Krasnow Institute for Advanced Study

Permanent URI for this collection

https://hdl.handle.net/1920/6554

The Krasnow Institute seeks to expand understanding of mind, brain, and intelligence by conducting research at the intersection of the separate fields of cognitive psychology, neurobiology, and the computer-driven study of artificial intelligence and complex adaptive systems. These separate disciplines increasingly overlap and promise progressively deeper insight into human thought processes. The Institute also examines how new insights from cognitive science research can be applied for human benefit in the areas of mental health, neurological disease, education, and computer design.

Browse

Now showing 1 - 20 of 24

Essays on the Drivers of Political and Ideological Extremism
(2016) Alizadeh, Meysam; Alizadeh, Meysam; Cioffi-Revilla, Claudio
The problem of interest in this dissertation is the phenomenon of drifting toward opinion extremes. The process is called Radicalization and has received a great deal of attention in social psychology and sociology from its inception. Although it is extremism of behavior that is of greatest interest, it is important to study opinion extremism since it has been shown to be a plausible preceding step of violent extremism. One of the longstanding quests in counter-radicalization studies is to know what drives extremists to adopt extreme ideologies. However, it is unlikely that extremists will volunteer for experimental studies.
A General Social Agent-Based Model of Opinion Dynamics with Applications to STEM Education and Radicalization
(2016) Harrison, Joseph Francis; Harrison, Joseph Francis; Cioffi-Revilla, Claudio
Many aspects of our society are affected by how opinions change and ideology spreads (e.g., interest in STEM and political radicalization), but the underlying processes are not well understood. Previous attempts at modeling these phenomena have suffered from a lack of empirical data and/or insufficient grounding in social-psychological theory. Moreover, the field of opinion dynamics would benefit from a broader view of the discipline that captures the commonalities between different domains.
Agents in Conflict: Comparative Agent-Based Modeling of International Crises and Conflicts
Masad, David P.; Masad, David P.; Axtell, Robert
Inter-state conflicts are a key area of study in international relations, and have been approached with a variety of techniques, from case studies of individual conflicts, to formal analysis of abstract models and statistical investigations of all such conflicts. In particular, there are a variety of theories as to how states make decisions in the face of conflicts -- such as when to threaten force, when to follow through, and when to capitulate to an opponent's demand. Some scholars have argued that states may be viewed as rational decisionmakers, while others emphasize the role of psychological biases affecting individual leaders. Decisionmaking is challenging to study in part because of its complexity: the decisionmakers may not just be individuals but organizations, following internal procedures and reflecting institutional memory. Furthermore, the decisions are often believed to be strategic, reflect- ing the decisionmakers' anticipation of multiple other actors' potential responses to each possible decision. In this dissertation, I demonstrate that agent-based models (ABMs) provide a powerful tool to address this complexity, and advance their use as a bridge between different method- ologies. Agents in ABMs can be used to represent countries and endowed with a variety of internal decisionmaking models which can operationalize a variety of theories drawn from case studies, psychological experiments or game-theoretic analysis. The specific decision model agents utilize may be changed without altering the sub-models governing how the agents interact with one another. This allows us to simulate the same overall interactions utilizing different decisionmaking theories and observe how the outcomes differ. Furthermore, if these interactions correspond to real-world events, we may directly see how much explanatory or predictive power the outputs of the model variants provide. If one variant's outputs correspond closer to the empirical data, it provides evidence supporting that variant's underlying theory. I implement two agent-based models, extending well-established prior models of international conflict: the International Interaction Game (Bueno de Mesquita and Lalman, 1992) and the Expected Utility Model (Bueno de Mesquita, 2002). For each, I start with their original agent decisionmaking models, and develop several variants grounded in relevant theories. I then instantiate the models with historic, empirically-derived data and run them forward to generate sets of simulated outcomes, which I compare to empirical data on the relevant time periods. I find that non-rational models of decisionmaking in the International Interaction Game provide similar explanatory power to the purely rational model, and yield rich satisficing behavior absent in the original model. I also find that the Expected Utility Model variant implementing a Schelling (1966)-inspired model of coercion yields richer dynamics and greater explanatory power than the original model. In addition to providing evidence in support of particular theories and hypotheses, this work demonstrates the power of the comparative modeling methodology in studying international conflict. Future work will involve adding more statistical controls to the model output analysis, comparative analysis between the outputs of the two overall models, and extension of the decisionmaking models for each. The same methodology may also be expanded to other formal and computational models of international relations, and social science more broadly.
