Actions

Complex Systems Summer School 2016-Projects & Working Groups

From Santa Fe Institute Events Wiki

Complex Systems Summer School 2016

Note: if you are not going to continue one of the project ideas below during the CSSS please put them in the archived projects section below.

Contents

Projects

Cities as Ecologies of Plans

Keywords

ecology, economics, cities, civil engineering

Summary

The idea that the macro economy is an ecology of interacting plans has been gaining force among heterodox, complexity-aware economists. Similarly, cities are increasingly understood as not only having ecologies, but as ecological systems themselves. The project would be to create an agent-based model (likely in NetLogo) that bridges between the literature in economics and civil engineering, envisioning cities as ecologies of interacting plans.

Plans are created by individuals as a kind of dynamic chain of actions and strategies to attain some kind of goal. Goals can be anything: money, love, power, career, altruistic ends. Plans interact with the plans of others: if you want to move into a city, all other things being equal and assuming no vacancies, you need someone else to move out. In that sense, your plan to move in is coordinated with someone else's plans to move out. The plans and goals of people who move in/out of and live in cities directly affect the city as an ecological system. A suggested model: look at how a city's resources, spatial distribution, and other metrics are affected by the plans (dynamic strategies) and goals of its inhabitants. Is a lack of overall coordination (disharmony with the plans/goals of other inhabitants) with negative consequences for other inhabitants reflected in the ecological system of the city?

Some References

Group Contact

Abigail Devereaux (abigail.devereaux@gmail.com)

Interested Participants

  • Ben Zhu (b.zhu@tudelft.nl) (I am from the field of Industrial Ecology, using biological metaphors to solve social/industrial sustainable problems)(I am interested in applying universal Darwinian thinking on economy)
  • Danilo?

Inferring the Structure of a Network

Keywords

critical infrastructure, information theory, network theory, agent-based model

Summary

Deregulation of the US power grid has opened the door to speculators who trade futures contracts that are based on the minute-by-minute price difference for a megawatt hour between two generators (think nodes). The generators are connected by transmission lines (think edges) that must, in general, follow Kirchhoff's laws of current and potential. We want to explore what information exists in these location based prices regarding the structure of the underlying network of generators, consumers, and transmission lines. Public historical data on the price at each generator and the power both forecasted (24 hours ahead) and realized at each generator exists for certain regions of the US and I have built a small database of 2 years worth of this data for roughly 15 zones of the New York interconnect. An interesting side note, the power forecasted and demanded at one particular node tends to be periodic and easy to forecast. The price (which theoretically is driven by demand) at that same node exhibits multifractal properties, apparent heavy-tails in the increments, and long memory. The process that converts bids, offers, and the physical constraints of Kirchhoff's laws is usually a linear program that minimizes cost across all generators that can contribute to the demand.

Group Contact

James Thompson (j7sthompson@gmail.com)

Interested Participants

  • Daniel Biro (daniel.biro@med.einstein.yu.edu)
  • Charlotte (charlotte.james@bristol.ac.uk)
  • Ryan McGee (ryansmcgee@gmail.com)
  • Mika Straka (mika.straka@imtlucca.it)
  • Harrison Smith (hbs@asu.edu)
  • Aina Ollé Vila (aina.olle@upf.edu)

Resource Curse: paradox of plenty

Keywords

resource curse, agent-based models, networks, economics, competition, self-similarity

Summary

Economist have noted that an abundance of natural resources (usually non-renewables) is apparently linked to worse economic indicators, an absence of democracy, and a number of other negative effects regarding the general health of the nation or region. A few of us have discussed exploring the dynamics of this alleged curse using the tools of complexity science. One possible question to explore is does the resource curse exhibit self-similarity at smaller scales of time, resource quantity, and competitors for the resource? Anecdotally one could think of this as the difference between say a mineral rich African nation and a Walmart store on Black Friday. Both have finite resources and competitors attempting to reap the resource while it exists. And both have been known to invoke violent responses.

We are open to exploring any other ideas related to the resource curse.

Group Contact

James Thompson (j7sthompson@gmail.com)

===Interested Participants=== please join the slack channel # resourcecurse and email Catriona (c.sissons@auckland.ac.nz) to get added to the Dropbox folder

  • Catriona Sissons (c.sissons@auckland.ac.nz)
  • Daniel Biro (daniel.biro@med.einstein.yu.edu)
  • Scott Armstrong (armstrongsb1@cardiff.ac.uk)
  • Charlotte (charlotte.james@bristol.ac.uk)
  • Evelyn Strombom (evelyn.strombom@gmail.com)
  • Rana Azghandi(rana.azghandi@gmail.com)

Emergence of symmetry

I'm super fascinated by the question of why there is so much symmetry in nature? For example, in the biological system, what are the evolutionary models that can explain such high percentage of species which exhibit certain type of (approximate) morphological symmetry? What can we learn from this mechanism?

Inspired by different examples in nature, we can try to invent network models that exhibit emergence of certain symmetry and structure formation. We can also analysis the chaotic behavior and entropy flow in the model we invented, and might discover some new phenomena of emergence!

All ideas and discussions are welcome!

Group Contact

Marcus Nordström manordst@gmail.com; Lin-Qing (lchen@ptip.ca)

Interested participants, please sign up below

Julia Adams jadams@wellesley.edu Markus.junginger@mhp.com

Movie citation network

Notice: we will be merging with the other group that works on movie data (Temporal social networks (and dialogues) in film). If interested, please see the section for details.

Summary

IMDB has an open access dataset at: http://www.imdb.com/interfaces. Our project is based on the dataset and we have current several potential research questions:

1) The sleeping beauty of movies. Based on the "shoutouts" of one movie to another, a "citation" network exists that connects millions of movies. The "sleeping beauties" had been identified in scientific paper citation networks, which are papers that became frequently cited a long time after publication. We are interested in finding similar (or different) patterns in the movie citation network.

2) If atypical combination of movie properties/citations/genres could predict its popularity.

3) A more general idea is the gender issue in movie industry, such as the different network structures of actor and actress.


We welcome any feedback!

Group Contact

Yizhi (Elise) Jing (jingy@indiana.edu)

Lu Liu

Interested participants, please sign up below

Cheng Jin (chengkim@zju.edu.cn)

Emergence of power structures in networks

Summary

Note: This is now a separate project from the prestige goods project.

High-level questions: How do (social) power structures emerge in complex networks? How do hierarchies restrict/modify information flow through networks? How does structural power (e.g., due to network hierarchy) differ/relate to intrinsic power (e.g. some nodes/people have more "skill/worth" regardless of their network position).

One idea I have is looking at the interplay of these dynamics by simulating (social learning) games on networks; for example, games where agents have to share information to estimate the worth of external items (like stocks, prestige goods etc.). Other ideas would involve analysis of how information flow is affected by hierarchy etc.

At the first meeting we talked about how power is multi-faceted, and informed by multiple interaction or resource types. I really liked the idea of taking a multi-layer approach, running games on multiple networks (that may also relate to each other) which both contribute to power structure/hierarchy.

Group Contact

Lula Chen (nchen3@illinois.edu)
Will Hamilton (wleif@stanford.edu; wrote this is description based on chat with Lula; not my area of research)

Interested participants, please sign up below and join the Slack channel [sfi-csss-2016.slack.com/messages/power, email Will if not yet on Slack]

Santiago Guisasola (sguisaso@uci.edu)
Joris Broere j.j.broere@uu.nl
Crosato Emanuele (crosato.emanuele@gmail.com) - I can use information theory to investigate directional dependencies, might be useful?
William Leibzon (william@leibzon.org) - partially my area of research. question is what you mean by "power" - I'm not on slack, pls send email
Kelly Finn (kfinn@ucdavis.edu) - I have done some work with social power networks in monkeys
Moriah Echlin (moriah.echlin@gmail.com) devrim ikizler

Meetings


We're meeting in the coffee shop at 6:30, June 15th.

A genetic model of music evolution

Keywords

mutation, selection, recombination, cultural evolution, market ecology, music

Summary

We would like to build a genetic inspired model to describe the mechanism of music evolution in history. Basic idea is that different elements of characters in music is analogous to genes. Mutation creates new music character and language material, hence it is the source of musical diversity (genetic diversity). Migaration and cultural exchange propagate a music style to different community. Recombination allows different elements to form certain combinations. Natural selection, which is influenced by audiences' musical culture background, aesthetic values and market ecology. Selection chooses certain combination over others and when the number of differences reaches a critical point, it can be identified as a new style.

All ideas and discussions are welcome!

Interesting addition and comments to the idea

Group Contact

Lin-Qing Chen (lchen@pitp.ca) ; Chenling Xu (chenlingantelope@gmail.com)

Interested Participants

  • Jesús Arroyo (jarroyor@umich.edu)
  • Santiago Guisasola
  • Donovan Platt (donovan.platt@students.wits.ac.za)
  • Marcus Nordström (manordst@gmail.com)
  • Mark McCann ( Mark.McCann@Glasgow.ac.uk)
  • Elise Jing (jingy@indiana.edu)
  • Cheng Jin (chengkim@zju.edu.cn)

Meeting time

1) June 15, 8:15~9:00 pm in the first floor coffee shop.

Quantifying Trust

Keywords

Agent Based Modeling, Information Theory, Game Theory

Summary

To me there are some simple rules of trust:

  1. The more we trust someone the less questions we are asking about their activity
  2. By trusting someone, a person reduces the amount of information he needs to process. e.g. If a pedestrian does not have a clear view of the road then he might follow other pedestrians in front of him. By following them he is reducing the amount of information he needs to process.
  3. If there is a "perfect trust" in a social system, then the system seems to be more ordered. For example in a military system a soldier blindly obeys the order of its officers. This makes the system more ordered.
  4. On the other hand the less there is trust the more disorder in the social system. So quantifying trust could be a way to quantify the disorderedness or entropy of a social system.
  5. But there is a catch. It is easier to propagate a misinformation in a system where everyone blindly trusts each other.

