Upcoming course: Model Thinking

When my colleagues explained to me what this blog was for, I was really pleased to hear that it would be a forum where a modeling novice could gain some orientation and learn shortcuts that weren’t available, or at least not easy to find, when I was just learning. Modeling, at least the computational side of it, is still in many ways a rarefied specialization in the social sciences, and reliable guideposts are few.

The concept of modeling itself is vast and vague. For some, images of flow-charts and mental maps come to mind. For others, it could mean miniature trains or linear regressions. There are many different ideas about what models are or what they are used for. It doesn’t help that there are two very different career paths that are both called “modeling” (from what I can tell, the crossover rate has been pretty limited).

When someone new begins to dig into the pursuit of modeling, they’re likely to come up against that stumbling block to end all stumbling blocks: MATHEMATICS. Differential equations. Graph theory. Markov chains. Point processes. And coupled to this is an array of seemingly unrelated computer programming languages and development environments with documentation that is not always easy to navigate. If you don’t come from a mathematics or computer science background, it can be difficult to know where to begin. But more importantly, it may not be clear what modeling is actually for or why anyone would ever want to do it.

Enter the Coursera course “Model Thinking”, taught by Michigan’s Scott E. Page. In the first lectures, the reasons why anyone would want to learn about models are laid out in plain English. From the course website:

1. To be an intelligent citizen of the world
2. To be a clearer thinker
3. To understand and use data
4. To better decide, strategize, and design

Page tells us that, in the era of Big Data, the ability to identify key components and apply knowledge are crucial. Data by itself, no matter what quantity, is useless without the ability to harness its informational potential through identifying patterns and understanding processes. Models, it is argued, are just the kinds of tools we need to use data wisely.

It should be stated up front that this is not a course designed to teach you to how to program. There are lots of different courses, tutorials, and other materials on programming which are available, and this site is doing its part to help provide some direction. Instead, Page offers valuable tools for thinking about complex problems using the power of models.

In some ways, it’s like a best-of album: all your favorites are there. Segregation. The Prisoner’s Dilemma. Forest Fires. The Game of Life. These models are used to demonstrate principal concepts in modeling and complex systems, such as aggregation, tipping points, bounded rationality, and path dependence. Real-world case studies are used to show how models like these can illuminate core dynamics in what are otherwise very complex and intractable systems, such as banking networks, electoral politics, or counterterrorism.

The course doesn’t deal outside of mathematics entirely, but introduces the necessary concepts in a fairly straightforward and basic way. The online format suits this well: if you’re not familiar with a concept, you can simply pause the video and Google it.

If you’ve taken a few MOOCs, you know that production counts for a lot, but sometimes it can be distracting. A voice-over with someone’s lecture slides is bound to put you to sleep; too many animations or an overdone background or wardrobe can draw your attention from the lesson. Most of these videos begin with Page in front of a blank background, waist up and gesturing, with key words being displayed at the bottom of the screen. This usually transitions into a demonstration with simple but effective graphics and live-drawn overlays for emphasis (see here). This approach seems to balance the issue of too little/too much production. The weekly quizzes are thoughtful but not overwhelming. I found them particularly good for someone new to MOOCs. In addition, if you’re watching the videos on the Coursera site, many of the lectures will stop part way through and ask a multiple-choice question to make sure you’re paying attention. This does a reasonably good job of reinforcing the lesson.

For someone who does modeling on a regular basis, the course is great for clarifying and compartmentalizing different ideas which you may already be using but don’t know much about their background or how to interface them other other concepts. For someone who is new, it has the potential to shed light on some of the reasons models are used and give some direction in terms of how to use models in your life.

The next session begins on February 3rd and runs for 10 weeks, with a recommended workload of 4 to 8 hours per week. The signup page at Coursera can be found here.

Image “5th Floor Lecture Hall.jpg” courtesy of Xbxg32000 @ Wikimedia Commons

Modelling Across Millennia

Wind back the tape of life to the early days of the Burgess Shale; let it play again from an identical starting point, and the chance becomes vanishingly small that anything like human intelligence would grace the replay.  (Gould 1989; p. 14).