An Essay on Micro Heterogeneity and the Evolution of Inequality
Shin, Hyungsik; Shin, Hyungsik; Axtell, Robert
The level of inequality has increased over the past several decades and reached at the historically high level in many developed countries. Yet, the traditional theory of supply and demand in labor has failed to elucidate the emergence of prevailing inequality from bottom up. In fact, its scope on inequality has been limited and overlooked the emergence of inequality through interactions of heterogeneous sub-systems in micro level. In this dissertation, therefore, I have used Agent-based Models as an alternative to traditional methods to understand the evolution of inequality given micro heterogeneity and bridge the gap between the theory and real world. In the first two chapters, I have studied how behaviors of ill-motivated heterogeneous individuals cause inequality to rise and severe crises. In the first chapter, it has focused to compare and contrast macro behaviors such as inequality and growth when the compen- sation system is vulnerable to free riding or not in an otherwise identical economy. The model results have demonstrated that if the compensation is ill-defined, individuals became selfish "free riders" and the economy became volatile hence experienced constantly rising inequality followed by crises. In the second chapter, it has examined the effect of such ill-defined compensation system on inequality and growth via consumption channels -- demand-driven recession. The simulation outcomes have revealed that even a mild income concentration in the hands of a few in the beginning could cause a terrible economic recession eventually because such an income squeeze led a decrease in consumption of the poor and middle class which in turns, drags down a whole economy. It has confirmed the wide-spreading belief that if the poor and middle class cannot afford to buy goods and services, it causes a collapse of an economy eventually. In the last chapter, I have explored that when information -- characteristics -- about goods is the object of preference rather than goods per se, contrast to the traditional theory of supply and demand, how such a change in the preference object affects market equilibrium and inequality in a barter-based economy. The simulation results have demonstrated that, depending on the ratio of characteristics of goods, even after finitely many numbers of trades, MRSs of goods do not necessarily converge to the optimal level. In other words, many agents are not content with their after-trade endowments but can't be helped because existing goods do not suit them to be Pareto optimal. When it comes to inequality, GINI coefficient, an index for measuring the level of inequality is limited to capture dynamics of the change in wealth over trades within a period. Thus, it is not clear that the change in the object of preference affects the evolution of inequality over trades as well as over time.
A Computational Social Science Approach to the Social Determinants of Cancer
(2016) Metgher, Cristina; Metgher, Cristina; Axtell, Robert
Cancer is a complex system of systems - in which heterogeneous actors interact dynamically with each other and with the environment across time. It is the second leading cause of death in the U. S. after cardiovascular disease, striking people from all occupations and backgrounds. The U. S. President Nixon declared the “War on Cancer” in 1971. Since then, much progress has been made towards understanding this disease. However, due to the reductionist thinking that has been predominant in cancer research for the last decades, the mortality rates due to cancer have hardly changed.
Explaining Box Office Performance from the Bottom Up: Data, Theories and Models
(2016) Russo, Holly Ann; Russo, Holly Ann; Axtell, Robert
Every week, there are more than 50 movies playing in theaters from which movie-goers can choose. Analyses of the relative box office success of these films shows that it is Pareto-distributed, with roughly 20% of them earning 80% of the overall revenue. Arthur De Vany studied the potential causes of this ‘winner-take-all’ distribution through equation-based analyses, and theorized that the Pareto-distributed box office revenues we observe emerge from the micro-level complex adaptive behavior of movie-goers with imperfect information.
Social Preferences, Learning, and the Dynamics of Cooperation in Networked Societies: A Dialogue Between Experimental and Computational Approaches
(2016) Cotla, Chenna Reddy; Cotla, Chenna Reddy; Axtell, Robert L
In this dissertation, I empirically investigate cooperative behavior in networks using the framework of network public goods games. To do so, I use a dialogue between behavioral experiments and agent-based models. I design and conduct behavioral experiments to generate data to construct boundedly rational agents that behave like humans and reproduce stylized facts in public goods environments. The human-like agents are deployed in a small-scale agent-based model to make novel quantitative predictions that can be statistically tested using a new set of behavioral experiments. This ensures that the behavioral specification of agents carries predictive value so that quantitative predictions made using it can be reproduced with human subject experiments. The high fidelity agent-based model is then extended to study the dynamics cooperation in networked environments. The dissertation is organized into three chapters.