My intuition is that we can use both information theory and game theory to quantify trust this way. The problem is I don't have a dataset yet that can validate the rules above. But we can use the above observations to formulate an agent based model where we can tune the parameters to see if we can get any interesting observation.

Interesting insights to the idea

We will be dumping our insights and ideas in this google doc from now on:

https://docs.google.com/document/d/1LI4RhwwaosKljJsXmLOS3DPh6nA0-dNwL1EfMSl5WAY/edit?usp=sharing

We can also discuss and chat about the idea in Slack in the channel #trust. Anyone does not have an account can send me an his/her email so that I can invite into slack.


Please add your comments and idea about the project here.

Are there shortcuts to building trust? For example, in the area of financial instruments, can the use of credit ratings substitute for more detailed due diligence? What are the implications of such short cuts? [Nai Seng]

Group Contact

Syed Arefinul Haque, haque.s@husky.neu.edu

Interested Participants

Donovan Platt (donovan.platt@students.wits.ac.za)

Frank Marrs

Nai Seng Wong (fsgwns@gmail.com)

Mika Straka

Lu Liu (luliu@psu.edu)

Xiongrui (John) Xu (xuxr007@gmail.com)

Marcus Nordström (manordst@gmail.com)

Rana Azghandi

Matteo Morini

Rudi Minxha

Meetings

We have created a public group #trust in slack for the online conversation. We will also post the face-to-face meeting schedules below:

Wednesday 7:30, June 15, 2016

We are meeting in front of the coffee shop at 7:30 for a general discussion

Semantic word game

Keywords

cognitive science, language, games, network routing, free association

Summary

See http://pybossa.socientize.eu/pybossa/app/Semantics/ to play the game. The goal is, given a starting word, navigate through forced choices to a known target word. The global idea is to understand the effect of context and how individuals can manipulate semantic space online (like we constantly have to do in natural conversation. A project idea is to construct random models where we can see influence of the target word. For example, if we have directed search from the start node to the end node as well as a directed search path from goal to start, can we combine these paths to capture human performance.

Data is available for download at https://pybossa.socientize.eu/pybossa/app/Semantics/tasks/export and looks like (JSON) {"info": "ILLUSION~FANTASY~16~FANTASY~ISLAND~6~ISLAND~OCEAN~4","user_id": 1732, "task_id": 13792, "created":"2014-04-01T11:26:06.479946", "finish_time": "2014-04-01T11:26:06.479966", "calibration": null, "app_id": 430,"user_ip": null, "timeout": null, "id": 3211605}, This user went from illusion to fantasy (16 seconds), then from fantasy to island (6 seconds) and so on.

Research question is still a bit unclear but a while back there was more than 10478 games. Additionally something similar has been done using wikipedia http://snap.stanford.edu/data/wikispeedia.html

Group Contact

Nicole Beckage nbeckage [at] gmail.com

Interested participants, please sign up below

Cooperative game: Modeling communication and innovation

Keywords

agent based models, network science, cooperation, iterative games

Summary

Data available. Game design:Each player is tasked with building a 'team' of 5-6 characters out of a league of 40-60 players. The team receives a numeric (deterministic) score. This is played iteratively for a fixed (known) set of rounds. Cooperation comes in because at any time each player can see the teams and scores of all other players and may copy, select characters from or otherwise utilize all other players teams. The goal is to maximize the score of the group. Group size varies from 1 to 9 people in a session. Repeated games are played between the same team and difficult varies systematically within a full set of games.

Research question is still not nailed down but some ideas model the communication structure (in terms of copying other peoples choices) in a network and look if certain structural elements are correlated with improved performance model learning and improvement in terms of sharing information over iterative runs and across group size look at instances of sweeping change as innovation and study propagation of this event through the group.

Group Contact

Nicole Beckage nbeckage [at] gmail.com

Interested participants, please sign up below

  • Mika Straka
  • Joris Broere, j.j.broere@uu.nl

Meeting schedule

Tomorrow at SFI first thing after we start working on projects


Temporal social networks (and dialogues) in film

Keywords

movies, films, social network analysis, pop culture

New subpage here: Movie Project

Summary

We have three cool datasets for analyzing social networks and interactions of characters in films (i.e., the "social" networks of the characters in the movies, based on who appears with and talks with who).

  • Dataset 1: pre-processed dialogues from around 300 movies + interesting IMBD metadata (this dataset is public; www.mpi-sws.org/~cristian/Cornell_Movie-Dialogs_Corpus.html).
  • Dataset 2: detailed information on the temporal scene structure (who is on screen when + scene breaks etc.) for thousands of movies + metadata (e.g., rotten tomatoes scores); this data is more private/sensitive and is currently somewhat messy, but we can definitely use it for this project. It should be easy to extract dynamic networks from which characters are on screen together.
  • Dataset 3: high-level static (i.e., non-temporal) network structures for character interactions in 700 movies (this dataset is public; www.moviegalaxies.com; not totally clear how the data was constructed).

The plan is to use this data to give some really interesting insights into the art of storytelling and motifs/archetypes that re-occur.

Main questions (that we have data for):

  • Can we predict genre, critic scores, etc. from the network interaction structures?
  • How is gender represented in the interaction structure? And how has this evolved over the last 40 years?

Group Contact

Will Hamilton (wleif@stanford.edu)

Interested participants, please sign up below and join the Slack channel [sfi-csss-2016.slack.com/messages/movie-networks, email Will if not yet on Slack]

  • Juste Raimbault (juste.raimbault@polytechnique.edu) [definitively in, but not as main project. Can propose analyses such as NLP on dialogues or spatial analysis if there is spatial metadata.]
  • Andy Mellor (mmasm@leeds.ac.uk) - Very interesting project but perhaps a little to close to my PhD work. Definitely interested in the tempora network aspect/temporal motifs.
  • Michael Schaub (michael.schaub@uclouvain.be)
  • Moriah Echlin (echlinm@uw.edu)
  • Elise Jing (jingy@indiana.edu)
  • Thomas Zhang(tomzhang@umich.edu) (Would very happy to join the discussions and contribute ideas. Probably would not do this as my main project here at SFI due to similarity to my own research.)
  • Lu Liu
  • Catriona Sissons (c.sissons@auckland.ac.nz)
  • Dan Biro (daniel.biro@med.einstein.yu.edu)
  • Harrison Smith (hbs@asu.edu)

Emergence of project groups in CSSS 2016

Keywords

subgroups, team participation, satisficing, networks

Summary

We want to model team emergence in cross-disciplinary academic research.

- Is diversity important? 
- What makes teams productive? What makes teams fun?
- How do teams form information networks?
- How do members satisfice?
- How do members select their teams and how does that align with success of the group?
- What are the conditions for maverick rigor?

Merged from another project on same topic field study of group formation in Summer Schools we can use data on sex, nationality, profession and bio sketches data on previous summer schools could be available as well

(Simon Carrignon about the bio I was playing with it today: | here some lines of python code to grab the bio of all the people of this year. I played with mallet and the basic R cluster tools and it gives some stuff like the tree on the right

tree test

there is funny things if one look carefully, but is really just naive tests I am not a specialist at all. But at least the python code can be usefull)

There is a nice paper about collective intelligence http://science.sciencemag.org/content/330/6004/686 which can also be useful

We checked with Juniper and she can provice data for sex,nationality, science interest, but probably only going back to 2011

Group Contact

Ross Buhrdorf rbuhrdorf@gmail.com
Jacob Hunter, jacob.hunter@pnnl.gov

Interested

Justin Williams

Will Lee, wilee@vt.edu or wlee@mitre.org

Andy Mellor (mmasm@leeds.ac.uk)

Ulya Bayram (ulyabayram@gmail.com) This could be an interesting side project for me, I'm definitely interested! And I think this project can be merged with the one up above I think.

Anjali Tarun (anjalibtarun@gmail.com)

Dmitry Alexeev (exappeal@gmail.com)

Mark McCann Mark.McCann@Glasgow.ac.uk

William Leibzon <william@leibzon.org> - similar to others I have interest in this area but don't see as being a main project

Abigail Devereaux

Santiago Guisasola <sguisaso@uci.edu>

Simon Carrignon

Resources

An interesting lecture on diversity and similarity from an agent based modeling perspective: [1]

Who is leading? the distribution of the number of edition made by the users on this page (Complex_Systems_Summer_School_2016-Projects_%26_Working_Groups) during the 2 first days (meta meta study :D )

Distrib.png

NY subway microbiome

Keywords

networks, ecology, subway , city , communities

Summary

I am working with international consortia - Metasub one of the projects is well known NY subway map of microbes http://d2zahwnsqpmout.cloudfront.net/map/

Chris Mason from Cornell is willing to share the data they have collected in different stations and different parts of the stations

Microbial communities reflect history of the stations as well as people migration - interesting would be to create a model for microbe spread along the subway and exchange

Other data on NY could also be applied

Original paper from Cell Systems http://www.cell.com/pb/assets/raw/journals/research/cell-systems/do-not-delete/CELS1_FINAL.pdf

additional information on network inference http://psbweb05.psb.ugent.be/conet/karoline/documents/conferences/Detecting_bacterial_associations_in_human_microbiome_Bertinoro_2012.pdf

mutual information computation - see section on mutual information https://www.dropbox.com/s/vhm33gj2tsxqla0/PIIS2211124715015442.pdf?dl=0


Group Contact

Dmitry Alexeev (exappeal@gmail.com)

Interested participants, please sign up below

  • Daniel Biro (daniel.biro@med.einstein.yu.edu)
  • Michael Schaub (michael.schaub@uclouvain.be)
  • Ruichen Sun(r5sun at ucsd dot edu)
  • Nicole Beckage (nbeckage@gmail.com)
  • Chris Revell (cr395@cam.ac.uk)
  • Emanuele Crostao (crosato.emanuele@gmail.com)

Genome of Law

Summary

Off-the-cuff discussion about how a given set of laws/statutes/legal regulations might have similarities to genetic regulatory networks. More of an exploratory look, but might have some compelling things to be found. James has study from a colleague at MITRE looking at networks of laws and how they reference each other, and there are similar toe-dipping studies, but nothing rigorous.