How can we attempt to understand the complexity of life today when we cannot run repeated experiments on the evolution of life? If we could go back to the beginning, would we find that each evolutionary change was contingent upon the previous step? Would stochasticity make every new run of the “tape of life” completely different from the last?

Of course, actually replaying the evolutionary tape to answer these questions is impossible, but through the use of agent-based modeling we may be able to run experiments on a system to understand how the “tape of life” created the complexities we see today.  By exploring the behaviors of agents the modeler can deduce what occurred in real systems. These models allow scientists to study systems in space and time, which are often too large and too long for more traditional measures of study. How to understand Big Data is a problem (see this article for a discussion of Big Data issues) but agent-based modeling provides a way to not only generate that big data, but with proper use, sort it and answer questions of interest that only Big Data can answer.

Repeated calls have recently been made to apply agent-based modeling to contemporary affairs to not only understand crises as they unfold, but also to anticipate them (e.g. see Buchanan 2009; Cabrera 2008; Epstein 2009).  Archaeology is essential for these efforts. It provides that long-term view that a myopic study of our modern problems cannot truly address.

In a 2012 special issue of Ecological Modelling, several archaeologists put forth their uses of modeling to understand past societies, and I would argue, these studies help us further our understanding of current problems. Their geographic regions span from the U.S. Southwest to the South Pacific, the Mediterranean to Mongolia. Crabtree and Kohler provide a good background to the models presented in the issue. They say:

“Modelling of ancient socio-ecological systems is in its infancy. Problems to be faced include both the incompleteness of data imposed by the archaeological record, and the difficulty of developing satisfactory frameworks for characterizing the behavioral plasticity of humans and the evolvability of the cultures they create. We do not pretend that all the problems are yet satisfactorily addressed, but we believe it is important to begin, nevertheless. Traditionally, humans have relied on culture to provide them with a framework for addressing current problems. But as contemporary societies lose their traditional cultural knowledge, archaeology provides our best hope for deriving lessons from ancient cultures to address today’s problems. These articles report on our attempt to build a capability to study the past in ways that make it useful for thinking about our future.  While we may not be able to “wind back the tape of life” archaeology and agent-based modelling offer us new ways to understand human/environment interactions, providing us with a clearer picture of what may have occurred.”

These articles address such issues as the fragility of human existence in unstable environments, how humans can construct their own niches to better survive in these environments, and how small decisions can have dramatic effects to society.

With agent-based modeling still in its infancy in archaeology, this issue of Ecological Modelling should be of special interest to archaeologists, ecologists and modelers alike. It shows how important ABM can be to understanding archaeological systems, and reports fully on five distinct uses of ABM in archaeology. Importantly, it collates these studies into one easily digestible package, and allows for comparison of the different modeling approaches.

To read the articles:

Crabtree and Kohler, summary and intro

Rogers et al. on pastoral Mongolia

Murphy on Hohokam irrigation

Kohler et al. on Ancestral Puebloans and the Village Ecodynamics Project

Kirsch et al. on Hawaiian intensive agriculture

Barton et al. on Mediterranean environmental change

The full issue is here

–Stefani Crabtree


Buchanan, Mark

2009   Meltdown modeling: Could agent-based computer models prevent another financial crisis? (News Feature) Nature 460(6):680-682.

Cabrera, Derek, James T. Mandel, Jason P. Andras, and Marie L. Nydam

2008    What is the crisis? Defining and prioritizing the world’s most pressing problems. Frontiers in Ecology and the Environment 6(9):469–475.

Epstein, Joshua M.

2009   Modelling to Contain Pandemics: Agent-based computational models can capture irrational behaviour, complex social networks and global scale—all essential in confronting H1N1. (Opinion) Nature 460(6):687.

Gould, Stephen J.

