The Department of Logic & Philosophy of Science present

"Social Dynamics Conference"

Friday-Sunday, March 15-17, 2013
8:30 a.m.-3:00 p.m.  *All talks held in the Beckman Center "Board Room"
Arnold and Mabel Beckman Center, 100 Academy Drive, Irvine, CA 92617

Beckman Center Floor Plan

Map to Beckman Center

This conference will focus on applying game theory and evolutionary game theory to the formation of conventions, cooperation, signaling and the social contract. The area spans philosophy, economics, political science, and biology.

Brian Skyrms, Distinguished Professor of Logic and Philosophy of Science at UC Irvine and Professor of Philosophy of Stanford University (Fellow of the National Academy of Sciences)
Simon Huttegger, Associate Professor of Logic and Philosophy of Science at UC Irvine


J. McKenzie Alexander, Carl Bergstrom, Cristina Bicchieri, Rosa Cao, Christof Hauert, Ed Hopkins, Gerhard Jaeger, Jorge Pacheco, Rory Smead, Peter Vanderschraaf, Elliott Wagner, Kevin Zollman

Discussants: Justin Bruner, Greg McWhirter, Hannah Rubin, Vitor Vasconcelos

Schedule of *Talks:

Thursday, March 14, 2013

7:00pm: Welcome Dinner: Las Brisas Restaurant, Laguna Beach

Friday, March 15, 2013

8:30am: Breakfast

9:00am: Christof Hauert, University of Brittish Columbia: "Cooperation in spatially structured populations - a brief history"

Abstract: Over the past decades the interplay between individual behaviour and population structure has attracted interest across disciplines. Early game theoretical work considered lattice populations - mimicking spatial extension by arranging individuals on a grid where they interact with their nearest neighbours. The spatial clustering enables cooperators to survive under conditions where they would not in unstructured populations. This modelling setup is promising, for example for bacterial populations, but in other species geographical constraints are less important, which applies in particular to human interactions. This resulted in a shift of focus towards more general population structures and, in particular, towards scale-free and small-world networks, which are better models for human social interactions. The heterogeneity of these structures may further promote cooperative behaviour. The logical extension came in the form of dynamically changing population structures, where individuals are allowed to adjust their social ties - which is, of course, a crucial aspect of social networks. However, modelling the creation and severing of social ties is itself a challenge, as it bears all the hallmarks of a game already. We now propose a new approach to model dynamical networks that removes the dichotomy between strategy and structure and provides a natural link dynamic, which follows the principle that social networks are defined by social interactions and social interactions define an individuals' strategy: the structure encodes the strategy or, the strategy determines the structure.

10:30am: Rosa Cao, New York University: "Brain Plasticity and Coordinated Population Action"

Abstract: How can collectives in the brain coordinate their activity to produce functional behavior as circumstances change? Collectives of synapses face problems of “judgement aggregation” similar to those characterized in elsewhere (e.g. in decision-making by human groups, Pettit 2009). Beyond fast neural action potentials, processes involving other cell types at slower time scales (from hundreds of milliseconds to minutes to days) play an important role in brain function. Plasticity allows function to be shaped over the course of an individual’s development and experiences to fit its particular set of challenges. One way of thinking about plasticity is as the consequence of selection on populations of synapses. The function of a given collective – from aggregating sensory information to producing motor output or predicting rewards for the organism – is shaped by the nature of the selection processes facing its members (as well as by the gross constraints imposed by features of anatomy that are common across organisms). One mechanism of selection is modulation, in which changing the activity level of a synapse in turn changes how likely it is to persist (and in what form). I will discuss two prominent types of modulation: feedback effects responsible for phenomena such as synaptic homeostasis (Liu 2004) and feedforward effects responsible for consolidation (e.g. Sabatini 2012). These mechanisms contribute to coordinating action across populations of synapses.

12:00pm: Lunch

1:00pm: Peter Vanderschraaf, UC Merced: " The Evolution of Information-Based Authority"

Abstract: I present a model of the evolution of an authority equilibrium in a variable state public good game. A player, or set of players, in a game is an authority over other follower players with respect to a subset of strategies when the follower players follow strategies specified by recommendations or orders from the authority. Previous studies of authority equilibria have focused on how existing authority is maintained or lost according to the interactions between the authority and the follower players. Here I explore how agents can learn to follow an authority equilibrium in a public good game where there is no preexisting authority. The specific game is an 8-player state dependent threshold public good game defined in terms of Stag Hunt games. When only one player knows the true state that defines the threshold, this leader player becomes an authority as sufficiently many of the other players learn to obey the leader's orders to contribute to the provision of the public good and consequently generate this public good at appropriate states. This leader's authority is entirely information-based, that is, the followers maintain the leader's authority only because the leader's orders transmit information useful to them, not because they expect any punishments directed at any player who disobeys. I compare this account of authority based entirely upon differential information against another limiting case in the literature where authority is based entirely upon fear of expected punishment for disobedience, and argue that actual authorities in the world are likely to be maintained by some combination of information transmission and punishment.