Group Contact

Temporary: JP (JP@santafe.edu)


Interested Participants

Catriona
James Thompson
Moriah
Charlotte

The assembly of plant-pollinator networks

Keywords

networks, mutualisms, ecology, restoration, succession, assembly

Summary

I have a 10 year dataset of ~1500 observations of pollinators (bees, flies, butterflies, wasps etc.) visiting plants in a native plant restorations (hedgerows) in the Central Valley of CA. The assembling communities are paired with unrestored field margins (controls) and mature (non-assembling) hedgerows. The goal would be to examine how and why the structure of the network is changing through time. How are the individual species changing their interaction patterns? What does this mean for the topology/resilience of the network? There is also a spatial dimension (the meta-population dynamics of the networks?) that could be explored.

For more information on the dataset, please see https://nature.berkeley.edu/~lponisio/wp-content/uploads/2014/12/ponisio-2016-704.pdf

Group Contact

Lauren Ponisio (lponisio@gmail.com)

Interested participants, please sign up below

Ryan McGee (ryansmcgee@gmail.com)
Daniel Biro (daniel.biro@med.einstein.yu.edu)
Dima Alexeev exappeal@gmail.com
Lindsay Todman (Lindsay.todman@rothamsted.ac.uk)
Chris Revell (cr395@cam.ac.uk)
Julia Adams (jadams@wellesley.edu)
Marilia P. Gaiarsa (gaiarsa.mp@gmail.com)
Asher Mullokandov (asher.mullokandov@gmail.com
Michael Schaub (michael.schaub@uclouvain.be) (I am interested, but I feel there might be too many people already..)
Frank Marrs
Anjali Tarun (anjalibtarun@gmail.com)
Nicole Beckage (nbeckage@gmail.com, I am interested in hearing about the project and method ideas but would want to be more in the background and slightly less involved)

Early Warning Signals in a Disease Network

Keywords

networks, disease, regime shifts

Summary

In 2011 a major avian influence outbreak occurred in the domestic ostrich industry in the western cape of South Africa. This outbreak resulted in massive losses to the industry, and the eradiation of a significant number of farms. I have published papers looking at how the movement of birds may have caused the network to become more vulnerable over time (http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0086973), and also looked at how the system self organized following the crash (http://onlinelibrary.wiley.com/store/10.1111/1365-2664.12486/asset/jpe12486.pdf?v=1&t=ipgcjacz&s=b6f493d011a1e9cafa5849eadd8cb8d59be7b055).

I would like to investigate whether there were early warning signals of a regime shift. This work could relate to anticipating critical transitions, and looking to see whether this could present an empirical example of the presence of early warning signals.

I currently have data from 2005-2014 but could easily obtain data until 2016. I am really open to ideas of how to work with this data.

Interesting addition and comments to the idea

Please add your comments and idea about the project here.

Group Contact

Christine Moore (christine.moore@ouce.ox.ac.uk)

Interested Participants

  • Jeffrey Emenheiser (jemenheiser@ucdavis.edu)
  • Frank Marrs
  • Usama Bilal <ubilal@jhmi.edu> . This sounds very interesting! My experience is restricted to humans, but many of the things that interest me include regime shifts, so I think I'll enjoy studying regime shifts in a different organism
  • Jesús Arroyo (jarroyor@umich.edu)
  • Thomas Zhang (tomzhang@umich.edu)
  • Marilia Gaiarsa (gaiarsa.mp@gmail.com)
  • Charlotte James (charlotte.james@bristol.ac.uk)
  • Cheng Jin (chengkim@zju.edu.cn)

Meetings

We're meeting over breakfast at 8:15 on June 16th.


A preliminary model of the coupled human-natural system of swidden agriculture

Summary

Swidden agriculture, also known as Slash-and-burn is about as old as agriculture itself. It exists in diverse variants practiced by 200 to 500 million people in different regions of the world; all of these forms involve the slashing and burning of portions of forest, hence its name. There has been a historic controversy regarding swidden agriculture, with some publications presenting it as a destructive force that contributes to global deforestation and other publications highlighting its sustainability and ecological benefits when practiced as a means of subsistence. This controversy shows the need of tools to further research and assess the benefits and costs of swidden agriculture.

This project idea will address the following research question: How do human activities interact with the ecological landscape and the sustainability of swidden agriculture?

This project idea aims to produce a simple, preliminary model with the following components:

  • A simplified social network of swidden farmers exchanging agricultural labor, inspired by Downey (2010)
  • A landscape in the form of a 2-D grid in which cells are patches of forest / crops

Possible model outputs include:

  • Yearly harvest
  • Biodiversity
  • Yearly biomass production in forest and fallows
  • A representative indicator of the "net production" of the complete system
  • Sustainability of the complete system

Biodiversity, biomass production, and sustainability will need reasonable, simple definitions that follow modern literature on the subject.

Software and language

Every idea is worth discussing. NetLogo might be an option that fits with our limited calendar. Git and GitLab would be used to manage the source code.

Group contact

Fabio Correa <facorread@gmail.com>

Interested in swidden agriculture, please sign up below

  • Lindsay Todman (Lindsay.todman@rothamsted.ac.uk)
  • Chris Revell (cr395@cam.ac.uk)
  • Christine Moore (christine.moore@ouce.ox.ac.uk)
  • Evelyn Strombom (evelyn.strombom@gmail.com)
  • Santiago Guisasola (sguisaso@uci.edu)

Modeling within-city residential migration (Updated 6/21/2016)

Keywords

residential mobility, neighborhood change, network analysis

Summary

Title: Measuring Within-City Residential Migration Flows and Exploring Attractors and Barriers to Mobility and its Consequences

Objectives:

  • Measure within-city residential migration
    • Characterize the mobility network
    • Quantify flows of mobility across neighborhoods
  • Explore attractors and barriers to residential mobility
  • Quantify changes over time in the mobility network
  • Explore the consequences of neighborhood mobility
    • Impact of increased residential mobility throughput on neighborhood disorder.
    • Predictive model of housing prices


Study setting City of Madrid (Spain), from 2004 to 2015
Note: Madrid is divided into ~2400 census sections (~1500 people in each) nested into 128 neighborhoods (~20-30,000 people in each).

Data sources:

  • Spanish Continuous Administrative Census (Padron)
  • Mobility data: in 1 year windows (from 2004-2014) how many people have moved from each census section to every other section, by age and level of education
  • Sociodemographic data: cross-sectional cuts every year (from 2005 to 2015), including data on education, age and country of origin, for all census sections

Other sources

  • Property value: from 2001 to 2015, by neighborhood; cross-sectional cut in 2016 with all houses under sale with their characteristics.
  • Tax registry: geocoded housing units with data on year of construction and size (n~1.6 million housing units). Complete history from 2001 to 2016
  • Commercial spaces: geocoded list of all commercial spaces (and open/closed and licensed use) from 2012 to 2016.
  • Traffic: every 15 minutes, from 2013 to 2016, all traffic lights
  • Traffic fines: date/time + geocoded address, type of fine (2014-2016)
  • Complaints: date/time + geocoded address, type of complaint (2013-2015)
  • Other: unemployment, labor data, election results, noise, pollution, gdp, vehicles..


Working Packages:

  • Management (overarching task)
    • Data cleaning and management, codebook management.
    • Code, data interfaces
  • Mobility as networks (1st stage):
    • Quantification of Mobility (flows)
    • attractors/barriers models
  • Consequences of Mobility (2nd stage)
    • Measuring disorder
    • Predicting property value

Group Contact

Usama Bilal <ubilal@jhmi.edu> Slack #citymobility

Group Members (as of 6/21/2016)

  • Usama Bilal
  • Scott Armstrong
  • Jesús Arroyo
  • Michael Schaub
  • Andrew Mellor
  • Anjali Tarun
  • Ellen Badgley
  • Catriona Sissons

Complexity insights into Circular Economy

Specific wiki subpage here


Keywords

Complexity economics, Network optimization, Urban systems, Agent-based modeling

Summary

The 'circular economy' is recent approach on questions of sustainability. The current 'take, make, waste' way of producing is considered to be not sustainable anymore. Current thoughts on 'closing the loop' mechanisms on producing and consuming are booming in many different academic fields. However, not yet from an integrated point of view. The idea of 'closing the loop' mechanisms corresponds to complex system paradigms, however complexity science has hardly been utilized in this field. In this project we want to start brainstorming on how complexity science can contribute to the field of circular economy. We think there are many ways in which complexity science can contribute, both in methods and in theory. Based on the competences in the group we would like to brainstorm on suitable research questions, that can range from very practical to theoretical issues.


Insights

Please contribute here by adding your insights

  • (Juste) From the point of view of theoretical geography, I think this question directly enters the frame of the understanding of human settlements. Distribution of population and ressources in space and time is a fundamental question for which many theories have been proposed (Evolutive Urban theory, Scaling theories, Human territoriality, etc. among others) and that many disciplines have studied (economics, geography, planning, archeology, more recently physics). To my point of view this project has a strong component linked to this issue, as it tackles the question of sustainable territorial organizations and the possibility of the existence of closed subsystems. As a consequence, the study of circular economic systems should necessarily consider them as being, among others (1) multi-scalar, (2) non-ergodic and (3) non stationary [beside buzz-words : 1) urban systems exhibit by nature multi-scalarity, and thus a circular organization of ressources management would also be as it is necessarily embedded into an urban structure ; 2) path-dependancy and geographical context in urban systems generally imply that universality is difficult to establish and makes direct comparison hazardous : urban systems all over the world do not correspond to realizations of the same process at different times (see Pumain, 2012) ; 3) urban systems as economic systems in general are never at equilibrium and the nature and positions of attractors changes in times, i.e. parameters of the dynamic do not have a stationary distribution]. For these reasons, complexity approaches are a natural way to study them, and build associated models that include these properties.
  • (Lorraine) Also from a cities perspective: taking the city as a non-equilibrium system, one could say the city forms a dissipative structure (Bristow and Kennedy, 2015) - this is also the case for ecosystems (Schneider and Kay, 1994). Dissipative structures maximize destruction of energy gradients (i.e., as the structure becomes more ordered/grows, more energy flows through). This has direct implications for increasing resource consumption in growing cities. Furthermore, scaling relationships for cities show that consumption of resources scales superlinearly with population (Bettencourt, 2013; Bettencourt et al, 2007), which implies that as cities grow, more and more energy is consumed. It would be interesting to explore the concept of a circular economy in such a system, i.e., can a "circular economy" policy mechanism actually change the organic dissipative/scaling nature of cities? Further, since ecosystems are also dissipative structures, are there examples of "circular economies" in ecology that can serve as a model for cities? Just some ideas, would love to discuss more.
  • (Ben) I have worked on the topic of cleaning data sets for sustainable industrial production systems. Some of the work have been documented in http://enipedia.tudelft.nl/wiki/Industrial_Symbiosis_Data_Sources. I can also access to Europe Waste/pollution registry (stored as RDF format). It covers major industrial polluters in Europe.

