CSSS 2010 Santa Fe-Blog

CSSS Santa Fe 2010 Home Surveys/Evaluations U.S. Visa Requirements Travel Logistics Faculty Participants Schedule Readings Background Readings Projects & Working Groups Tutorials Blog Student Notes After Hours Photos TShirts Project Presentations Final Papers

Use this page as an informal forum to share your opinion and discuss anything at CSSS'10.

• Michael Szell: Some people asked me about the URL of my online game. Here it is: http://www.pardus.at - Enjoy! :) (The tutorial may some time to complete..)

• June 7 (Ana Hocevar): So I know my official day of contribution on the blog is not until tomorrow, but I wanted to share this with those who are interested. When I get inspired by someone giving a good talk, I have a tendency to write down some of the statements I find encouraging or funny and so on. So here are my favorite quotes from the first day of lectures (and I hope the lecturers don't mind this):

"If there is one thing you should learn at the summer school, it's to speak up." -Dan Rockmore

"Science is a social thing." - Dan Rockmore

"You start with a beautiful idea and end up with reality." - Dan Rockmore

"Commit to taking advantage of this opportunity." -Dan Rockmore

"Let it all marinate up there." -Dan Rockmore

"And wear sunscreen." -Ginger Richardson

Dan Rockmore: Here is a link to the NYR piece I mentioned today http://www.nybooks.com/articles/archives/2010/jun/24/other-side-science/ Good to meet you all and looking fwd to a great CSSS.

Monday June 7

Maria Opazo

Alison Snyder

I thought it might be interesting to give my perspective on today as a journalist and writer. What I starred and underlined in my notebook, what stuck in my head and why.

First as a journalist…

Schooling Fish

I’m familiar with some of Iain’s research – the beautiful images and descriptions of phenomena that many curious people outside the sciences have seen and thought about lends itself to visual storytelling -- but today was the first time I heard him discuss his fish work. In particular, I’m going to follow what he’s looking at in terms of how individuals influence others in the group and the (preliminary?) finding that some individuals consistently influence the group. Because it is unpublished, I’ll check in with him periodically to discuss in more detail and see how the research is progressing as well as whether and when he might expect to publish the research.

Moiré-ing

Have you ever seen someone on TV who's striped shirt competed with the gravity of what they were saying? In TV production we pay editors large amounts of money for hours of work to fix moiré-ing so the term “numerical moiré-ing” caught my attention.

Chaotic Mixing

Liz mentioned her CU colleague who described clams that open up to exploit chaotic mixing to mix up their gametes. The counterintuitive idea of systems not only embracing chaos but exploiting it, is a provocative theme to explore in a general audience story. To most people, chaos is something to be avoided.

Ideas that got my attention as a writer…

Gas Gauges and Indelible Images

In opening her lecture, Liz used the metaphor of her old car’s gas gauge to illustrate non-linearity. The gauge stayed on full then plunged when the tank was near empty rather than dropping as the gas level dropped. While it was a brief aside about a concept that is very easily understandable to a scientist, it might not be to a non-scientist. To someone who has never heard of or thought about non-linearity, they will understand it the first time they hear this metaphor and they’ll remember it. Even if they forget what the gas needle was illustrating, they can back track. Non-linearity becomes an indelible idea.

As a writer, the metaphor also reminded me of the power of a simple idea in telling a story. One of the most difficult tasks is parsing the scale of what a story is about. Sometimes the most meaningful way to tell a story about the Empire State building is to describe one of the bricks.

In other words, today was full of lots of ideas about research to follow and reminders about how to effectively communicate even the most basic of ideas and the power of doing just that.

Looking forward to tomorrow…

Kyla Dahlin

My notes from today are mostly graphs, equations, and Dan's quoted quote about "what if biologists had tried to develop a theory of gravity." I'd love to see a debate between the math/physics/universality camp and the biology/ecology/sociology/everything is different camp. Seems like an interesting tension within Complex Systems folks.

So far the self-organization into project groups seems to be going well, and I'm sure it will evolve over time. Tonight a few of us talked about the pressure to collaborate being somewhat foreign - so much of PhD work is so solitary. I'm sure the sociologists would love to track the networks that are forming, the different players, and our final results.