1989   Wonderful Life: The Burgess Shale and the Nature of History.  W. W. Norton and Company. New York.

Computer Applications and Quantitative Methods in Archaeology conference

CAA is currently the largest annual conference focusing on computing in archaeology. It usually hosts a session about computational modelling and/or simulations but this year seems to be particularly prolific for complexity science. Here’s a quick tour of what we particularly look forward to:


(W12) Workshop: One hour, one model: Agent-based Modelling on-the-fly”

Organised by myself, Ben Davies, Tom Brughmans and Enrico Crema this workshop will aim at brining together researchers working with complexity science tools. We will divide into small groups and work in parallel on the most common building blocks of archaeological simulations (diffusion of an idea, innovation, environmental change etc) to see how different our approaches are and if different models could produce  different outcomes. We also hope to build a small library of code snippets.

(W11) Workshop: Introduction to network analysis for archaeologists

Run by Tom Brughmans, Ursula Brosseder and Bryan Miller it’s a half day hands-on workshop (so you can come to W12 as well) introducing the basic techniques of network analysis. No previous experience required.


(S25) Session:  Agents, Networks, Equations and Complexity: the potential and challenges of complex systems simulation

A full day session organised by the same team as the ‘One hour, one model’ workshop (Ben Davies, Iza Romanowska, Tom Brughmans, Enrico Crema). Last time I checked we had 18 papers in our session with presenters from all six continents and an enormous breath of applications, case studies and techniques. From Early Palaeolithic dispersals (that’s me! but also another paper by Dario Guiducci, Ariane Burke, James Steele which I’m really looking forward to) to Tierra del Fuego societies to  sea faring in Oceania to modelling 17th century Polish epidemics – you get 12 hours (!!!) of Agent-based Modelling, Network Analysis, Neural Networks and even a few theoretical papers. You can find the abstracts here:  S25. Agents, Networks, Equations and Complexity.

(S23) Session: Modelling approaches to investigate population dynamics and settlement patterns over the long term

Another giant session, thankfully not overlapping with S25. Focused on population  dynamics, settlement patterns and land use this session takes a leap forward from the traditional static, GIS approaches and looks for more dynamic modelling techniques such as simulation.

(S24) Session: Modelling approaches to study early humans in space and time

I had a pleasure to participate in this session at the CAA2013 in Perth and it was a fantastic combination of papers showcasing various techniques (databases, least-cost path analysis, ABM) used to approach the topic of mobility in prehistory.

(S20) Session: (Re)building past networks: archaeological science, GIS and network analysis 

Network analysis seems to be getting a strong hold in archaeological computing. This session shows a few of the most common applications (inter-visibility, transport/trade, connectivity of islands) as well as some new ideas.

Satellite Event: The Connected Past

On Saturday, the Connected Past team will hold a satellite conference on Network Analysis in archaeology. You can find their call for papers and all the details here: The Connected Past.

Grand Challenges in Archaeology, or, How archaeology (and complexity) can save the world

photo by Nate Crabtree photography

Few issues beg for urgent attention more than the possibility that the Earth cannot support continued population growth and increased use of limited natural and energy resources. But how common has been the problem of societies outgrowing the resources available to them given their technical capacities? Why do societies collapse? (Kintigh et al. p. 7)

As a species we have become increasingly aware of the great dangers that are facing us. Climate change, political unrest, food scarcity… these issues are talked about daily in the press. And like the threat of nuclear winter during the Cold War these issues provide an ever-present (and rather unpleasant) umbrella for a lot of discourse. (Unlike the Cold War we don’t get the security blanket of practicing “duck-and-cover” drills under our desks; how does one duck and cover from a melting icecap?) Scientists are addressing these challenges by studying them, which can then help inform policy. But since these issues have only recently been undertaken with scientific rigor, how can we truly understand human groups’ reactions to such challenges?