An Agent Based Model of Community Authority Structure Resilience
(2016) Mcfarlane, Hugh James; Mcfarlane, Hugh James; Cioffi-Revilla, Claudio
This dissertation presents a theoretical model based on social exchange theory that explains the resilience and adaptation of authority structures in urban communities. Communities where non-state actors undermine or replace government institutions are a persistent public policy concern in many cities. The structure of instrumental relationships between authorities and residents in these communities is a key variable associated with a wide range of human security and governance challenges. Altering these structures is often necessary to enable other public policy goals. However, there is an absence of theoretical frameworks that address the dynamic characteristics of these structures. This hinders policy development by limiting insights into the effects of efforts to support or undermine particular groups or to alter social conditions on authority structures. The theory developed here describes authority structures as an emergent feature of a community-level complex adaptive social system. In this system, individual actors select relationship partners based on past experiences, preferences, environmental conditions, and information from other actors. Changes to structure are the result of changes to the set of actors in the system and how these actors value particular relationships. A comparative case study of three sub-Saharan African communities located in Nairobi, Cape Town, and Lagos and several additional experiments are performed using an agent-based model implementing this theory. The results demonstrate the practical application of the theory to public policy analysis and support the choice of social exchange theory as the basis for actor decision-making. More broadly, this effort extends existing theoretical and agent-based models of contentious polities and authority structures. It also demonstrates the utility of computational modeling in advancing the research programs in social exchange theory, political authority, and comparative urban politics.
Implementing a Complex Social Simulation of the Violent Offending Process: The Promise of a Synthetic Offender
(2016) Dover, Thomas J.; Dover, Thomas J.; Cioffi-Revilla, Claudio
There are limitations to traditional methods of capturing the dynamics of violent interactions. These limitations are due to outcome driven approaches, data sampling issues, and inadequate means to capture, express, and explore the complexity of behavioral processes. To address these challenges, it is proposed that “violent offending” be re-framed as an emergent feature of a complex adaptive social system. This dissertation abstracts and computationally implements a theoretical framework that forms the basis of a complex social simulation of the violent offending process. The primary outcome of this effort is a viable synthetic offender that emerges from simulated interactions between potential offenders (subjects) and potential victims (targets) within an environment. The results of calibrating this model to a real-world murder series are discussed, as well as, the comparison metrics used to assess goodness-of-fit of simulated and real-world event-sites. A synthetic offender promises valuable insights into individual offending trajectories, offender tactical processes, and the emergence of geospatial and temporal behaviors. Furthermore, this approach is capable of reproducing the violent offending process with sufficient detail to contribute new scientific understanding and insights to criminology and the social sciences.
Towards Emergent Social Complexity
(2015) Rouly, Ovi Chris; Rouly, Ovi Chris; Axtell, Robert L.; Crooks, Andrew
Complexity science often uses generative models to study and explain the emergent behavior of humans, human culture, and human patterns of social organization. In spite of this, little is known about how the lowest levels of human social organization came into being. That is, little is known about how the earliest members of our hominini tribe transitioned from being presumably small-groups of ape-like polygamous/ promiscuous individuals (beginning perhaps as early as Ardipithecus or Australopithecus after the time of the Pan-Homo split in the late Pliocene to early Pleistocene eras) into family units having stable breeding-bonds, extended families, and clans. What were the causal mechanisms (biological, possibly cognitive, social, and environmental, etc.) that were responsible for the conversion? To confound the issue, it is also possible the conversion process itself was a complex system replete with input sensitivities and path dependencies, i.e., a nested complex system. These processes and their distinctive social arrangements may be referred to favorably (as one notable anthropologist has called them) as, “the deep structure of society.” This dissertation describes applied research that used discrete event computer modeling techniques in an attempt to model-then-understand a few of the underlying social, environmental, and biological systems present at the root of human sociality; at the root of social complexity.
Delay and traffic rate estimation in network tomography
(2015) Etemadi Rad, Neshat; Etemadi Rad, Neshat; Mark, Brian L.; Ephraim, Yariv
Network tomography deals with estimation of computer network features from measurements on links or terminal nodes. The area was pioneered with the work of Vanderbei and Iannonou in 1994 and Vardi in 1996. Of particular interest are estimation of source-destination traffic rates from link packet counts or from aggregated packet counts in input and output nodes, and estimation of link delay from source-destination delay measurements. Traffic rate estimation, and link propagation delay estimation, are inverse problems which require the solution of under-determined sets of linear equations. Iterative solutions based on moment matching and the expectation-maximization algorithm were proposed for traffic rate estimation, and a maximum entropy approach was developed for link propagation delay estimation. Traffic rate estimation was also performed using a Bayesian estimation approach. Estimation of link delay densities commonly involves exponential mixture models which entail independence of the delay on various links. Network tomography is useful for monitoring the performance of a network, and thus maintaining and expanding the network.