Organization

Timetable

As many meetings are scheduled, we will meet both tonight at 7pm (Wednesday) and tomorrow at 9pm (after physics lab) at the coffee shop. People interested can join one or both. A summary of the first meeting will be updated here.

Repository

Git repository for the project : [2] . please email project contacts to obtain full write access (or do PR directly).

Group contact

Joris Broere j.j.broere@uu.nl

Juste Raimbault juste.raimbault@polytechnique.edu

People interested please sign up below

Insights in Scientometrics

Summary

I have multiple scientometric datasets scraped from the "Web of Science" website, which include published works from different (hard sciences only, sorry) domains. Author(s), Year, Journal, ..., and citations are available. I have some code ready to pre-process those data, e.g. for linking articles by co-citations, or bibliographic coupling. I've been working on these data for some time, and am eager to offer the data in exchange for fresh, brilliant ideas.

Group Contact

Matteo Morini

Interested people please sign up below

Xiongrui (John) Xu (xuxr007@gmail.com)
Thomas Zhang (tomzang@umich.edu)

Using DOTA 2 video game data as a proxy to improve cyber intrusion detection algorithms

Keywords

Adaptive strategies, co-evolutionary systems, cybersecurity, DOTA 2, Inductive Game Theory, time-series analysis, potential NLP, potential genetic algorithms, potential network analysis

Key Points

Main question: how to a) detect and b) quantify the rate of change of strategies in co-evolutionary systems?

Motivation: improve cyber intrusion detection algorithms

Method: use the competitive, strategic, multiplayer online game DOTA 2 data as a proxy for a cyber attack. Extract sets of strategies from game observables and develop an algorithm to detect when strategies change. This could be from within the games or from the drafting of characters (heroes).

Summary

Most metrics for scoring how well a given cyber intrusion algorithm performed involve static determinations after an attack has occurred and usually consist of simple false positive or false negative type scores. This does not enable the granularity necessary to improve these algorithms in appreciable ways. The goal is to develop new metrics to assess in real time when and how attackers adapt to the protective mechanisms they encounter so that the security teams can adjust their own defensive tactics more quickly. This could decrease the long lead time for patches and stave off additional damage.

Usable data from within the cyber domain that may help to strengthen detection algorithm scoring is nearly nonexistent. If actual attacks are documented in the real world, this data is rarely made available and may be network specific (hence, not generalizable). Thus, a proxy for the data is necessary. An online battle arena video game called Defense of the Ancients 2 (Dota 2) is a rich data source of real-time adaptive adversaries. Dota 2 is a strategic, competitive, multiplayer game where two teams of five individuals each compete against each other to complete objectives and to destroy the other team’s base in a time frame of ~20 to ~90 minutes. The players deploy various in-game and between-game tactics and procedures to achieve a specific measurable objective. Professional players vie for tens of millions of dollars in prize pools each year, and over 2 billion games have been played. Currently, we have access to a Mongo database with 500 GB of data (~2.5 million games played over one year).

This is akin to the classic ‘Red Queen hypothesis’ in evolutionary biology, but in this case we are interested in human behaviors where a strategy can be considered most generally as a ‘meme’ of sorts. Our hypotheses include: 1) a mapping exists between game observables and a set of strategies, 2) a measurable signal can be extracted from which to ascertain the adoption, stabilization, and decay of specific strategy traits, 3) changes in strategies occur over time, and 4) strategy changes are driven (at least in part) from behaviors of the opposing team. We postulate that testing these hypotheses will require an understanding of the co-evolutionary dynamics of the overall environment. In particular, the human behavioral components underlying the adversary/defense team actions will need to be assessed. The use of data from a multiplayer online game for this purpose assumes there is a valid mapping between the ‘game-space’ to ‘cyber-space’ behaviors from which useful inferences can be made. Through this exercise we hope to understand what types of data (if any) are useful for this purpose and how one might develop proxies to characterize strategies. Ultimately, we hope that the analysis could be applied against realistic data specific to a cyber intrusion.

See DOTA 2 DATA OVERVIEW for additional details.


Technology

Parsed data is in JSON format. We should have access to two databases: one is a mongo database discussed above for aggregated information (e.g., fights, deaths (characters respawn), characters played, items acquired, positions for the first 10 minutes, league IDs, player IDs, wins), and the second is raw data for every move and every action each player took throughout the entire game.

Interesting addition and comments to the idea

Please add your comments and idea about the project here.

Group Contact

Anne Sallaska, alsallaska@gmail.com (or asallaska@mitre.org, harder to check so slower to respond potentially). Slack channel has been set up! If you need an invite to Slack, let me know. That will be the easiest way to organize and communicate.

Interested participants, please sign below

  • Ellen Badgley (flyingrat42@gmail.com) - I'm happy to provide support for the Java parser
  • Will Hamilton (wleif@stanford.edu)
  • Marla Stuart (marlastuart@berkeley.edu)
  • Salva Duran (salvador.duran at upf.edu)
  • Daniel Biro (daniel.biro@med.einstein.yu.edu)

Complexity in fruit fly learning behavior

Keywords

Fruit fly, learning, behavior, phenotypic space

Summary

I designed and built a high-throughput apparatus (currently in the process of patent application), in which I use laser to deliver heat stress to a single fly. When the fly moves, the laser turns on, when the fly stops, the laser shuts down. In order to avoid the heat, the fly has to learn not to move. Some flies have difficulty in making correct decisions under this stress. Different genetic mutations related to dopamine/serotonin systems may have contributed to the incorrect decisions made.

In the following figures, the first one is a fly’s movement trajectory before training; the second one is the same fly’s movement after one training (usually one fly will experience two training sessions or more).

Before training

Beforetraining.jpg

After one training session

Aftertraining.jpg

Objectives

The goal of this project is two-fold: 1) Quantify the complexity hidden in the learning curve of individual flies, and characterize variations from fly to fly. 2) Create a mathematical model for the fly’s learning process, and verify it in different animals, as well as to understand how mutation in the genome affect learning behavior from the perspective of parametric analysis in mathematical models.

My time-series trajectory measurement dataset consists of approximately 700 individual flies, including wide-type and mutant flies. It is a rich dataset ready to be analyzed.

Group Contact

Ruichen Sun (ruichensun0 at gmail dot com, or r5sun at ucsd dot edu)

Interested participants, please sign up below

Cheng Jin (chengkim@zju.edu.cn)
Chris Revell (cr395@cam.ac.uk)
Kelly Finn (kfinn@ucdavis.edu)

An Interdisciplinary Approach to Morphogenesis

Presentation

This is not properly speaking a project proposition as the idea is quite ambitious and the result may be more at a theoretical level than concrete and practical developments. If some are really interested but are already busy with other projects, I think this one can be done without heavy time consumption if it stays at a theoretical level.

The concept of morphogenesis is a powerful approach to many kind of complex systems. It aims at unveiling the processes responsible for the formation of patterns, and link the emergence of these patterns to functional properties of the system. This definition would sound wrong to anyone who has been involved with morphogenesis, as each field has its own definition and use of the concept. Proposing an interdisciplinary approach to this concept has already be done (see e.g. Bourgine & Lesne 2010, available here ) but to my point of view its potential richness has not been exhausted. The idea would be to gather insights from various people having worked on morphogenesis and to study the potentialities of a self-consistent theoretical framework that would include the different views, and be autonomous as an approach to complex systems. The work would be both on an epistemological and theoretical level, but also potentially practical applications (e.g. develop an existing model someone has worked on, given the insights of other domains and of the theory).

Group contact

Juste Raimbault (juste.raimbault@polytechnique.edu)

Interested, sign up below

add your field and if you worked on morphogenesis, which particular approach

  • Juste Raimbault [worked on various models of urban morphogenesis, and currently trying to develop a theory of co-evolutive territorial systems which one of the core components is morphogenesis]
  • Lars Hubatsch [I work in C. elegans which is a classical biological system for morphogenesis]
  • Jesus Mario Serna (familiar with geomorphogenesis and the possible relations to psychological structures)
  • Emanuele Crosato (I studied Artificial Life and I've been investigating Artificial Self-assembly, and also produced a literature review on it)
  • Moriah Echlin (I have some experience with pattern formation in networks and experience in theoretical biology/modeling)
  • Chris Revell (cr395@cam.ac.uk - I work on computational modelling of self-organisation in the mammalian embryo)
  • Aina Ollé Vila (aina.olle@upf.edu)-I study tissue organization and cell differentiation
  • Chenling Xu (chenlingantelope@gmail.com) - I have not studied morphogenesis but I am interested in learning about a species of frog that has a few different skin patterns (spotted, horizontal stripe and vertical stripe) for imitating a poisonous frog. I don't know the genetic basis of how the patterns are produced but I think if we can figure out how to reproduce the natural patterns by varying parameters in a diffusion system and predict what the minimum number of genetic changes is required for producing all the variety we observe in the nature, and how much of the variation is just caused by noise.
  • Daniel Biro (daniel.biro@med.einstein.yu.edu) (I work on computational models of evolution that incorporate aspects of development, particularly robustness)
  • Lin-Qing (lchen@pitp.ca)(I study physics, I haven't worked on morphogenesis yet, but I'm very interested in different mechanism of structure formation.)

