Agent_Zero: Toward Neurocognitive Foundations for Generative Social Science (Princeton Studies in Complexity) - Hardcover

Epstein, Joshua M.

Hardcover

ISBN 10: 0691158886 ISBN 13: 9780691158884

Publisher: Princeton University Press, 2014

Synopsis

The Final Volume of the Groundbreaking Trilogy on Agent-Based Modeling

In this pioneering synthesis, Joshua Epstein introduces a new theoretical entity: Agent_Zero. This software individual, or "agent," is endowed with distinct emotional/affective, cognitive/deliberative, and social modules. Grounded in contemporary neuroscience, these internal components interact to generate observed, often far-from-rational, individual behavior. When multiple agents of this new type move and interact spatially, they collectively generate an astonishing range of dynamics spanning the fields of social conflict, psychology, public health, law, network science, and economics.

Epstein weaves a computational tapestry with threads from Plato, Hume, Darwin, Pavlov, Smith, Tolstoy, Marx, James, and Dostoevsky, among others. This transformative synthesis of social philosophy, cognitive neuroscience, and agent-based modeling will fascinate scholars and students of every stripe. Epstein's computer programs are provided in the book or on its Princeton University Press website, along with movies of his "computational parables.?

Agent_Zero is a signal departure in what it includes (e.g., a new synthesis of neurally grounded internal modules), what it eschews (e.g., standard behavioral imitation), the phenomena it generates (from genocide to financial panic), and the modeling arsenal it offers the scientific community.

For generative social science, Agent_Zero presents a groundbreaking vision and the tools to realize it.

About this title

About the Author

Joshua M. Epstein directs the Center for Advanced Modeling at The Johns Hopkins University, where he is a professor in the Department of Emergency Medicine with joint appointments in the Departments of Applied Mathematics, Economics, International Health, Environmental Health Sciences, Biostatistics, Civil Engineering, and the Institute for Computational Medicine. He is also an external professor at the Santa Fe Institute. This volume forms a trilogy on agent-based modeling, with Growing Artificial Societies: Social Science from the Bottom Up (MIT), with coauthor Robert Axtell, and Generative Social Science: Studies in Agent-Based Computational Modeling (Princeton).

From the Back Cover

"Agent Zero offers a solution to some of social science's great puzzles. Its behavioral basis is the interplay of emotion, cognition, and network contagion effects. It elegantly explains why so many human actions are so manifestly dysfunctional, and why some are downright evil."--George Akerlof, Nobel Laureate in Economics

"Rarely has a book stimulated me intellectually as much as this one. Particularly exciting is the incorporation of agents who feel (affect) and deliberate, as well as influence one another through social interaction. Epstein is a brilliantly creative scholar and the range of applications showcased here is stunning. In sum, this is a pathbreaking book."--Paul Slovic, University of Oregon

"Joshua Epstein proposes a parsimonious but powerful model of individual behavior that can generate an extraordinary range of group behaviors, including mob violence, manias and financial panics, rebellions, network dynamics, and a host of other complex social phenomena. This is a highly original, beautifully conceived, and important book."--Peyton Young, University of Oxford

"In social science generally and most notably in economics, the rational actor model has long been the benchmark for policy analysis and institutional design. Epstein now offers a worthy alternative:Agent_Zero, a mathematically and computationally tractable agent whose inner workings are grounded in neuroscience. Much like you and me,Agent_Zero is influenced by emotion, reason, and social pressures. Epstein demonstrates that collections of Agent Zeros perform amazingly like real groups, teams, and societies and can therefore serve as the fundamental building blocks for what he calls Generative Social Science. The rational actor now has a true competitor. Agent_Zero is a major advance."--Scott Page, University of Michigan

"This is social science based on how our brains actually work. Epstein's computerized 'agents' can feel passion and fear, and can influence each other emotionally. And when they interact, we see many of the realities of social life, from the dynamics of juries to racist violence to Arab springs. A remarkable and original piece of work."--W. Brian Arthur, Santa Fe Institute

From the Inside Flap

"Agent Zero offers a solution to some of social science's great puzzles. Its behavioral basis is the interplay of emotion, cognition, and network contagion effects. It elegantly explains why so many human actions are so manifestly dysfunctional, and why some are downright evil."--George Akerlof, Nobel Laureate in Economics

