Econophysics
Many influential theories in economics can be traced back to physics in one way or another. To name a few, Option-Pricing Theory (Black & Scholes, 1973; Merton, 1973), General Equilibrium Theory (Debreu, 1959), Macroeconomic Models (Tinbergen, 1959; Klein, 1970), Game Theory (von Neumann & Morganstern, 1944; Nash, 1951), and Samuelson’s seminal Ph.D. dissertation (1947) are all anchored on physical laws. Even stronger are the parallels between physics and finance which share the same equations from thermodynamics, brownian motion, and fluid dynamics among others. Therefore, not surprisingly, this synergy between physics and economics, has led to the birth of what is now commonly known as “econophysics” (Mantegna & Stanley, 2000).
Econophysics can be regarded as a new interdisciplinary field where principles and methods from physics are applied to the solution of problems in economics. This ambitious endeavor has led to the creation of other research areas such as networks science at the Santa Fe Institute and the Harvard-MIT Observatory of Economic Complexity. However, despite the heroic triumphs of physics, economic systems are far more complex than physical systems. Economic systems are driven by interactions between people and institutions where psychology plays a leading role in the decision processes that ultimately shape their fate.
In recognition of the need for incorporating psychology in our understanding of economic systems, our research entails a multidisciplinary approach that incorporates principles from physics, computer science, psychology, statistics, engineering, and economics with the common goal of developing artificial economic systems capable of emulating real world interactions between humans.
Astrostatistics
The extraordinary quantity and quality of data available from remarkable surveys such as the SDSS, GALEX, SIRTF, HST, DEEP2, VVDS, WMAP, and the Magellan telescopes (among many others) will for the first time make it possible to carry out a comprehensive census of the matter and energy content of the Universe. This wealth of data has resurrected an interest in collaborations between astronomers and statisticians and led to the emerging field of Astrostatistics (see e.g., ASAIP). Astrostatistics, while still in its infancy, holds great potential for discovery at the intersection of Astrophysics, Statistics and Computer Science.
Astronomical data are an ideal testbed for advanced statistical and computational data mining tools because of their large volume (especially true for the expected data from the LSST), inherent complexity, diversity, and challenges they pose to scientists in uncovering hidden universal laws in these massive datasets. For astrophysicists it is common practice for us to conduct research that is highly unidirectional by specializing on cosmological probes such as e.g., galaxies, the cosmic microwave background (CMB), stars, or planets only. However, to better understand the evolution and structure of the universe we need a comprehensive approach to this endeavor. Therefore, in response to this need, our work focuses on developing statistical tools capable of integrating observations from different astronomical sources with the goal of providing a unified characterization of our cosmic puzzle.
Human Causal Reasoning
Modeling human reasoning presents daunting computational and mathematical challenges that extend beyond current technologies and understanding. However, limitations in computing resources can be overcome by sophisticated and intelligent algorithms founded on sound mathematical principles, led by a profound understanding of the underlying cognitive processes. Our current research in human (and animal) causal reasoning lies at the forefront of the promising synergy between Statistics, Cognitive Science, Neuroscience and Computer Science. A promising result thus far has been the successful learning ability of our algorithms from sequential data, and their consistent outperformance of human learners. When presented with noisy multidimensional data, our algorithms only require about 25% of the observations that humans need for establishing the correct cause and effect relations, yet they can infer the causal structures 90% of the time compared with only 60% for humans.