Patterns in Globalization

Keywords

patterns, globalization, trade, fractals, trends, British empire, Trieste

Summary

What if can we see how is the face of globalization? Globalization is a phenomenon lasting centuries. I believe we can begin to look better at its face from the data of trade.

Thanks to the scaling power of fractals, we can look for patterns into a small, well defined environments, such as the free ports.

I have the data regarding the flow of goods in the free port of Trieste in the 19th century and specifically those regarding Trieste and the British Empire. Trieste was the main port of the Habsburg empire, one of the 10 most important port in the world at the time. In the 19th century we had a kind of globalization that was centred on the British financial system, so it is important to see the grade of integration with the British world System of a single free port out of the British empire.

I have data regarding:

vessels: tons and value of goods, flag, provenance and destination

goods: amount in tons and value of import-export for each typology, general data regarding trade in Trieste

In addition I have the data regarding the Generali Insurance company. Based in Trieste, the company was and is one of most important insurance company in the world. I have all its balance sheets from 1830s to World War 1. We can look at insurance on vessels and the goods they transported as a proxy for the general development of trade in Trieste and to evaluate trade risk over the time.

We can begin concentrating on 3 goods, for example, and to see what kind of networks of countries we find out, what trends they developed. In general we can choose from the huge data set I have what can be more interesting to our purpose. What models we can build, what math we can use and what tools fit better to find the patterns, which ultimately will led us to begin to see on the horizon the face of globalization.

Interesting addition and comments to the idea

Please add your comments and idea about the project here.

  • (Lorraine) It would also be interesting to explore how trade impacts the growth of the city itself. Materials entering and exiting, how materials are stored, how the city is built, etc...
  • (Gaetano) very well, I have data about it too, for example this on number of houses (case) and population (abitanti):

Selection 097.jpg

Group Contact

Gaetano Dato, gdato7@gmail.com

Interested Participants

  • Ben Zhu (b.zhu@tudelft.nl)
  • Usama Bilal <ubilal@jhmi.edu> This is a topic that really interests me. I'm mostly focused ("back home") in short-term changes but I do believe in the enormous (actually, much more important!) importance of long-term trends
  • Lu Liu A paper on product space and golbal economy which might be relavent to the project : Hidalgo, César A., et al. "The product space conditions the development of nations." Science 317.5837 (2007): 482-487. (Hi Lu, I've encountered this paper as well, and I've shown it to Gaetano. They also have interesting visualizations of the product space for each country. More to see here: http://atlas.media.mit.edu/en/ - Anjali )

Catriona Sissons (c.sissons@auckland.ac.nz)

a small gift to everybody: a link on a simulation showing a single mass on a spring

For those who likes to play with this kind of things: http://www.myphysicslab.com/spring1.html

Consensus, conflict and coalitions: exploring agent interactions within health policy systems

Keywords

Decision making, Health policy, Network influence

Summary

Decision makers in health choose which policies to support from a range of options. These decisions are theorised to be based on; what they see is the best option for society (i.e. public health), what they see is the best option for themselves (financially or politically), what they are told is the best option by other decision makers, what they think based on previous experience, or a combination of various processes.

Health policy outcomes are determined collectively via interactions within policy systems. Policy actors (e.g. government departments, charities, businesses, media etc.) work to influence each other and final policy decisions. Policies with high levels of support are more likely to be implemented, while disagreement between actors may stop one decision and lead to another less beneficial public health outcome.

In cases of conflict, rival policy advocacy coalitions (network components) can form. Coalition structures may affect the flow of information between decision makers for future decisions, even when levels of conflict may be lower. This could in turn influence how well the policy process benefits public health, or benefits certain actors.

This project aims to explore how the characteristics of actors within policy systems can lead to different policy outcomes and system configurations. We aim to construct health policy networks based on existing research and policy theory, develop models to explain the evolution of policy networks over time, assess how our model outcomes compare to outcomes in real policy systems, and determine how changing parameters of the model could change the outcomes of the system.


The project will involve the following strands:

1. Identifying policy system structure; comparing UK and US policy systems (Based on political theory, health policy literature, or other suggestions from collaborators)
2. Determining the ways in which actors take decisions (Based on data envelopment analysis, or other suggestions)
3. Constructing a model to show changes in policy systems over time (Boolean Network, other suggestions)
4. Determining are there certain characteristics of policy systems, policy actors, or policy options that lead to certain outcomes

The findings of the models could help us better understand conflict and coalitions in policy systems, variation in the use of evidence, and the emergence of communication breakdowns in policy enivronments.

Group Contact

Mark McCann Mark.McCann@Glasgow.ac.uk

Interested participants, please sign up below

  • Jesus Mario Serna
  • Rana Azghandi
  • Marla Stuart

Complexity in Material Stocked in Japan

Keywords

cities, industrial ecology, materials, GIS, fractals, scaling

Summary

Fishman et al. (2014) have developed a database of the mass of materials stocked in buildings and infrastructure for all of Japan, at a scale of 1km^2 and 10km^2. Looking at a map of the mass of stocked materials (see figures from Tanikawa et al. 2015), there are clear links to complexity (i.e., scaling and fractals). The overall objective of the project is to explore the complexity inherent in materials stocked in Japan. This could allow us to make links to physical infrastructure growth and the mass of cities.

We currently have GIS data for 2009 (buildings) and 2010 (infrastructure). We have reached out to the authors of the study to see about gaining access to the full time series. Below are the research questions for each option:

Analysis with 2009/2010 data:

  • Is there a scaling nature of stocked materials in Japanese cities? Is there scaling with respect to population, city rank, and other indicators? Does the spatial definition of "city" impact scaling?
  • What is the fractal nature of stocked material in Japanese cities?

Analysis with time series: (could brainstorm potential approaches with or without data)

  • Apply non-linear techniques to an econometric analysis of socio-economic analysis of material stock accumulation
  • Investigate the fractal nature of material accumulation over time
  • Track physical growth of specific cities after a catastrophic event (i.e., severe bombing in cities during World War 2)

Group Contact

Lorraine Sugar (lorraine.sugar@gmail.com)

Team Members (as of 06/21/2016)

  • Ellen Badgley
  • Devrim Ikizler
  • Lu Liu
  • Andrew Christian
  • Ben Zhu
  • Xander Brehm

Complexity and P = NP

Summary

It seems intuitive that if I were to A) give you the output (and only the output) of the logistic map with paramater 4 and initial condition 0.4, you would have a hard time recovering the function and parameter set, and B) if I gave you the logistic map and parameter values, you could quickly verify that they produced the output. This project would be about studying the P=NP problem and the relevant information theory and chaos theory.

Admittedly this is something I know little to nothing about, and whole fields of time series analysis are devoted to recovering functions and their parameters: establishing that there exists a function that produces output that is not computably invertible but is easily verified would make a big statement on the P = NP idea. The activities of this group would be reading and understanding information theory and chaos theory and working towards a paper that provides some insight on P=NP and relevant complexity theory.

JE: I think symbolic regression could actually recover the return map rather quickly, although it would certainly be difficult (impossible) to confirm the uniqueness of this map. I'm very curious about this topic, but it could be as simple as a conversation with Jim Crutchfield sometime.

Group Contact

Xander Brehm (alexander.brehm@linacre.ox.ac.uk)

Interested participants, please sign up below

  • Jeffrey Emenheiser (jemenheiser@ucdavis.edu), but I have too many projects already.

Markus Junginger

Complexity in speech patterns

Summary

Have you ever:
- Been bored by listening to a monotone lecturer?
- Been inspired by a dynamic motivational speaker?
- Counted the number of “umms” in a conference talk?
- Thought about how lovely it is to listen to Morgan Freeman?

And more importantly, are you curious about what properties in speech patterns elicit these reactions?

There are many properties in speech time series (fluctuations in tone, pattern of sound/silence, word pattern, etc. etc.). I’m interested in investigating patterns in these different properties, specifically how various complexity measures of speech relate to listener perception.

Possible data: online lectures, political candidate debates or speeches (could be fun, especially given current politics)
Possible measures of perception: “likes” or internet popularity, or a short study having CSSS participants rate short audio clips of people talking.

  • could also play around with speech simulations


Other ideas welcome!

  • Was chatting with MITRE folks over dinner tonight and it was mentioned an experiment where people had to decide whether or not to exchange phone numbers based on listening to an audio clip with the words muffled out - only the speech pattern was perceptible. I can try to dig up the reference if anybody is interested. - Ellen

Next Meeting

Friday at SFI after lunch (~ 1:30).

Some materials

Zootopia Sloth speech scene

Group Contact:

Kelly Finn (kfinn@ucdavis.edu)

Interested Participants:

  • Andrew Christian (CommonTime@gmail.com)
  • Ruichen Sun (ruichensun0 at gmail.com)
  • Rudi Minxha
  • Pavel Senin - I worked on a MFCC voice decomposition pipeline before, we can do some quantitative analytics using that, like counting umms automatically, etc..

Tournaments of the World

Summary

Some of you may be aware of the (important) things that are going on in the world today: the NBA finals, Euro 2016, the 100th Copa America, the NHL finals, et al. A question that I am interested in is trying to deduce what the probability was that the winner of the tournament was going to win based on the data generated by the tournament (the results of all of the matches).

My ansatz is that this will involve some Bayesian analysis to create models of the various teams, then a Monte Carlo phase of simulating tournaments based on these models. Once this MC dataset is created, one can ask any number of fascinating questions (e.g., What is the predictive power of the tournament? How sensitive is it to the seeding? What is a better tournament algorithm? And, of course, how unlikely is it for the US to win a major soccer tournament?)

I feel like I need to work some more [complexity] into this idea. Suggestions are welcome. Perhaps there is something to be said regarding the information generated by sporting events (as they are not simply random events as insinuated above).