Enter archaeology. While many lay-people associate archaeology with excavation (which we do), or Indiana-Jones style adventures (which we rarely have), archaeology is rapidly moving to become a field that uses the study of past societies to help anticipate future challenges. To do this, archaeologists have begun using sophisticated techniques developed in the hard sciences to address our unique suite of problems. For example, instead of looking at a pot and describing its artistic characteristics, we now can analyze this pot for how it was interconnected in a larger sphere of exchange, inferring the social relationships that came with the trade of this pot.

In a recent article from a working group at the Santa Fe Institute, Kintigh et al. describe what they see as the biggest issues that archaeology can uniquely address. They list five main categories:

  1. Emergence, Communities and Complexity
  2. Resilience, Persistence, and Collapse
  3. Movement, Mobility, and Migration
  4. Cognition, Behavior, and Identity
  5. Human-Environment Interactions.

Each of these five points sounds like a talking point for a modern UN conference on world affairs.

By using complexity science, archaeologists can use the rich record of the past to better understand, for example, why cities collapse, or how large populations in the past dealt with food scarcity. While these societies were not as technologically advanced as what we have now, the underlying structure of human groups is sufficiently similar that this knowledge can potentially help us to anticipate key events.

Kintigh et al. don’t provide a road map of how to accomplish this research, but use this as a call-to-arms to mobilize our efforts. Likewise, the article does not provide a theoretical framework for complexity approaches in archaeology, but provides a unifying thread of what research questions may be the most important for archaeology to address.

Through careful approaches of agent-based modeling, analyzing the networks of human interactions, and critical approaches to understanding the scalar nature of many systems (power-laws, increasing returns to scale), as well as many other approaches from complexity science, we can use the past to better understand our future.

Kintigh et al’s article, Grand Challenges For Archaeology, will appear in American Antiquity 79(1) 2014 in the Forum section and a shorter version on PNAS here.

–Stefani Crabtree

Student Conference on Complexity Science

Student Conference on Complexity Science run this year by the Institute for Complex Systems Simulation at the University of Southampton has just released an open call for sessions. Having participated in the last two editions I can happily vouch for it. Originally designed as a forum for the students of UK doctoral training centres in Complexity Science (Bristol, Southampton and Warwick) it grew to become a truly international and interdisciplinary event. It is one of those gatherings where pretty much everyone is out of their depth – 250 students spread through every imaginable field of science from the most hardcore maths to the soothing storytelling of humanities. The program looks exciting with the first day devoted to hands-on workshops, a few high-level keynotes and plenty of opportunities to drink coffee and wine and chat with likeminded PhD student. Not to mention Brighton is probably the ‘hipsterest’ of hipster towns. See you all there in August!

See their CfS below.

We are pleased to announce that the Call for Sessions at the 4th Student Conference on Complexity Science (SCCS2014) is now open.

The Conference

The Student Conference on Complexity Science (SCCS) is the largest UK conference for early-career researchers working under the interdisciplinary framework of Complex Systems with a particular focus

© Tessa Coe from the series 'complexity'. http://www.tessacoe.co.uk
© Tessa Coe from the series ‘complexity’. http://www.tessacoe.co.uk

on computational modelling, simulations and network analysis. Since 2010, this conference series has brought together PhD students and early career researchers from both the UK and overseas interested in areas ranging from quantum physics to the economics of happiness. The 4th Student Conference on Complexity Science will be held between 19-22 August 2014 at the University of Sussex, Brighton, UK.

If your work comes under the general topic of complexity science, we want to hear from you!

The Format
© Tessa Coe from the series ‘complexity’. http://www.tessacoe.co.uk

The conference will consist of 4 hands-on workshops and 10 parallel sessions of 6-9 papers. We particularly encourage sessions crossing traditional disciplinary boundaries and/or dealing with the general theoretical underpinnings of complexity science. The general themes of the conference include but are not limited to:

  • Complex Systems Simulation in Physics, Engineering and Mathematics
  • Complexity Theory
  • Artificial Intelligence and Swarm Robotics
  • Network Science
  • Complexity Science applications in Earth and Social Sciences, Digital Economy and Digital Humanities
The Logistics

To submit a session please provide a short abstract and a list of potential speakers at http://sccs2014.soton.ac.uk/sessions.php. The list of chosen sessions will be announced on 17th February 2014.

