This year’s Conference on Complex Systems, which took place in Cancun, Mexico, was the second CCS conference outside of Europe and the first in Latin America. The conference reflected the great and under-appreciated talent across the Americas—with, for example, plenary talks by Mario Molina, a Nobel Laureate from Mexico; Venezuelan Marta González, a highly respected researcher of patterns of human mobility; and César Hidalgo from Chile, who is a major talent in the field of collective dynamics. Many more researchers from all over Latin America were able to participate in the conference through parallel sessions and posters. 

Some of the most interesting talks were those by Luís Bettencourt on scaling laws in cities (“Cities as Open-Ended Complex Adaptive Systems”), by Marta González on modeling human mobility (“A Complex Systems Approach to Study Human Mobility”), and by Peter Fennell on reducing Simpson’s Paradox (“Statistical Methodology Uncovers Feature Importance and Correlations in Human Behavioral Data”). Each of these talks demonstrated how human behavior is not a phenomenon reserved for “soft” science, but is also an object of fascination by physicists and those in other “hard” fields. Physicists, however, may gloss over the diversity of human behavior — humans are not as simple and uniform as an atom — so care must be made when modeling data, as demonstrated in the talk by Peter Fennell. 

In Luís Bettencourt’s talk, he discussed that wealth creation and other socioeconomic output, such as wages, increase super-linearly with a city’s population (it increases as N × 7/6, where N is the population), while infrastructure, such as the miles of roads, increases sublinearly (it increases as N × 5/6). Therefore, cities achieve both greater wealth accumulation and economies of scale, and their features have various non-obvious scaling exponents. Deviations from the mean value of a conservation law suggest ways to improve city efficiency, and city properties, such as non-uniform spatial distributions of electricity, can increase social disparity. Not only are facets of cities described by unexpected physics-like scaling laws, but the findings have clear implications about city planning.

Marta González looked deeply into the patterns of human mobility, briefly discussed in Luis’ talk, to better understand both how to predict human behavior (in order to increase infrastructure efficiency and energy consumption) but, even more fundamentally, how to understand whether simple laws can explain the seemingly complex behavior of humans. Through a method, called "TimeGeo", developed by Marta and her collaborators, one can successfully model both where and when people arrive down to a few minutes. This may seem almost Orwellian, but recall that humans have variance that even this model cannot completely capture. The overall patterns, however, can be used in the same way that correctly timed green lights can dramatically reduce gridlock.

Finally, Peter Fennell discussed an underappreciated aspect of social science: that there is a wide variance in how people behave. (In other words, people do not just “think”, but “think different”, to quote IBM and Apple slogans, respectively.) The consequence of this simple notion is that aggregating data, such as that on gender biases in the UC Berkeley graduate schools, creates a statistically significant conclusion — e.g., that the graduate schools may favor male over female applicants — even though, when the data are split more finely, for example by department, one may find the opposite conclusion. This property is known as Simpson’s Paradox, and it demonstrates how important conclusions can be lost in a sea of other data. Given the number of parameters that a data set may have, researchers often suffer from the curse of dimensionality, in which those important morsels of data can be hidden among the many potential ways that the data can be split up. Therefore, Peter Fennell and Kristina Lerman are creating an automated way to find how to best split data such that Simpson’s Paradox can be reduced. They pointed to examples where their method searches Twitter data and finds whether users adopt novel hashtags (e.g., “#GameDay”) via social influence or other probable ideas. Unlike much previous research, they find that social influence had a very small effect on user behavior.

The conference was great! The only significant limitation of this conference was the lack of incentive for researchers to view the nearly 200 conference posters. Natural places where researchers might congregate, such as at snack bars, were far from the poster room, and the room itself lacked places to sit down. Therefore, almost all networking was done far from the posters themselves, and one would have to isolate themselves to view the posters. That said, the talks and posters were very interesting, and demonstrate the field of complexity is both growing and filled with researchers with great aptitude.

In conclusion, the Conference on Complex Systems had fascinating talks that demonstrate the importance of complex-systems research in our everyday lives. Furthermore, it demonstrates the goals of SIAM, with mathematics used to both understand and directly help our society.