And I think we all are trying to maintain a balance of fun, thinking, working, and staying level in a world with constantly flowing coffee.

Lucas Antiqueira

Some first impressions:

This first day of lectures was really nice. I was exposed to subjects I'm not quite familiar with, lectures were excellent and inspiring, projects started to evolve. The process of getting to know people continued, there are so many brilliant minds here.

The summer school is a dream coming true for me. I've been in Santa Fe for only a few days, and I can definitely tell that these days are among the most gratifying in my academic life. And things have just begun!

Many thanks for the SFI/St.John's staff for the support! Not to mention the food, which is excellent by the way!

Tuesday June 8

Andrew Banooni

Some of you have heard me use this analogy, and I’m certain that many others of you have heard it elsewhere, but they say that the amount of material there is to learn in medical school is a bit like trying to drink water from a fire hydrant. One thing that I had not realized until attending yesterday and today’s lectures, however, is how homogeneous that information seems, now that I’ve gotten a taste of the multidisciplinary approach here at SFI. Sure, there are a tremendous number of disease processes that involve vastly different physiologic systems, and don’t even get me started on all the bugs and drugs we have to learn, but it’s all medical. We don’t sit and discuss econometrics (actually I don’t think I’ve ever had a discussion about econometrics), or the influence of individual fish on collective behavior. Over the past two days, I’ve dusted off cobwebs from parts of my brain that I feel I haven’t used in years. I don’t mean to say that I am rusty in certain topics (and completely new to many others), but rather that I feel that in the past 48 hours I have been thinking in ways I haven’t had to think in my graduate training. In medicine, you can survive by learning a tremendous number of facts and then regurgitating them at the appropriate time. We observe, we recognize patterns, we diagnose, we treat. I am so thankful for this refreshing reminder that there is so much more enrichment to be had, that the medical problem I am trying to solve can share many similarities with a physical, financial, or aeronautic one my colleague might be struggling with. I feel inspired by every one of you, very lucky to be here, and I look forward to three amazing weeks!

Andreas Ligtvoet

Liz

• It's a skill to explain difficult stuff in an easy way.
• Lots of dusting off of things I should know, but forgot.

Peter

• Some interesting discussion about the end of theory due to access to large amount of data. It doesn't help you understand stuff, though.
• Walmart seems to have enormous amounts of data. A colleague of mine suggested finding ecologies of consumer products. What type of furniture goes best with red napkins?
• We can create huge amounts of non-existant networks and have fun with them.
• An average node degree of about 4 leads to hairballs. What can we do with those?

Iain

• The larger the group of naive individuals, the (relatively) easier it is to influence it. For 10 individuals you need 50% 'leaders', whereas for larger groups only a few %.
• Another interesting remark either yesterday or today: just because you have a nice model for some fish species, doesn't mean you can apply it to cows or bugs or. ..
• THe details of the system do not matter in a collective transition.
• As soon as you behave differently, you are nailed.

Tom

• How much to put in a model? Not too much, says Tom. However, what good is an "economic" model that shows a Bolzman distribution? His answer is that this is a natural law in other words, you will never make an economy where the Bolzman distribution does not exist. It is even worse if you allow people to invest.
  • In the state of Ilinois, Pi was legally defined as 3.
  • Stochastics from στοχαστικός, from στοχάζομαι ‘aim at a target, guess’, from στόχος ‘an aim, a guess’.
  • Mention power law and get published!
• Immunize Pig/Mexican/Flu: make use of power law in a network => your friends are more connected than you are! Fascinating...
• We don't know all that much about networks: we don't know what matters.
• Agents with genomes: 10010101. DIstribution split of genome crossover matter!
  • T-shirt sniffing leads to selection of different immune systems.
• In complex systems there is a lot of exploratory stuff.
• Tom is a (scientific) heretic.

Drinks

@Cowgirls. What's this ID thing about anyway? Don't I look old enough?

Xin Wang

The topics today are involved in nonlinear dynamics, complex networks and collective behaviors. I am very interested in those topics before this summer school, and through lectures I get the deeper understanding about those three areas.