"Rarely has a book stimulated me intellectually as much as this one. Particularly exciting is the incorporation of agents who feel (affect) and deliberate, as well as influence one another through social interaction. Epstein is a brilliantly creative scholar and the range of applications showcased here is stunning. In sum, this is a pathbreaking book."--Paul Slovic, University of Oregon

"Joshua Epstein proposes a parsimonious but powerful model of individual behavior that can generate an extraordinary range of group behaviors, including mob violence, manias and financial panics, rebellions, network dynamics, and a host of other complex social phenomena. This is a highly original, beautifully conceived, and important book."--Peyton Young, University of Oxford

"In social science generally and most notably in economics, the rational actor model has long been the benchmark for policy analysis and institutional design. Epstein now offers a worthy alternative:Agent_Zero, a mathematically and computationally tractable agent whose inner workings are grounded in neuroscience. Much like you and me,Agent_Zero is influenced by emotion, reason, and social pressures. Epstein demonstrates that collections of Agent Zeros perform amazingly like real groups, teams, and societies and can therefore serve as the fundamental building blocks for what he calls Generative Social Science. The rational actor now has a true competitor. Agent_Zero is a major advance."--Scott Page, University of Michigan

"This is social science based on how our brains actually work. Epstein's computerized 'agents' can feel passion and fear, and can influence each other emotionally. And when they interact, we see many of the realities of social life, from the dynamics of juries to racist violence to Arab springs. A remarkable and original piece of work."--W. Brian Arthur, Santa Fe Institute

Excerpt. © Reprinted by permission. All rights reserved.

Agent_Zero

Toward Neurocognitive Foundations for Generative Social Science

By Joshua M. Epstein

PRINCETON UNIVERSITY PRESS

Copyright © 2013 Princeton University Press
All rights reserved.

ISBN: 978-0-691-15888-4

Contents

Foreword, xi, 
Preface, xiii, 
Acknowledgments, xv, 
Introduction, 1, 
Part I. Mathematical Model, 19, 
Part II. Agent-Based Computational Model, 81, 
Part III. Extensions, 107, 
Part IV. Future Research and Conclusion, 181, 
Appendix I. Threshold Imputation Bounds, 195, 
Appendix II. Mathematica Code, 197, 
Appendix III. Agent_Zero NetLogo Source Code, 213, 
Appendix IV. Parameter Settings for Model Runs, 221, 
References, 227, 
Index, 243, 

CHAPTER 1

PART I

Mathematical Model

In this part, we specify explicit mathematical models for the emotional,deliberative, and social components of the Agent_Zero framework. Thesechoices are not cast in stone, and different components should certainly beexplored, as discussed in the Future Research section. First, however, we reviewsome underlying neuroscience of fear and its throne: the amygdala.

This review is worthwhile because the Rescorla-Wagner equations (usedfor the affective model component) do not presuppose that fear acquisition islargely unconscious, while this is a crucially important fact from a social sciencestandpoint, and the amygdala discussion demonstrates that it is a neuroscientificallysound modeling assumption. Also, important evidence of emotionalcontagion comes from fMRI studies of the amygdala, and if we didn'tknow anything about the amygdala, these images would mean very little.

Understanding, then, that unconscious fear acquisition is what we havein mind, we now discuss the elementary neuroscience of fear as prelude tothe famous Rescorla-Wagner equations of conditioning, all en route to ourmore general model of behavior in groups.

I.1. THE PASSIONS: FEAR CONDITIONING

Humans are born with a variety of innate endowments or capacities. Oneof these is the capacity to acquire fear (and other) associations through aprocess of synaptic change in which, as Donald Hebb (1949) presciently putit, "neurons that fire together wire together." That is, after certain pairingsof an initially neutral stimulus (e.g., a tone) and a stimulus that is innatelyaversive (e.g., a shock), the initially neutral stimulus will evoke the sameresponse as the innately aversive stimulus. This associative process—oftentermed conditioning—is generated by synaptic change, or "plasticity." For alucid nontechnical exposition, see LeDoux (2002). We, of course, cannot cutopen a human and observe her fear, but we can intelligently speak of a fearcircuit—a distributed neurochemical computational architecture—whoseproper functioning is of obvious evolutionary value and whose activation isstrongly correlated with physical, autonomic, and other observable symptomsof fear (e.g., freezing). Indeed, LeDoux and others have mapped thefear circuit's operation in considerable detail and have made huge stridesin explaining the observed capacity for associative fear acquisition, retention,and extinction by Hebbian plasticity and long-term potentiation at thecellular-synaptic level (LeDoux, 2002, pp. 79–80).