Individual and Social Learning: An Implementation of Bounded Rationality from First Principles
(2015) Palmer, Nathan Michael; Palmer, Nathan Michael; Axtell, Robert L.
This dissertation expands upon a growing economic literature that uses tools from reinforcement learning and approximate dynamic programming to impose bounded rationality in intertemporal choice problems. My dissertation contributes to the literature by applying these tools to the canonical household consumption under uncertainty problem. The three essays explore individual and social approaches to learning-to-optimize and how these may be brought to data.
Climate Change and the Potential for Conflict and Extreme Migration in the Andes: A Computational Approach for Interdisciplinary Modeling and Anticipatory Policy-Making
(2015) Magallanes, Jose Manuel; Magallanes, Jose Manuel; Cioffi, Claudio
I present an agent-based model to support the thesis that extreme migration and social conflict can emerge by simply extending the current social and natural conditions, and by replicating simple mechanisms at the individual level. To carry out this work, every available information on water supply and demand has been collected and organized using official data sources, producing a baseline dated in 2011; and the basic demographics of the population has been implemented using the last official census (2007). Based on these data, many computations have been made to find, calibrate and represent the trends in population growth and water balance. For the basic mechanisms at the individual level, field work guided both by theoretical considerations and ethnographic findings has been done. A key assumption on the processing of information, not identified from the field work, has been introduced via a Bayesian belief updating mechanism.
Modeling Adaptive Economic Agents With PID Controllers
(2015) Carrella, Ernesto; Carrella, Ernesto; Axtell, Robert L.
I provide here a counterpoint to the rational agents that dominate economics: rather than adding rigidities and information limits on an otherwise classical feed-forward agent, I build a new feed-back agent that achieves equilibrium without knowledge of the model or the market it is in.
When People Rebel: A Computational Approach to Violent Collective Action
(2014-08) Pires, Bianica; Pires, Bianica; Crooks, Andrew T.
Why an individual rebels, why an individual joins collective action, and how that manifests to violence are not new questions, but are questions that continue to pose a significant scientific challenge. It is the role of the conflict analyst to answer these questions by exploring the underlying dynamics, interactions, and individual behaviors of the conflict. Violent collective action, a subfield of conflict studies, is a complex system, consisting of individuals with unique attributes that interact with other individuals through interconnected networks on a heterogeneous environment. In order to represent a complex system, we must model it from the "bottom-up," as the only way to generate the macro-behaviors is by modeling the individual, micro-level components of the system. In its ability to model complex systems, a computational approach is ideal. While various computational models have explored the use of agent-based modeling (ABM), social network analysis (SNA), and geographic information systems (GIS) in the field of violent collective action, most have explored the techniques in isolation. The models presented in this dissertation build on the value of integrating these approaches. Computational methods (i.e., ABM, SNA, and GIS) are used to develop three instantiations of more general models of violent collective action. The instantiations, or case studies, were selected for their diversity in terms of geographic location, temporal and spatial scale, and the political and cultural issues underlying the violent collective action. In addition, the case studies serve as building blocks; as I add layers to the environment, develop more sophisticated cognitive frameworks, and create agent-to-agent and agent-to-environment interactions that more closely represent reality. In addition, with the final case study I will demonstrate the value of integrating the three computational methods. Using empirical data for which to create the modeling world and inform the agents, qualitative agreement with actual events modeled are sought. The research question this dissertation addresses is: Can a bottom-up approach provide us with useful insight into the formation, spread, and strength of violent collective action? By covering a variety of different situations of violent collective action while building on the complexity of each computational technique used, the use of a computational approach to gain a better understanding violent collective action is given greater legitimacy. Through such understanding, this dissertation contributes to the existing body of knowledge on the topic of violent collective action.
Using Social Media Content to Inform Agent-based Models for Humanitarian Crisis Response
(2014-05) Wise, Sarah; Wise, Sarah; Crooks, Andrew T.