But what impresses me most is that the students here are very active and it is different from the situation in China. In the meantime, I enjoy the joy of communicating with people from different backgrounds and getting cross-field collaboration.

Ana Hocevar

I am not a blog person really, I've never written a blog before, but I guess CSSS is also a great opportunity for trying out new things unrelated to science. So, here I go.

As many others, I can't get over how amazing Iain Couzin's talks were. I am very impressed and greatly inspired. Usually attending physics conferences that left me wondering what it was that I was missing, Iain I think made me realize I prefer systems that include things with eyes and wings. Or perhaps gills? Amazing work in my opinion, really.

I should also thank Liz Bradley for giving incredibly clear lectures on nonlinear dynamics. From the courses I had on chaos, I wouldn't imagine such an intuitive presentation of the topic is possible. It certainly isn't easy, so: Liz, you rock!

I haven't been enjoying only the lectures, though. Dancing to no music with Andrew, listening to Jonathan playing the violin, stimulating discussions over lunch or breakfast, great food (we even have ice-cream!) and so much more...

So with such great company, such cool science and so many wonderful things still ahead of us, I can only say I am very grateful to be here.

Borys Wrobel

I will not be the first to say that Liz Bradley's lectures are something to behold. I am constantly taking notes on not what she explains but HOW she does it. Especially the mix of powerpoint and whiteboard/transparencies. Hopefully I can decipher my handwriting when it comes to explain some of this stuff when I am teaching. Iain Couzin's are also very inspiring, especially on the advantages of keeping things simple. It was even more inspiring to talk to him @cowgirls, especially his views that it is important to discourage grad students/postdocs to keep TOO long hours... and the importance of having a 'fun lab' in attracting people that would like to work you. It was fascinating to see how Tom Carter engages the audience, so late in the evening (and after dinner, too!). I am not sure that we Europeans agree on his view on how the economic inequalities can be addressed, he seemed very pessimistic about the possibility of it, and many people in the room, I think, were asking themselves how introducing taxes etc. would affect the distributions.

Other remarks

Some people may be interested in the http://decoi.collectivae.net/ Design Of Collective Intelligence series - those are 1 week long versions of what SFI is trying to do, but mainly focused at multi agent systems.

Also interesting, and this is even a longer shot, is http://www.nextgenerationinfrastructures.eu/academy if you are interested in (physical) networks and infrastructures.

Here is an article about the spread of contagion in a social network written for a general audience, as an example of science journalism:

"Infectious Personalities", The Economist, May 13, 2010. http://www.economist.com/node/16103882

AS

Wednesday June 9

Today was a sad day for Dutch politics. Ligtvoet

Thank you Paige and Coco. Samuel_Scarpino

Great Kerouacish moment. Thanks hairy thighs! Giovanni Petri

(jp) Man, I don't know what went on last night...

Ingrid van Putten

Lots can be said about the visual art - but the bar has definitely been raised by mathmeticians and computer geeks over the past two days at SFI. Not only did we have buildings tumbling down or having fish swim out of them at the Santa Fe Complex last night, but today liz completely outdid the lot by making music and then ... dance. Those performances should make maths, physics, and computer science interesting to anyone!

Thomas Maillart

This morning Peter told about network attack tolerance. The results obtained by Albert et al. Nature (2000) for scale-free networks are a direct consequence of the "robust yet fragile" (also called HOT - Highly Optimized Tolerance) concept first coined by Carlson et al. Physical Review Letters also in 2000. However, Newman et al. criticized the HOT model (in Physical Review Letters, 2002), arguing that while "optimized", scale-free systems, the optimization process should occur at some costs. Thus, they introduced the COLD model ("constrained optimization with limited deviations"), by adding a cost to optimization (i.e. preventing the formation of scale-free systems). Under these conditions, the tolerance to attacks should be improved. The interesting point behind the message, is that there is an intrinsic trade-off between optimization and risks, i.e. optimized systems are prone to suffer more extreme damage (heavy-tailed distribution of damage).

On the same topic, Doyle et al. (PNAS 2005) argued that the model by Albert et al. (Nature 2000) does not hold for the Internet because the most connected nodes are not in the core of the Internet, which is typically low connected. The main point of this article is to point the importance of looking at the mesoscopic structure of real networks, and not only their degree distribution or general purpose metrics.