The same Hebbian picture is mirrored in the higher-level Rescorla-Wagner(RW) equations, which we shall employ in the affective componentof the model. These operate not at the neuronal level but at the level of theperson, or subject, where certain conditioning stimuli (the bell) become associatedwith specific unconditioned stimuli (the shock) through repeatedpairings. There is certainly an underlying mathematical theory of neuronalfunction (action potentiation and firing), of which the cornerstones are thefamous Hodgkin-Huxley model (Hodgkin and Huxley, 1952) and its relatives,notably the Fitzhugh-Nagamo (Fitzhugh, 1961) model. As suggestedearlier, one can imagine filling in the gap between the cellular-synaptic accountand the high-level RW equations with such intermediate models.This is an important scientific challenge. Here, we attempt only a crudeplausible synthesis of simple emotional, cognitive, and social components.But to begin at the beginning, let us examine some basic features of fear.

Fear Circuitry and the Perils of Fitness

A snake is suddenly thrown in your path. You automatically freeze. Why?From an evolutionary perspective, a reasonable hypothesis is that we freeze(are "scared stiff") because the predators faced by our evolutionary ancestorsused motion detection to home in on prey, and animals (i.e., species)that didn't freeze were wiped out. Animals hard-wired to freeze enjoyeda selective advantage, in other words, and have passed the relevant wiringdown as part of our genetic endowment.

Wiring: The Amygdala in a Nutshell

As LeDoux writes, "The basic wiring plan is simple: it involves the synapticdelivery of information about the outside world to the amygdala, and thecontrol of responses that act back on the world by synaptic outputs of theamygdala. If the amygdala detects something dangerous by its inputs (discussedfurther below) then its outputs are engaged. The result is freezing,changes in blood pressure and heart rate, release of hormones, and lots ofother responses that are either preprogrammed ways of dealing with dangeror are aspects of body physiology that support defensive behaviors." (LeDoux,2002, pp. 8–9). A simple depiction is given in Figure 1 for an auditorythreat stimulus.

Having classified an auditory stimulus as threatening (innately or throughconditioning), the auditory thalamus projects (emits an action potential) tothe lateral amygdala (LA) and auditory cortex, which also projects a morerefined signal to the LA. The central amygdala (CE) then activates varioussystems to produce responses, such as those shown: freezing, increases inblood pressure, and the release of various hormones. (Further responses arediscussed later.)

In somewhat greater detail, the neural mechanism of amygdala inputsand activation, and amygdala output, are conveyed in the diagrams of Figure2. Inputs are depicted in the top, and outputs are shown in the bottomdiagram (Figure 2).

The blue almond-shaped structure here corresponds to stunning micrographsof stained brain slices like the one shown below in Figure 3 (LeDoux,2008).

One essential point is that this architecture supports a critical delay betweenunconscious and conscious responses to stimuli.

Inputs: High Road and Low Road

For example, "auditory inputs reach the lateral amygdala from the auditorythalamus and auditory cortex ... These provide a rapid but impreciseauditory signal to the amygdala. Cortical inputs from the auditory andother sensory systems ... provide the amygdala with a more elaborate representationthan could come from the thalamic inputs. However, becauseadditional synaptic connections are involved, transmission is slower" (Ledoux,2007). Hence, LeDoux (2002) calls these "the low road and the highroad," as depicted in Figure 4.

I instantly freeze at the snake (low road) but then evaluate it as beinga benign garter snake (high road), not a true black mamba, for instance.While the extreme rapidity of the unconscious response is of immenseevolutionary value, we will see that, from a social standpoint, the lag betweenit and conscious appraisal is a decidedly mixed blessing.

Outputs

Continuing, "once the amygdala detects a threat, its outputs lead to the activationof a variety of target areas that control both behavioral and physiologicalresponses designed to address the threat," (Rodrigues, LeDoux,and Sapolsky, 2009, p. 294). Beyond freezing, amygdala activation inducesthe release of numerous neurotransmitters (e.g., serotonin and dopamine),increasing arousal and vigilance. Endocrine and autonomic responses arealso dramatic, "including increased blood pressure and heart rate, divertingstored energy to exercising muscle, and inhibiting digestion" (Rodrigues,LeDoux, and Sapolsky, 2009, p. 295).