We are pleased to announce that thanks to the generosity of the Institute for Complex Systems Simulation we will be able to offer a number of travel bursaries to the attendees.

The Deadlines
  • Call for Sessions closes: 12 p.m. (UTC) 27th January 2014
  • Sessions announced and Call for Papers opens: 17th February 2014
  • Call for Papers closes: 12 p.m. (UTC) 26th May 2014
  • Papers announced: 16th June 2014

You can  on Twitter.
Or like our Facebook page.

Why Simulating Complexity?

The Tale of Complexity Science

In the old science things were simpler. As a scientist you would look at a problem, dissect it into components and then research them one by one. The general belief held that the more you know about the components the better you understand the system and hence the closest you are to solving the problem. And so we run in a similar vein for a few hundred years and we were doing just fine.
Until a few decades ago, when more and more voices started uttering uncomfortable questions and pointing out to some perplexing results of their research. Normally, you could dismiss pretty much any piece of research which gives counterintuitive results but this time the fussers were using two things that are really difficult to dispute: maths and computers. And so, as Thomas Kuhn would call it, a paradigm shift started to happening.

The Theory of Complexity

The definition of complexity is multidimensional and often varies adjusting itself to the interests and assumptions of a given scientific field. There are, however, a few core elements that seem to characterise complex systems regardless if they’re made out of particles, neurone, mice or men. You can start with what it says on the tin: complexity science deals with complex systems. And a complex system is a system consisting of a large number of independent components which interact with each other and sometimes also with the environment, a good example of which is a society, an ant hill or a brain. So far, this is in line with what we called the ‘old science’ view of the world but there is a key difference. In the complexity science view the interactions between the system’s elements, even if following very simple rules, can lead to large-scale complex patterns which would be difficult to predict just by looking at the individual components.

A great example of which is a flock of birds – it is a magnificent show involving hundreds of birds creating one multi-element “organism” able to effectively escape predators, travel thousands of miles but also, on a cheerful note, turn the Italian capital into a smelly pile of bird’s poo. It may come as a surprise, but bird flocks are leaderless and it was shown that the individual birds need to follow only three rules to form a flock: 1. align with nearby birds; 2. adhere to the nearby birds; 3. but try to avoid collisions (don’t believe? check out the simulation for yourself: NetLogo Flocking).

Other commonly quoted examples of complex systems include: human brain, global economy, cities, ecosystems etc. They often have a great ability to adapt rapidly to the constantly changing circumstances but they’re also prone to very dramatic changes if a specific threshold is crossed (think economic crashes). All this happens without a centralised control, a leader and often without any influence of exogenous factors (you may be mortified to learn that trafic jams can occur with no reasons whatsoever).

Complexity Science and Archaeology

The main two tools of complexity science are simulations and network analysis. A number of successful applications of both of those techniques to archaeological case studies showed the great potential of this framework. This includes a great case study of growing the artificial society of the Anasazi, getting to better understand early hominid’s behaviour and food sharing or the numerous simulations of the Neolithic wave of advance. However, at the moment complexity science in archaeology is undergoing its adolescence growth spurt with the number of new applications growing exponentially. Being still in its formative phase it is vulnerable to uncritical applications and it’s lacking a solid theoretical framework although its potential is being more and more widely recognized.

Screen Shot 2013-05-11 at 21.08.23

Network created by agents moving between specific points in the landscape


Complexity science offers an exciting opportunity to bring archaeology closer to more quantitative approaches as the tools it uses enforce binary description of the system. It is also a fantastic playground for testing out new (and old) models. Luke Premo called the simulation environment a ‘behavioural laboratories’ in which one can “eliminate plausible scenarios that are nevertheless unlikely to have occurred” ( Premo 2006, 108). Finally, complexity science tools are the only way in which one can fully understand the nature of complex systems of which past human societies are definitely one.

From the world of Complex Systems Simulation in Humanities