Vessela Daskalova

Some thoughts inspired by Tanmoy Bhattacharya's lecture on inference in historical processes and by Tom Carter's modeling workshop:

They both tackled two classical questions regarding the methodology used in different fields.

1) Are we trying to explain past observations or are we trying to predict future events? And how do models in different fields reflect this goal?

2) Do we want to build in what we already "know" in models we are building/ regressions we are running?

Obviously, people from different fields will answer these questions differently. It would be cool if at least one person from each field wrote about what is accepted in their area.

Here one point of view from economics:

on 1) In contrast to historians, who try to give a good explanation of what happened in the past and why, economists try to make some predictions regarding the future. Therefore, models/regressions in economics include fewer variables than models/regressions in history and political science. Tanmoy Bhattacharya argued in favor of including many explanatory variables when using maximum likelihood for historical inference. However, if the goal was prediction rather than inference, the number of explanatory variables would have to be reduced to avoid introducing a lot of past noise in the predictions.

on 2) In the social sciences (especially economics) there is a tendency to build in some assumptions of what matters in your model. This seems useful (at least to me) as many of the phenomena we are analyzing are a result of social interaction, there is a large degree of randomness in the environment and unless we use our a priori knowledge (as Tanmoy Bhattacharya argued yesterday), we cannot expect to come up with some deterministic equation that accounts for social phenomena.

Coming back to prediction in economic models, let me end with a popular economics joke: "Why do meteorologists make weather forecasts?" "To make economic forecasts look good." Just to clarify, this is not meant as an insult to meteorologists, but as self-irony:)

===Kasia Samson===

Borys Wrobel

It is really too bad that Peter Dodds did not have the time to present his last lecture, I was really interested in hearing it. I feel this was the cost of the historical perspective (his first lecture)... Santa Fe Complex sure was fun, it sure seemed a geek paradise. It was great to see the American culture of innovation at its best... But it sure was a grueling day, having to get up earlier than on the other days, running to the buses, going from one place to another. Not enough time to explore the SFI (and its library) really, hopefully it will be possible next Wed... It is amazing that so many people decided to stay late at night and walk back to St John's...

Thursday June 10

Chaitanya Gokhale

"The trick about life is to make it look simple."

It would seem quite ironic if I say this at a “complex” systems summer school, but the more I get immersed in the subject the more I think that it is relevant statement. I mostly apply nonlinear dynamics to develop mathematical models of biological system and apply evolutionary game theory to biologically, socially relevant concepts. Please note that both of these is not data driven to begin with.

Oxford Dictionary defines “model” as follows (model in the context as we use it),

...3 something used as an example. 4 a simplified mathematical description of a system or process, used to assist calculations and predictions. 5 an excellent example of a quality....

A model is not the actual system. Tom Carter gave an amazing demonstration of that fact. The models he was discussing were not of as much importance as the message he was trying to put across. We do not need complex models to understand complex systems, rather simple models is what we should be aiming for because the goal is not to replicate the complexity but to understand it in simpler terms. The same thing was iterated by Owen Densmore at the SF Complex and saying that what they are doing is complex would be the biggest understatement. Today again Jure just amazingly deflated the whole network complexity into a simple 2 \times 2 matrix confirming the idea ever more.

Any fool can make things bigger, more complex and more violent. It takes a touch of genius and a lot of courage to move in the opposite direction --Albert Einstein

On a different note, I loved Liz Bradley’s lectures. Though I have been using the methods for quite a while now, I had never seen such a crystal clear overview of nonlinear dynamics before.

Jing Li

When Google first topped the 100 best companies to work for, Fortune's comment was 'It's the kind of place where, yes, you're going to work but you also know you are going to have a fun time as well'. I would like to say exactly the same for CSSS 2010.

What a beautiful day today! Liz's final presentation is as excellent as previous: science is not only interesting, but also COOL; Nathan is as impressive as his project at CSSS; Jure is so passionate about his research. Dan, your comment is the highlight of the presentation, as always.

A friend of mine who was in CSSS 2009 wrote 'I hope your time at SFI is as wonderful for you as it was for me'. Yes, it is!