Note: This is not just about sports, but these tournaments provide a timely example of data that is akin to the human subject data that I often encounter in my work (so don't feel bad about joining in if sports aren't your thing).

Group Contact

Andrew Christian (CommonTime@gmail.com)

Interested Participants

  • Ryan McGee
  • Ellen Badgley - Not joining but I would suggest looking at the DOTA 2 data/description linked below, as this is an incredibly competitive e-sport and we have tons of tournament data for it.

Meeting Information

First meeting will be during the America game and the LeBron James game (7 pm Thursday the 16th) at the BoxCar in downtown Santa Fe.

Removing bias from predictive algorithms

Keywords

Machine learning, Justice policy

Summary

In many states across the US predictive algorithms are used in an effort to support judges in more objective decision making. These so called risk assessment algorithms are used to predict whether or not a person is likely to appear to court hearings, whether they are likely to repeat their criminal behavior after being released etc.. A recent report by ProPublica shows that one of these algorithms, which is in use in different states, is heavily biased against Afro-Americans.

We are wondering whether there is a systematic way of removing such a bias, either by modifying the algorithm after training, or by systematically varying the set of criteria which is used to inform the algorithm in order to remove the bias without negatively affecting its predictive power.

A dataset of 7000 convicts and their criminal history after release from prison is available [online].

More generally, this could be used to avoid identifiability of specific groups in situations where such algorithms are used. This could apply for example to online shopping, in order to avoid discrimination.

Group Contact

Lars Hubatsch (lars.hubatsch@crick.ac.uk)

Interested Participants

Human Group Dynamics; Coupling a Thematic and Acoustic Analysis on an Interdisciplinary Operative Group Model

Keywords

Thematic analysis, Group Dynamics, Acoustics, Interdisciplinarity, Empirical Research, Psychoanalysis

Abstract

We do an acoustic analysis on human group dynamics, bridging qualitative and quantitative data. The data comes from recordings on an interdisciplinary discussion group. We then analyze the thematics and group dynamics and couple it with the acoustics from the waveform.

The most sensible question is that, even though in psychology there are working models to think about group dynamics, they frequently they frequently lack quantitative groundings (mostly due to the difficulty of quantifying extremely complex information, so most of the researchers either go radical on behaviorist oversimplification or just drop quantification altogether), hence an acoustic analysis might help bridge this verification gap.

Operative Group theory is based on the dialectical view of the functioning of the groups and the relationship between dialectic, homeostasis and cybernetic. Thus, Nonlinear dynamical systems give us huge metaphorical insight soicopsychological models, network theory and agent based modelling gives us tools to further ground these notions, but what links could we actually discover from an acoustic analysis of human group processes? That is question for which we have yet to find more answers.

Following our first rough data analysis, we could center around the thematic nodes in the discourse, the edges they have with the agents, and look at fluctuations in the wave. For example, the relation in voice frequency synchronization between agents: if they agree on the subject and add on it, is there a link in voice synchronization? On the contrary, we might speculate that if the agent is linked to the thematic node, but contradicting it in the content, the degree of synchronization might be different.

Further connections on silence and outburst of laughter might give us indications of importance of a thematic node in time, as well as the degree of it's dynamics. Finally, another fruitful analysis might prove to be introducing a general entropy factor. This example might show up in a session with very low energy and less interaction between thematic nodes and agents, vs. one that's more dynamic, mixed, and fluid, etc.


Method

We have 4 recorded sessions with a group of 8 people. After the qualitative aspects of the data are analyzed (thematic and group dynamics) we will pair it with the quantitative aspects from the acoustic analysis of the waveforms.

The models at work come from acoustics and the "Operative Group" (Pichon-Riviere, inspired from Psychoanalysis) enlarged with a non-linear dynamical systems perspective. Next we search for possible links from the qualitative and quantitative information. From an NDS perspective, can think the operative group functioning as a magnetic model: agents interacting, at first individually and "coldly", then mixing up as the system "heats". In this model the energy is what pushes the language interaction and flow between the agents, and as the group dynamic takes hold and the energy increases, the interaction and ideas speed up and intensify .

Following out first data analysis our general approach will be centered around finding the thematical nodes in the discourse and the edges they have, see if they are related to voice frequency synchronization between agents (for example, if they agree on the subject and add on it ad we find a link in synchronization) or if they are not synchronized ( which might happen if the agent is linked to the thematic node, but contradicting it in the content). Further connection on silence and outburst of laughter might give us indications of importance. Finally, another fruitful analysis might prove to be an entropy factor (for example in a session with very low energy and less interaction between themathic nodes and agents, vs. one that's more dynamic, mixed, and fluid)

Summary for participants on the Interdisciplinary Discussion Group

How we conceive and relate to our models, its pros and cons and, what can we learn from our different fields? What steps are needed for implementing awesome mathematical models into human reality so societies can actually integrate them? Hence, what are some of the gaps and challenges that we often encounter in the transmission of scientific knowledge into political, social and cultural systems? What is the model for consciousness in your Complex Systems Field? Is it rational agents? Self emerging systems? Politics and health systems? Relation of bees and flowers? Human interactions in networks? Frankenstein supercomputers taking over the world? Sentient structures? Maybe no conscience at all! With this interdisciplinary soup we can see what emerges in our interactions.

You can think of it as a little oasis of reflection in the midst of the project frenzy…with a side project! So it’s taking a breath to reflect on what we are actually doing. And moreover, what does the application of our models imply? If what we do is useful, what use could that be, and how can we efficiently communicate that to people outside our fields (very often in the decision making processes or the receiving end). In that sense, it is more like a workshop, with the final point integrating some of our exchanges. If the qualitative aspects for this interaction are non-predictable, we can hopefully arrive to some general conclusions and acquiring some meaningful insight in our fields...or bring up some interesting new questions!

So no brain frying needed into the wee hours of the night, it’s just taking some (restricted) time in 4 sessions to reflect on what our work implies in an interdisciplinary environment in an open and convivial setting.

Concretely, what you need to do is: Show up in time to the 4 sessions and exchange in dialogue. We present our models and how do we relate to them, why do we think they are important or working, or lacking, and then see if we can learn something from other models and transdisciplinary applications.

There will be a second phase to analyze the information gathered. It’s not mandatory to participate (for example due to time constraints from your main projects) but the people willing to do it can chip in with their skills to find a way to apply some of our models to our own process (in a way, like a fractal intake of our own medicine) and see what we can get out of this experience. For example, beyond the analysis on group dynamics, ideas and concepts sprouting, can we spot patterns from the waveform of the audio recorded from the sessions? Are there peaks correlated to certain ideas or sensible points? Would a time series analysis of this give out something of interest? As the group is established, do we see a form showing up, some bifurcation or change of state that is different form the end in regards to the beginning?

Group Contact

Jesus Mario Serna (jesus.mario.sv@gmail.com)

Gaetano Dato (gdato7@gmail.com)

Participants

Meeting with Jelena Monday at 17h15

3rd Group Meeting Session: Wednesday 29th , 19h @ Chamisa

4th Group Meeting Session: Friday 1st , 16h @ SFI Pad

Dynamic Information Routing on Complex Networks

Keywords

information flow, network dynamics, oscillator networks, attractors, characterizing influence of network components

Summary

Flexible information routing fundamentally underlies the function of many biological and artificial networks. Yet, how such systems may specifically communicate and dynamically route information is not well understood. Kirst et. al (2016) present a mechanism that routes information on top of collective dynamical reference states in complex networks. Switching between collective dynamics induces flexible reorganization of information sharing and routing patterns, as quantified by delayed mutual information and transfer entropy measures between activities of a network’s units.

For this project, we are interested in characterizing nodes and network topologies that are especially influential on the routing of information flows on the overall network. Toward this end, we hope to identify attractors for this sort of high-dimensional complex network and, in so doing, identify the parameters that are most important in shaping information flows on the network. This work may also include re-exploration of metrics of directional information flow, network evolution, and/or application to neuroscience data.

Group Contact

Hamza Giaffar (hamzagiaffar@gmail.com)

Participants

Jeffrey Emenheiser (jemenheiser@ucdavis.edu)
Ryan McGee (ryansmcgee@gmail.com)
Andrew Christian (CommonTime@gmail.com)
Emanuele Crosato (crosato.emanuele@gmail.com) - I am working on collective moves of a school of fish, investigating lagged mutual information and transfer entropy
Sina Tafazoli (tafazoli@princeton.edu)
Lin-Qing Chen (lchen@pitp.ca)
Pinar Ozisik (pinar@cs.umass.edu)

Meetings

First thing (~9am) at SFI on Friday, June 17, we will discuss "dynamic information routing:" http://www.nature.com/ncomms/2016/160412/ncomms11061/pdf/ncomms11061.pdf

Expanding data sets via Chaotic Warping

Keywords

deep learning, time series classification, discretization

Summary

1.0. The goal.
Inspired by recent successes of deep learning in classification, we attempt to address a fundamental to the field problem of insufficient training data instances by developing a Chaos-based technique for data warping. Specifically, our goal is the development of a novel, deep learning-based solution for time series classification (TSC) problem based on Chaotic Warping. As the standard benchmark dataset (i.e [| the UCR dataset]) typically has very few instances in training data, which significantly affects training and cross-validation processes performance, we believe that our technique shall contribute to the whole community.

2.0. The approach taken.
Probably the best explanation of the problem and the approach, a different technique though, is given by Patrice Y. Simard, Dave Stdeinkraus, and John C. Platt in their work titled "Best Practices for Convolutional Neural Networks Applied to Visual Document Analysis" [1] ...
"...Synthesizing plausible transformations of data is simple, but the “inverse” problem – transformation invariance – can be arbitrarily complicated. Fortunately, learning algorithms are very good at learning inverse problems. Given a classification task, one may apply transformations to generate additional data and let the learning algorithm infer the transformation invariance. This invariance is embedded in the parameters, so it is in some sense free, since the computation at recognition time is unchanged. If the data is scarce and if the distribution to be learned has transformation-invariance properties, generating additional data using transformations may even improve performance. In the case of handwriting recognition, we postulate that the distribution has some invariance with respect to not only affine transformations, but also elastic deformations corresponding to uncontrolled oscillations of the hand muscles, dampened by inertia..."