The pupils dilate to allow more light to enter. The heart rate picks up,and the heart muscle contracts more strongly, driving more blood to themuscles. Contractions of selected vascular channels shift blood awayfrom the skin and intestinal organs toward the muscles and the brain.Motility of the gastrointestinal system decreases, and digestive processesslow down. The muscles along the air passages of the lungs relax,and respiratory rate increases, allowing more air to be moved in andout. Liver and fat cells are activated to furnish more glucose and fattyacids—the body's high-energy fuels—and the pancreas is instructed torelease less insulin. The reduction in insulin allows the brain to draw offa sizeable fraction of the glucose entering the bloodstream because, unlikeother organs, the brain does not require insulin in order to utilizeblood glucose. The neurotransmitter that triggers all these changes isnorepinephrine (Bloom, Lazerson and Nelson, 2001, p. 172).

For wonderful discussions, see also Darwin's The Expression of Emotionsin Man and Animals (1872). Contemporary scientific publications presentthese input-output (and additional feedback) pathways in various levels ofdetail. Highly detailed is LeDoux (2007).

On the experience, or "feeling," of fear, Öhman and Wiens (2003, p. 270)paraphrase LeDoux (1996):

The fear module is primitive in the sense that it was assembled by evolutionarycontingencies hundreds of millions of years ago to serve inbrains with little cortices. However, it now operates in a human braincapable of advanced thought, language, and the conscious experienceof emotion. Humans can talk about emotions, and they have emotionalexperiences. Awareness of an emotion not only depends on therecognition of an emotional stimulus but also originates primarily infeedback from the emotional responses that are elicited by the stimulus.For example, experiencing a racing heart when a shadow appearsfrom a dark alley contributes to the feeling of fear. In fact, in perhapsthe most classic of all classical contributions to the psychologyof emotion, William James (1884) proposed that such feedback is theemotion. To paraphrase, you feel the emotion when you experienceits effect on your body. Thus the feeling of fear is the experience of anactivated fear module.

Current research on the neurophysiology of fear shows James to havebeen remarkably prescient, despite lacking any modern tools. Very importantly,from a social standpoint, the fear circuit can be activated, and fearconditioning can occur, unconsciously.

Unconscious Activation and Conditioning

In humans, "the fear module can be activated, and fear conditioning canoccur without our conscious awareness." Indeed we need not ever becomeconscious of it. As LeDoux continues, "... unconscious operation of thebrain is the rule rather than the exception throughout the evolutionaryhistory of the animal kingdom.... And this, moreover, confers a selectiveadvantage ... if we had to consciously plan every muscle contraction ourbrain would be so busy we would probably never end up actually takinga step or uttering a sentence" (LeDoux, 2002, p. 11). Among the manydemonstrations that amygdala activation per se need not be conscious, theso-called backward masking experiments are particularly elegant.

In backward masking, "an emotionally arousing visual stimulus isflashed on a screen very briefly (a few milliseconds) and is then followedimmediately by some neutral stimulus that stays on the screen for severalseconds. The second stimulus blanks out the first, preventing it fromentering conscious awareness (by preventing it from entering workingmemory)" (LeDoux 2002). But the first still elicits the full suite of physiologicalresponses—increased heart rate, blood pressure, sweaty palms,and so forth. "Since the stimulus never reaches awareness (because itis blocked from working memory), the response must be based on theunconscious processing of the stimulus rather than on conscious experienceof it. By short-circuiting the stages necessary for the stimulus toreach consciousness, the masking procedure reveals processes that go onoutside of consciousness in the human brain" (LeDoux, 2002, p. 208). Inshort, the stimulus makes it to the amygdala by the quick and dirty "lowroad," but its arrival in working memory (the high road) never occurs.Cacioppo et al. (2007) write, "The amygdala is particularly sensitive tofear faces (Adolphs et al., 1999; Breiter et al., 1996) even when they arepresented so rapidly as to not be consciously perceived (Morris, Öhman,and Dolan, 1999; Whalen et al., 1998). For another nice discussion ofbackward masking, see Penrose, 1999.

As recent evidence of our capacity for unconscious conditioning proper,a very interesting study by Arzi et al. (2012) demonstrates that associativelearning can occur even while we are asleep.

Delayed Feelings

If we do become conscious of fear-inducing stimuli, moreover, we may doso only after the physiological responses. Only after we have ducked fromthe darting bat do we notice that our heart is pounding, and we ask, "Whoa,what the heck was that!?" The conscious experience of fear, in other words,is a brain state induced by the unconscious activation of neurophysiologicalprecursors driven by the amygdaloid complex (LeDoux, 2002, p. 208). Or,to paraphrase William James (1884), We don't run because we fear the bear.We fear the bear because we run.