Bradford Cross has a great post on learning about network theory that everyone with an interest in networks will probably find useful. Look at it or bookmark it for later, it compliments our previous lectures well.

The CSSS hash tag on twitter is #CSSS, the REU tag is #SFI.

Bruno Abrahao

One of the recurring themes in our lectures was power-laws in the degree distribution of complex networks. Since this feature was so prevalent across different networks, coming from many different domains, researchers started to create generative models that replicate this observation, with the goal of shedding insight on the root cause of this phenomenon. Some of the plausible explanations include preferential attachment, the copying model and others.

In 2004, Li-Alderson-Willinger-Doyle made the point that observing a power-law does not give us any insight on the underlying phenomenon. They show that a large number of different networks, with dramatically different characteristics, exhibit the same power-law behavior with the same power-law exponent. The paper was very well received by the computer systems community (it received the best paper award in the top conference in the area, ACM SIGCOMM), but it seems to be often ignored by other communities.

To make matters more complicated, Mitzenmacher made the point that it is virtually impossible to distinguish any generative process that exhibit power-law behavior from those that exhibit lognormal distribution.

Can we conclude that power-law observations must be accompanied by some form of validation of the root cause of the phenomenon in order give us useful information on complex systems?

Anna Pechenkina

Thanks, Bruno, for bringing this up! Networks are new to me, but the published findings of power laws in agent-based models is something that I've come across quite a bit. The question of model validation seems to lack a good answer (and expected practice, which is true at least in political science). If a model generates a power law distribution of some parameter of interest, normally, people do check for robustness of the finding. However, even when the model produces the power law across a large portion of the parameter space, this is no guarantee that the model is a good model of the phenomenon of interest (Tom Carter mentioned on Tues finite vs. infinite variance). The fact that there may be other ways to reach the same power law distribution calls for some sort of validation of the model's mechanism. I would like to hear what others believe a good example of "validation of the model's mechanism" can be (testing the model's *additional* implications against observational data? designing a test of whether the subjects use indeed the decision-making rule specified in the model? modeling different decision-making rules?) Cites of good examples are much appreciated!

Griffith Rees

Agent-based modelling is still a pretty new thing in sociology, and validation has always been a weak suit. Peter Hedström has attempted to mix empirical results with simulations to assess the relative impact of different effects, and the extent to which they explain macro phenomena. In his case, he is modelling diffusion of unemployment, and the extent to which being socially tied to other unemployed people increases your likelihood of becoming unemployed and increases the length of your unemployment. I don't think his validation really does the job, but it's an interesting approach.

Friday June 11

(Kang Zhao) I really like what Nathan Eagle's idea of Engineering Social Systems. Yesterday was the 3rd time I heard his talks. I feel that, in the emerging area of network science, people have been focusing mainly on understanding the patterns in all kinds of large networks. No offense to those outstanding research. But... OK, power-law distribution, highly-clustered social network, 6-degree of separation, then what? I have engineering background and always love research that can make a difference in the real world. Can we build more applications to exploit the network structure/pattern to make people's life better/easier? While there have been some efforts (such as the DARPA network challenge this year) and interesting applications, I don't think network science is quite there yet. (Of course, we need to keep privacy issues in mind when we mine social network data).

Florian Sabou

A recent article in The New Yorker, "Silver or lead" by William Finnegan, reveals the deep complexity of the drug war reality in Mexico. Rather than a typical armed conflict between authority and drug gangs, or between gangs themselves, the violence in Mexico presents characteristics of political, social and religious insurgence. Drug gangs offer public services that the central government fails to implement such as: construction of community sport and art centers, administration of justice, unemployment management, security and even drug-rehabilitation clinics for methamphetamine addicts. An overview of the article can be found at http://www.newyorker.com/reporting/2010/05/31/100531fa_fact_finnegan and for interested readers the printed magazine issue is available.