At the first prototyping iteration, for simplicity of the feasibility evaluation, we plan to re-use Diana Dabby's approach for variation (warping) of musical pitch sequences: "..The sensitive dependence property of chaotic trajectories offers a natural mechanism for variability. By affixing the pitch sequence of a musical work to a reference chaotic trajectory, it is possible to generate meaningful variations via a mapping between neighboring chaotic trajectories and the reference. The variations result from changes in the ordering of the pitch sequence. But two chaotic orbits started at nearly the same initial point in state space soon become uncorrelated. To counter this, the mapping was designed so that a nearby trajectory could often track the reference, thus tempering the extent of the separation. Tracking means that pitches in the variation appear exactly where they did in the source. However, regardless of whether the two trajectories track, the mapping links the variation with the original by ensuring only those pitch events found in the source piece comprise the variation..." For the next iteration we plan to re-use the Liz Bradley's technique or some other approach (if found)...

Essentially, we would expect the Complexity warping to work somewhat similar to the displacement field transform which is illustrated on the figures below:

ChaoticWarping01.png

However, our approach will differ as we are going to build it upon the symbolic discretization and time series bitmaps illustrated on the figures below:

2.1. Symbolic Aggregate approXimation.
SAX transform the input time series into a string:

Sax.png

As shown, the input time series processed with three steps: (i) it is normalized to the mean of zero and a unit of standard deviation (z-normalization), (ii) Piecewise Aggregate Approximation (PAA) applied to the normalized time series, (iii) PAA coefficients are mapped to a Symbolic Aggregate approXimation (SAX) alphabet of size 3.

This SAX process is applied to the input time series via sliding window to transform it into a set of representative strings -- i.e. a window of size W slid across the time series and the subsequences extracted via the window are SAX-transformed with the PPA size P and the Alphabet size A.

2.2. Time series bitmap.
Bitmap01.png

Bitmap02.png

2.3. Neural net-based TSC.

Neural net01.png

Some References

Group Contact

Pavel Senin (seninp@gmail.com)
Nicole M Beckage
Pinar Ozisik

Interested Participants

Migration Behaviors Modeling and Analysis for Ocean Animals

Keywords

migration, behavior modeling

Summary

Given the movement track a migratory individual/population from location A (e.g. breeding season) to location B (e.g. non-breeding season), are they following the shortest path? If not, how random is their trajectory? And why are they bothering not using the shortest path? Answering this would maybe require analysing the point data using various curve fitting methods and comparing to null models like a random walk for example.

Some References

  • Tracking marine apex predator movements in a dynamic ocean, 2011.
  • Datasets:

Dolphins dataset


Two Whales

Group Contact

Ulya Bayram

Interested Participants

Modeling the co-evolutionary dynamics of the Lobaria pulmonaria lichen symbiosis

Keywords

lichen symbiosis, co-evolutionary dynamics, diversification, speciation, Lobaria pulmonaria

Summary

The mode of reproduction strongly influences the genetic structure of a lichen population [1,2]. Due to the lack of genetic markers with high-marker resolution, few studies have characterized the within population genetic structure of lichen species (i.e. mycobiont and photobionts genotypes). Because Lobaria pulmonaria is one of the best-studied lichen species in the world, it is a good candidate for testing hypotheses related to the genetic structure of L. pulmonaria populations. L. pulmonaria is a tripartite lichen species consisting of an association among the green alga S. reticulata, the cyanobacterium Nostoc sp., and its main fungal host in Europe, Lobaria [3, 4]. Recently, eight fungus-specific [5, 6, 7] and seven alga-specific [8] microsatellite markers have been developed for L. pulmonaria, allowing for reliable identification of genetically distinct individuals in highly clonal populations. Microsatellite markers have not been developed for the cyanobacterium [9]. To study the population structure of L. pulmonaria, the fungal and algal symbionts of 1960 L. pulmonaria thalli from 62 populations in forests throughout Europe, parts of North America, Asia, and Africa were genotyped at eight and seven microsatellite loci [10]. The L. pulmonaria-S. reticulata symbiosis showed significant within-population genetic structure due to restricted gene flow and vertical transmission (i.e. co-dispersal of vegetative propagules) with identical genotypes found in 77% of fungal and 70% of algal pairs [11]. L. pulmonaria was dominated by micro-evolutionary processes with high somatic mutation in the alga (30%) and the fungus (15%), while recombination contributed little to both the algal photobiont (no statistical evidence) and the fungal mycobiont (7.7%) [12]. In the L. pulmonaria system, new genotypes continuously arise via mutation and selection [3]. If the nature of the fungal-algal partnership evolves over time, the lichen may be able to acquire properties that favor evolution in harsh and changing environments via the exploitation of new niches.

With this dataset, we used different complex systems approaches such as network theory and agent-based models to better understand the structural genetic diversity existing within 62 lichen populations of L. pulmonaria and we reproduced general features of this system. First, we analyzed the fungal-algal population network structures of the data (8 fungal-specific and 7 algal-specific microsatellites) from the L. pulmonaria fungal-algal partnership. We then reconstructed the empirical bipartite genetic network, and obtained common network metrics (e.g., nestedness and modularity). Moreover, we introduced the particularities of the fungal-algal interactions in a continuous evolutionary algorithm based on the widely used ECHO framework [15; Holland 1995, http://tuvalu.santafe.edu/projects/echo/]. With these analyses, we hope to obtain similar patterns of diversification as well as ecological interactions, allowing us to better understand the mechanisms driving the evolution of the symbionts in the L. pulmonaria system. Determining the co-evolutionary relationships and dynamics in the L. pulmonaria system will help us to better understand the role that symbiotic interactions play in the generation and maintenance of biodiversity in forest communities.

Group Contact

Participants

Simon Carrignon
Aina Olle Vila
Julia Adams
Salva Duran

Meetings

A Parsimonious Agent-Based Spatial Reconstruction of Variations in Language Usage

Keywords

language, evolution, ABM, simulation, sufficient conditions, necessary conditions, multi-modelling

Summary

The project was born as a data-driven attempt at providing a sufficient and parsimonious explanation for observed stylized facts in the temporal evolution of language. We have unspatialized (non-georeferenced) data. A few well-known drivers pushing language(s) evolution exist. The idea is to develop an agent-based model where individuals, grouped in sub-populations, roam around, either offering and adopting “variations”. Do mutation also occur? How? The population is constituted by groups/communities of heterogeneous agents (possibly, both at the micro- and the meso-scale) whose initial characteristics are bootstrapped with our data. An individual is more likely to be similar to individual belonging to the same community. The model(s) will be trimmed in order to give a realistic temporal evolution, matching as closely as possible the real data. In particular, we look for alternative spatial configurations leading to a close match vs. a bad match, offering a tentative causal explanation for the stylized facts observed in the data. Multiple models will compete in order to offer the best match.

Everybody is expected to step a bit out of their comfort zone (wrt data, modelling paradigms, …)

Group Contact

Matteo Morini matteo.morini@ens-lyon.fr

Participants

Hamza Giaffar Juste Raimbault William L Hamilton Danilo Liuzzi Philip Pyka Sina Tafazoli

Data Sets

MITRE Data Sets

The two data sets we have access to are Defense of the Ancients 2 (DOTA 2) and Polish Power grids. To access the data please contact Juniper she has it on a hard drive. If you have any specific questions about the data you can contact Matt Koehler at mkoehler@mitre.org.

DOTA 2 DATA OVERVIEW

MITRE Challenge Questions and Powergrid Data Overview

MITRE Challenge Questions Overview


Archived Projects ("Parking Lot")

This section is for projects that we decide not to continue with. Maybe they're ideas that can be picked back up later (hence the "parking lot").

Effect of observer on their environment

  • Still meeting informally.

Keywords

uncertainty, observation, agent-based modeling

Summary

I'm interested in looking at how an observer affects their environment or the system they are studying. I've talked to a few people about this, but the idea may need more refinement to make a good project.

I've added a channel named observer_effects to the SFI CSSS 2016 slack group.

Other Ideas

If anyone has more specific ideas, please add them here.

JE: My perspective on this is through information theory: information transfer from self-past to self-future, conditioned on self-present. or something. This would require that information to be transferred through the environment. This perspective does not require the "self" to be an "observer" per se, but I expect they are intimately related.

Group Contact

Jacob Hunter, jacob.hunter@pnnl.gov

Group Members

  • Lula
  • Anjali

Interested

  • Lula Chen (nchen3@illinois.edu)
  • Marcus Nordström (manordst@gmail.com)
  • Abigail Devereaux (abigail.devereaux@gmail.com)

Can we use metabolic networks to predict the next beneficial mutation?

ideas

Richard Lenski (who is also associated with SFI) has evolved an coli population to use a new sugar source, and tracked their changes for 40000 generations. This is a widely studied dataset (>200 papers published on it already...) in one of the most widely studied biological system but there is still a lot we do not understand about it. For example although we know what genes evolved during the experiments, we do not know why it is these particular genes that changed. I think it would be cool to try to use genetic network (which is also very well understood) to try to understand how new mutations rising in frequency changes the performance of a bacteria, and how the previous state of performance changes what the next mutation should be. The genetic interactions are pretty well mapped in e.coli.(Regulondb) Also they claim that the performance plateaued, but the mutations that accumulated were still beneficial in the following paper. How did that happen? (Genome-wide Mutational Diversity in an Evolving Population of Escherichia coli) From a more computer science perspective: 40000 generation for an iterative process doesn't seem like very long. Does the network structure of metabolism pathways allow rapid adaptation? What can we learn from these networks to apply to computer science problems? These ideas are pretty raw... but I think something evolution related, that looks not just at a property of a system but also allow the system to change would be really cool. My email is Chenling Xu <chenlingantelope@gmail.com>.

