. For cities physicality is secondary, social organization is primary. It’s a buzz of activity wedded to the infrastructure. Cities are places that facilitate human interaction.  They are hierarchical modular structures. Are cities and companies just very large organisms satisfying the laws of biology?
  • Systems that scale sub-linearly eventually die.  Animals, including humans and companies are examples.  Systems that scale super-linearly don’t generally die.  Cities are an example.

Faculty Panel Discussion

SFI Panel

  • Is this a paradigm shift?
    • Some pushback on the notion of paradigm shifts lately. Interdisciplinary approaches represent a paradigm shift.
    • The change seems less discontinuous than prior paradigm shifts. If there were some sort of generalized theory we might might feel more comfortable calling it a paradigm shift.
    • There’s definitely a shift in data analysis.  In the past we’ve treated data as hard and expensive to collect.  There are a lot of tools based on a small N world.
    • Fundamental ideas in complexity science include non-linearity, emergence, and moving away from normal distributions. People still act as if the world is linear and normally distributed.
  • How can we use some of this thinking to better understand cultural morphing?
    • Would like to model algorithms and the social structure.  Finding what’s important. Computational sociology.  How beliefs are adopted.  How networks are pruned.  Find simple models with few parameters.  Skeptical about most general models.  Trick is the combination of rules.
    • When does culture trump strategy and vice versa?  What is culture? Conditionality/substate arguments. Dramatic culture contrasts.  Institutions create the culture.
    • Diversity in a network scales with the size of the network.  Critical to creating super-linear outcomes.
  • Uncertainty about this uncertainty work…
    • Many of us are searching for the simplicity that underlies complexity, if there is any.  Put some of these challenges into quantitative terms.  Some phenomena do have regularties at a coarse grained level.  Can you be more predictive in terms of the growth and evolution of the system?  Where are the boundaries? Historical contingency and path dependent.
    • People who live in the world where complex is a dirty word are living in a fantasy.  There’s no way to get the complexity out of the real world.  SFI is trying to provide a language for these things and makes it less scary. Makes it less dirty and more interesting.
    • Study institutional design.  Some things we want to be complex, some we want to be more regular.  Discovery of the efficient equilibria.
Bookmark the permalink.

Leave a Reply