Adaptive Innate Capacity

A range of stimuli will elicit this unconscious activation—we instinctivelycrouch protectively at unexpected explosions nearby or when unexpectedprojectiles dart at our heads. In other words, certain sensory inputs will innatelygenerate the threat response. In rats, for example, cats are in this setof innate threats. In fact, rats bred in colonies completely isolated from catsfor many generations will freeze upon first exposure to cat urine (LeDoux,2002, p. 4).

Notice, however, that animals equipped only with a fixed set of specificthreats would be vulnerable to novel ones. So, it would be advantageous ifthe set could be expanded to include novel threats. And it obviously can.Pleistocene man never encountered a BMW, but we freeze when a car whipsaround the corner at us, just as he froze when huge animals charged suddenlyfrom the tall brush. We are harnessing the same innate fear-acquisitioncapacity—the same innate neurochemical computing architecture. Miraculously,synaptic plasticity permits us to adapt the evolved machinery to encodenovel threats. Detailed neurochemical accounts are given in LeDoux(2002, pp. 89–90).

Retention

There is little point is learning to fear hippos on Monday and then forgettingto on Tuesday. So, the retention of acquired fear associations is obviouslyessential in such cases and is achieved by various forms of long-termpotentiation (LTP) at the synaptic level. This is also becoming understoodneurochemically and is treated in Bauer, Schafe, and LeDoux (2002). Likeunconscious fear acquisition, this fear retention is also significant socially,as we will discuss later in connection with "extinction."

Observational Acquisition

Finally, it would also be advantageous if one could condition on the aversiveexperience of others—if you could acquire fear of the red-hot stoveby watching me get burned, without having to get burned yourself. As wewill review, this too is possible. Indeed, so-called mirror neurons may haveevolved for this very purpose (see the discussion on p. 62). The result is thatfear is, in a defensible sense, contagious (Hatfield, Cacioppo, and Rapson,1994). This will be further discussed shortly.

All in all, then, as LeDoux observes, "It is a wonderfully efficient way ofdoing things...." Rather than create a separate system to encode each newdanger, "just enable the [single] system that is already evolutionarily wiredto detect danger to be modifiable by experience. The brain can, as a result,deal with novel dangers.... All it has to do is create a synaptic substitutionwhereby the new stimulus can enter the circuits that the pre-wired onesused" (LeDoux, 2002, pp. 6–7).

Perils of Fitness

It is indeed a most wonderful machinery. But it is also terrible: it makes usdeeply vulnerable to the unconscious construction and retention of racial,ethnic, and other fears and biases [on race, LeDoux (2003); Telzer et al.(2012); on racial face masking, Öhman (2005)]. It predisposes us to rash,often violent, overreactions and opens us to all manner of nefarious manipulation.Indeed, fear conditioning has been a fundamental tool in most propagandasince time immemorial. But, equally disturbing, fear can spreadin a completely decentralized manner, propelling mass violence, financialcrises, and deeply misguided health behaviors for example. See LeDoux(2002, p. 124):

As Pavlov suspected, defense conditioning plays an important rolein the everyday life of people and other animals. It occurs quickly(one pairing of the neutral and aversive stimulus is often sufficient)and endures (possibly for a lifetime.) These features have no doubtbecome part of the brain's circuitry due to the fact that an animal usuallydoes not have the opportunity to learn about predators over thecourse of many experiences. If an animal is lucky enough to surviveone dangerous encounter, its brain should store as much about theexperience as possible, and this learning should not decay over time,since a predator will always be a predator. In modern life we sometimessuffer from the exquisite operation of this system, since it isdifficult to get rid of this kind of conditioning once it is no longerapplicable to our lives, and we sometimes become conditioned to fearthings that are in fact harmless. Evolution's wisdom sometimes comesat a cost." [Emphasis added.]

(Continues...)

Excerpted from Agent_Zero by Joshua M. Epstein. Copyright © 2013 Princeton University Press. Excerpted by permission of PRINCETON UNIVERSITY PRESS.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.

Edition details

• Publisher: Princeton University Press
• Publication date: 2014
• ISBN 10: 0691158886
• ISBN 13: 9780691158884
• Binding: Hardcover
• Language: English
• Number of pages: 272