Roberta Sinatra

… and so this first week at CSSS is over. And it has been a very "intense" conclusion indeed! I would like to make a couple of comments/reflections, scientific and not:

• Lectures of Leskovec and the need of large dataset analysis: at the conference NetSci2010, László Barabási stated that "by 2023 the number of stored bits will surpass Avogadro's number". In my opinion, if this is true, no matter how good will be our tools for the large scale analysis, we will never be able to deal with a "microscopic" description of data. In the same way, for example, we are not able, and anyway it will be not that useful, to describe all the molecules in a gas using their dynamical equations, but we can just deal with the statistical ensemble of all the possible configurations they occupy, and study the constraints the system meets and the properties that derive from them. Therefore, what probably we will need in the future, is try seeking in all our data generality and common features as well as analogous constraints, by looking at them as realizations of a big ensemble of possibilities that are, in some sense, all equivalent and could share the same properties.

• Lecture of Hübler: I enjoyed it so much! I have a background as a physicist and because of this I attend a lot of "lab courses"during my studies. Nevertheless, since I work on complex systems, I had never the chance to enjoy them from a pure "experimental" point of view. The opportunity to go to the lab with him will make us figure out that complex systems are much more of a bunch of data someone collects for us, but they are most of the times systems we can reproduce. ANd much better if these experiments are eatable ;)

• Power cut during the lectures: are we still able to teach or give a lecture without the use of slides, but just with a piece of chock and a blackboard?

• The dinner at the Santa Fe Institute has been very pleasant and I really appreciate its familiar, welcoming and friendly atmosphere. And a proof of this is given by the fact that a cat that lives there in the same way he could in one of our houses! (By the way, never seen a cat in a research institute! Do you?)

• Mafia game of friday night: some of us played for 4 hours (yes, 4 hours!) the famous - but not to me - mafia game. It was a very good opportunity to socialize with other people, because when far from lectures and "working" hours, none of us feels he has to "prove" to be smart and brainy, and it is easier to be natural and spontaneous.

Bogdan State

I sure hope no one got annoyed at my hunting for candid pictures today at SFI. I am a very awkward person, which makes my tries at being a photographer...well, awkward. So thanks to everyone who smiled and went along with my snapshot-happy self. I hope seeing the picture will make you more tolerant of the annoying guy with the pesky lens:

http://picasaweb.google.com/worldmeetsbogdan/SFIBarbeque#

As for the CSSS, I'm in total awe of the program. This week of lectures has felt overall like gym for my brain - particularly all the advanced math in Liz's class. Hopefully, brain gym will be easier to keep on a schedule than other kinds of gyms...cough...

Bruno Abrahao

This is a timely analysis of an interesting complex system. I personally like his results! ;-)

Cornell University professor predicts Brazil will win 2010 FIFA World Cup title His soccer study submits world's best teams to statistical analysis.

Lucas Antiqueira

Nature's article: High hopes for Brazilian science

Complex Emergence

Hübler: "So, definitions of complex systems? "
Erik van den Broecke: "Female."
Sarah Wise: "Clearly you never dated a male"

Hübler: "If an ant would ask you to teach her calculus, what would you tell her? You would tell her, sorry little ant, that's a noble request, but you just don't have enough neurons!"

(Saturday June 12)

Simple Pleasures:

Rides from Leroy (and Paige/Coco)

Proper Espresso

Griff's Parkour

Dutch Heaven

Gold Ones

I study English long time

Evolution of Secondary Sex Characters in Northern Europe

"If I were only at sea level . . . "

"I'm putting that on the blog"

Monday June 14

Andrew Hein
Tracey McDole
Sergey Melnik
Erik Van den Broecke

Tuesday June 15

Drew Levin
Leif Karlstrom
Mark Laidre
Borys Wrobel

Wednesday June 16

Joseph Gran
Micael Ehn
Damian Blasi
Daniel Jones

Thursday June 17

Yixian Song
Sam Scarpino
Giovanni Petri
Michael Szell

Friday June 18

Sandra Bennun
Susanne Shultz
Lynette Shaw
Sarah Wise

Monday June 21

Dan MacKinlay
Megan Olsen
Vanessa Weinberger
Jonathan Cannon

Tuesday June 22

Erika Legara
Gavin Fay
Bruno Abrahao
Zhiyuan Song

Wednesday June 23

Julie Granka
Nick Foti
Felix Hol
Oana Carja