Crisis response is a time-sensitive problem with multiple concurrent and interacting subprocesses, applied around the world in a wide range of contexts and with access to varying levels of resources. The movement of individuals with their shifting patterns of need and, frequently, disrupted normal support systems pose challenges to responders trying to understand what is needed, where, and when. Unfortunately, crises frequently occur in parts of the world that lack the infrastructure to respond to them and the information to inform responders where to target their efforts. In light of these challenges, researchers can make use of new data sources and technologies, combining the information products with simulation techniques to gain knowledge of the situation and to explore the various ways in which a crisis may develop. These new data sources - including social media such as Twitter and volunteered geographic information (VGI) from groups such as OpenStreetMap - can be combined with authoritative data sources in order to create rich, synthetic datasets, which may in turn be subjected to processes such as sentiment analysis and social network analysis. Further, these datasets can be transformed into information which supports powerful agent- based models (ABM). Such models can capture the behavior of heterogeneous individuals and their decision-making process, allowing researchers to explore the emergent dynamics of crisis situations. To that end, this research explores the gathering, cleaning, and synthesis of diverse data sources as well as the information which can be extracted from such synthetic data sources. Further, the work presents a rich, behaviorally complex agent-based model of an evacuation effort. The case study deals with the 2012 Colorado Wildfires, which threatened the city of Colorado Springs and prompted the evacuation of over 28,000 persons over the course of four days. The model itself explores how a synthetic population with automatically generated synthetic social networks communicates about and responds to the developing crisis, utilizing real evacuation order information as well as a model of wildfire development to which the individual agents respond. This research contributes to the study of data synthesis, agent-based modeling, and crisis development.
Computational Modeling of Climate Change, Large-Scale Land Acquisition, and Household Dynamics in Southern Ethiopia
(2013-08) Hailegiorgis, Atesmachew Bizuwerk; Hailegiorgis, Atesmachew Bizuwerk; Cioffi-Revilla, Claudio
COMPUTATIONAL MODELING OF CLIMATE CHANGE, LARGE-SCALE LAND ACQUISITION, AND HOUSEHOLD DYNAMICS IN SOUTHERN ETHIOPIA
The Blind Lawmaker
(2013-08) Koehler, Matthew; Koehler, Matthew; Axtell, Robert L.
Many have written about how the Common Law should evolve. The few attempts to demonstrate this empirically, however, have not found evidence that this evolution takes place. This study uses a representation of the Article III United States Federal Courts and an agent-based model to demonstrate that a judicial system may evolve while simultaneously emitting signals to the contrary by evolving via a punctuated equilibrium dynamic. The study then proceeds to demonstrate that agent-based modeling is a viable method for understanding the performance of judicial institutions. After reviewing concepts of jurisprudence and computational social science, the development of the model is discussed followed by a presentation of the results of the aforementioned experiments.
Coupled Dynamics of Labor and Firms through Complex Networks
(2013-08-20) Guerrero, Omar A.; Guerrero, Omar A.; Axtell, Robert L.
This dissertation bridges the gap between labor and firm dynamics through the study of complex networks in labor markets. With extensive use of large-scale employer-employee matched micro-data and agent-based modeling, we tap into the effects that networked structures (between individuals or between firms) exert in labor outcomes and employment dynamics. Some of the contributions of this work are: (i) the first characterization of a network of firms for an entire economy (connected through labor flows, i.e. labor flow networks); (ii) the study of the relationship between labor flow networks and employment dynamics; (iii) agent-based models that generate rich stylized facts about labor, firm, and social dynamics from microeconomic behavior; (iv) providing the microeconomic foundations of the formation process of labor flow networks by coupling job search models with models about the formation of complex networks. We show that the study of labor dynamics can be enriched by coupling firm dynamics. Using agent-based modeling is a natural way to deal with the heterogeneous experiences of workers and firms while maintaining a simple representation of the labor market. Despite their simplicity these models are grounded on empirical evidence obtained from large-scale micro-data and are capable of generating numerous stylized facts simultaneously. This approach has great potential for the design and evaluation of labor policies. Therefore, governments, regulators, and policy-makers would be greatly benefited from collecting large-scale labor micro-data, analyzing labor flow networks, and developing agent-based models of labor markets.
Innovation from a Computational Social Science Perspective: Analyses and Models
(2013) Casstevens, Randy M.; Casstevens, Randy M.; Axtell, Robert L.
Innovation processes are critical for preserving and improving our standard of living. While innovation has been studied by many disciplines, the focus has been on qualitative measures that are specific to a single technological domain. I adopt a quantitative approach to investigate underlying regularities that generalize across multiple domains. I use a novel approach to better understand the innovation process by combining computational models with empirical data on software development, on one hand, and the evolution of the English lexicon on the other. Innovation can be viewed as the recombination and mutation of existing building blocks. I focus on how building blocks are used to generate innovations. The building blocks are pieces of code (e.g., functions or objects) for the software development data and words for the written language. These data lie at extremes of time scales: innovation occurring over the course of a few days or a week in the case of software while language evolution occurs over decades or centuries. This allows the examination of innovation processes that range from highly-constrained to completely open-ended. Computational methods reinforce the findings from the data analyses and permit exploration of the general features of innovation processes through the construction of abstract models.

Browse

Recent Submissions