Interested please sign up below

  • Ryan McGee (ryansmcgee@gmail.com)
  • Chris Revell (cr395@cam.ac.uk)
  • Dmitry Alexeev (exappeal@gmail.com) - though its my field - i would like to help and watch the idea developing - as well as learn some tools)
  • Thomas Zhang (tomzhang@umich.edu)


Meta-Epistemology and the Complexity of Research Areas (tabled)

The project has been tabled

Keywords

epistemology, complexity, mathematics, irreducibility, methodology, models

Summary

There are a lot of reasons modeling techniques develop in different fields, some of which include: 1) certain techniques and methodologies are more suited for the phenomena being studied by the field, 2) there's often historical path dependence of modeling methodologies that may or may not be suitable for the phenomena being studied by the field, and 3) it's difficult to determine what kinds of instruments are suitable to the phenomena, so researchers may copy the methodologies of other fields. This project would be a general effort to categorize fields by the complexity of the phenomena being studied (obviously we will have to break many fields into subfields), determine whether the methodologies historically and currently being employed by the mainstream of the field are suitable for the level of complexity of the phenomena being studied, and look at the replication rate for mainstream findings on real data in the field to determine whether there is a relationship between the use of unsuitable methodologies and the lack of replicability.

Some References



Group Contact

Abigail Devereaux

Interested Participants

  • Simon: I just presented in that conference something about : "Computer modelling and simulation as heuristic tool to understand the past". It would be a pleasure to participate to the discussion!
  • Ben Zhu (b.zhu@tudelft.nl) I am interested in philosophy of science.


Acquiring Information: to learn an existing language or to develop your own?

Summary

In situations where a problem's solution is presented in an unrecognizable language, we are interested in the trade-offs between learning that language and finding a solution in your own language. We intend to study several forms of language and information; the project could take many possible directions of widely varied scope:

  • Linear algebra: is it "better" to diagonalize a matrix or to represent eigenvectors in a non-eigen basis?
  • Computational mechanics: is it easier to learn/synchronize to an existing (perhaps non-optimal) model or to develop your own e-machine?
  • (motivation) Indigenous lore: Is it better to learn about an ecological system by listening to the lore of a native population or by strictly adhering to western scientific methods?

Group Contact

Jeffrey Emenheiser (jemenheiser@ucdavis.edu) [I am almost definitely not going to be able to do this project. I have too many others at higher priority.]

Interested Participants

  • Santiago
  • Jesús Arroyo (I'm interested specially in the first question about eigenbasis)
  • Moriah
  • Chenling (chenlingantelope@gmail.com)

A costly signaling prestige good model on a social network

Summary

Note: I've changed the description/scope of the project.
Prestige goods, in a nutshell, are goods whose primary value is in conveying social status/ranking (think Rolexes and iPhones). There is a great deal of speculation about how they contribute to increasing social complexity, including the idea that they were an older signaling system that predated true social stratification. Aimee Plourde has developed a costly signaling model of prestige goods that AFAIK has not been a) implemented as an ABM or b) tested out on a actual hunter-gatherer social network. We would like to remedy both of these omissions to explore the dynamics of prestige-as-costly-signaling over time, and whether it gives rise to any suggestive emergent behavior.
The basic idea is to generate a "typical" hunter-gatherer social network (albeit based on historical ethnographic data). Individuals will start out with different levels of skill/success (the base case is low/high) and will interact with others in their network, resulting in gain/loss of skill and/or prestige.
Data for this is iffy, so primarily we would be looking at it as an exploratory model (but if anyone has data ideas let us know).
Kantner from the Santa Fe School of Advanced Studies has an alternate game-theory model for prestige goods that would be interesting to implement as well as time allows.

Group Contact

Ellen Badgley (flyingrat42@gmail.com)

Interested participants, please sign up below

  • Simon Carrignon: I wrote already a general ABM to study that kind of things, but would be really happy to start from scratch something new in python or whatever!
    • Awesome! I started reading your paper and it looks like a great starting point - maybe we could extend the model to distinguish between common goods (subsistence-level, go away after each season) and permanent prestige goods that convey social value, as well as looking into the implementations of prestige on agent actions. If you would like to continue working on this for CSSS let's talk. - Ellen

Donovan Platt (donovan.platt@students.wits.ac.za)
Will Hamilton (wleif@stanford.edu)
Joris Broere (j.j.broere@uu.nl)
William Leibzon (william@leibzon.org> - mostly observer/lurk mode

Group Organization

We have a Slack channel on https://sfi-csss-2016.slack.com. If you need an invite to the Slack team, email flyingrat42@gmail.com.


Meetings


We're meeting in the coffee shop at 6:30, June 15th.

Tracking the migrations of urban hipsters (aka spatiotemporal analysis and scaling of labor)

Summary

Note: we decided to set this aside for now, at least the original question, because Bettencourt beat us to it - the diversity of labor (by different measures) does scale with urban size. The geospatial distributions and the dynamics over time in how the proportions of certain industries change within a city (and spatially) are still intriguing. We have mostly switched to looking at the Japan data on material distributions, but if someone has a brainstorm and wants to dust this off, go ahead!
This is linked to the "Viscosity of Labor" question on the MITRE Challenge Questions list, which is copied below:
Given US Census collected data can we find relationships between labor and urban scale?

"Using US Census data (and other sources) a number of interesting scaling laws have been discovered that relate to the dynamics of urban human social systems. These scaling laws relate to such things as the generation of intellectual property, income, tax revenue, crime, and so on. What about labor? Does the scaling seen in income come from new or shifting categories of labor or simply increasing the income within an existing (static?) distribution of labor categories? Is there a spatial component? Does the spatial distribution change with the scale of the urban area?"

The above is the complete question from MITRE, but we can adjust/focus as needed depending on interest and ideas. I have a sample analysis in another context from a colleague (who actually came up with this idea) which gives some good starting ideas. One thing I would like to look into is appropriate diversity measures for labor distributions.

There is plenty of US Census data available for this:

We would probably start at the county level and go finer (census tract/block group) if time allows.
Matching the labor categories across census years, etc. is going to be the main challenge.
MITRE and SFI are actively working together to improve the 2030 Census, so we can reach out to SFI directly on this.


Group Contact

Ellen Badgley (ebadgley@mitre.org, flyingrat42@gmail.com)

Interested participants, please sign up below

  • Lorraine Sugar (lorraine.sugar@gmail.com)
  • Usama Bilal (ubilal@jhmi.edu) Really interested in exploring this kind of data. I once cleaned data on some occupations by census tract from 1930 onwards (for Baltimore city, but can be reproducible). I also have a couple R scripts lying around to extract stuff from ACS easily.

Group Organization

We have a Slack channel on https://sfi-csss-2016.slack.com. If you need an invite to the Slack team, email flyingrat42@gmail.com.

Meetings

Let's meet at lunch on Thursday June 16 to talk about this further.


Searching the various non-linear landscape motifs of coupled chemical reaction networks for sensitivity to evolutionary tunable parameters

Summary

The goal of this work is to show the potential of tight control over the the location in phase space of non-linear coupled chemical reaction networks via the manipulation of individual parameters. Chemical reactions, when sharing in a combination of products and reactants, can be represented by systems of nonlinear equations. Both reaction rate constants for each involved reaction and the abundance of open system components can be manipulated as parameters. After a purely mathematical investigation of various coupled reaction motifs, two applied systems (one biotic, the other abiotic) will be investigated for their ability both respond to, and drive, phase space location respectively.

Group Contact

Tucker Ely (tely1@asu.edu)


Group Organization

Meetings

Conceptual Drift

Keywords

history of ideas, evolution of philosophy, perception, behavior, feedback

Summary

I have talked with a few people regarding my curiosity surrounding the relationship between the way we perceive of ourselves in relation to the world around us and how these concepts are informed, embodied, and reinforced by a myriad of behaviors. In the process of questioning how philosophy can be considered a complex system some interesting ideas have emerged. After a brief conversation with one of our colleagues about "philosophical species," I identified an open-source data set that represents a Philosophical Influence Network (PIN). It represents quite a significant network spanning back to pre-socratic philosophy. I think this can be a decent starting point for some impressive narratives. Scottish philosopher David Hume, likened philosophical disciplines to species and this metaphor may be appropriate to utilize as we consider the evolution or "conceptual drift" of philosophical ideas.

I would also like to add that I am completely open to ideas surrounding this topic, i.e. I noticed someone mentioned the evolution of genes, memes, and temes; I think to some extent we have an overlap in interest.

Another caveat is the source of the PIN is problematic; it is sourced from the "influenced by" side bar on wikipedia for philosophers. Thus we encounter bias, fanboys, conflict, and other kinds of noise. However there are other databases that may be more objective i.e the Stanford Encyclopedia of Philosophy.

Potential Directions

Where ideas go to die: Considering the richness of the PIN, it was suggested we may be able to search where ideas do really well and where others go to die. Each philosopher represents and attractor, furthermore each attractor interestingly represents ideas or concepts embodied within philosophical characters. This focus allows us to consider the drift, mutation, and other evolutionary processes surrounding concepts embedded in networks.

Why do "good" ideas go "bad?" : In this vein a couple of people have mentioned how we may explore what drives religion into war. Religions historically are meant to promote a kind of moral reasoning however throughout history as well as in contemporary society we see that religious fanaticism can drive violent acts. To what extent can we account for religious extremism and the emergence of groups like ISIS, and to what extent can we trace how violent ideologies emerge from moral doctrines?

Philosophy Divided: As we participate in the CSSS 2016 I can't help but recognize many philosophical assumptions we seem to tacitly evoke and at the same time identify other areas as painfully philosophical. In many regards this tension between the sciences and philosophy date back to Bacon's rejection of Aristotelianism. And as the scientific revolution emerged, a focus on the techne (artifact) also emerged, as a technical understanding of the physis or natural world, accompanied with technological advance. Today it seems as though the Aristotelian notions of techne and physis are "coming back around again." Can we account for the evolution of philosophical ideas in such a way that we can reimagine ancient philosophical concepts as complementary to novel research methods.

Group Contact

Justin Williams (justin.williams@unt.edu)

Interested

Markus Junginger