2:30pm: Rory Smead, Northeastern University: "The Evolution of Learning in Changing Games"

Abstract: In strategic interactions, the best way to learn often depends on how others are learning. This complicates our understanding of how learning (or cognition more generally) might evolve in these settings. Recent game-theoretic studies in the evolution of learning have revealed that we should expect a diversity of learning rules to persist in populations. Most of these studies, however, have assumed a single fixed game being played between members of the population. Since different learning rules are successful in different circumstances, a changing game will likely change the relative success of learning rules as well. To address this issue, I will explore a model of the evolution of learning in a more dynamic interaction setting, where the game individuals are playing changes over time. I will begin with a basic model of the evolution of trial-and-error learning rules, which are variations of Win-Stay, Lose-Randomize, a very simple but successful learning rule in game contexts. The results of this model model illustrate why we should not expect a single learning rule to dominate in a fixed game. I then extend the model to include changing games and assess the importance of this additional complexity for the evolution of learning.

4:00pm: Break

4:30pm: Cristina Bicchieri, University of Pennsylvania:  "Altruistic punishment, compensation and reward in Ultimatum games"

Abstract. We measured the beliefs and behavior of third parties who were given the opportunity to add to or deduct from the payoffs of individuals who engaged in an economic bargaining game under different social contexts. Third parties rewarded bargaining outcomes that were equal and compensated victims of unfair bargaining outcomes rather than punishing perpetrators, but were willing to punish when compensation was not an available option. Beliefs of whether unequal bargaining outcomes were fair differed based on the normative context, but actual punishment, compensation, and rewarding behavior did not. This paper makes a contribution to the literature by comparing negative sanctions, positive sanctions, and compensation behavior by third parties.

6:00pm: Dinner

Saturday, March 16, 2013

8:30am: Breakfast

9:00am: Gerhard Jaeger, University of Tuebingen: "The Iterated Best Response Model of game theoretic pragmatics and its relatives" (Joint work with Judith Degen and Michael Franke)

Abstract: An important finding of the game theoretic research on signaling games is the insight that under many circumstances, a signal obtains credibility by incurring costs to the sender. Therefore it seems questionable whether or not "cheap talk" - signals that are not payoff relevant - can serve to transmit information among rational agents. This issue is non-trivial in strategic interactions where the preferences of the players are not aligned.

Researchers like Crawford & Sobel, Rabin, and Farrell demonstrated, however, that even in the case of partially divergent interests, cheap talk may be informative. They assume that signals have an exogenously given meaning that is common knowledge between the players, and they explore the conditions under which such a signal is credible.

This discussion has obvious relevance to the program of Gricean pragmatics in linguistics. According to Grice's "Cooperative Principle", this research tradition only considers scenarios where the interests of the players are aligned. Nevertheless the assumption of differential signaling costs introduces an element of non-aligned interests here.

The present paper proposes a framework that combines these two research strands. Using an inference protocol of "iterated best response", it gives a recipe how the interlocutors derive rationalizable strategies from exogeneously given "literal" meanings of signals. No special assumptions about the alignment of interests or signaling costs are made.

After introducing the formal model, the presentation sketches several applications of this models to problems in linguistic pragmatics, such scalar implicatures, the division of pragmatic labor, and the interpretation of measure terms.

In the final part of the talk I will present some recent results of behavioral experiments where participants play a signaling game. To fit the theoretical model to the experimental data, the "best response" rule is replaced by the "quantal response" rule from behavioral game theory that captures bounded rationality in a probabilistic way.

10:30am: J. McKenzie Alexander, London School of Economics: "Epistemic Landscapes and Optimal Search"

Abstract: In a recent paper, Weisberg and Muldoon present an argument for why the division of cognitive labour is useful and advantageous. They represent the space of truths as an "epistemic landscape", where the height of a point on the landscape indicates its significance, and agents move around on this landscape using various cognitive search strategies. They argue that a heterogenous population of boundedly rational agents, who use a mix of cognitive strategies, has greater success in discovering all (significant) truths within a reasonable time than a homogeneous population. We begin by showing analytically that their model contains a critical error which, when corrected, undermines a central claim of their paper. We then show how some alternative cognitive strategies perform better in homogenous populations than the mix of cognitive strategies identified by Weisberg and Muldoon. Finally, we consider the extent to which any of these cognitive strategies perform well in a model adapted from the NK-fitness landscapes of Kauffman.

12:00pm: Lunch

1:00pm: Jorge Pacheco, Universities of Mehno and Lisbon: "Diversity, Adaptation and the Evolution of Cooperation"

Abstract: Several mechanisms have been discovered to date which explain how cooperation may evolve in populations engaging in prisoner’s dilemma type game theoretical interactions whose sole Nash equilibrium is to defect. Here I shall employ the so-called network reciprocity mechanism to show how diversity plays such an important role in the evolution of cooperation. At a macroscopic, population-wide scale, networks which lead individuals to play different roles in the population help transforming a prisoner’s dilemma game into a stag- hunt game. I proceed to show that such diversity of roles is actually the most natural scenario, mostly whenever individuals have a choice in deciding whom they interact with, thereby reshaping the network with time. I will finalize by proposing diversity – while providing examples along the way - as a general concept to bear in mind regarding the evolution of cooperation in its different guises.

2:30pm: Ed Hopkins, University of Edinburgh: " Competitive Altruism, Mentalizing and Signalling"

Abstract: The human tendency to cooperate with nonkin even in short-run relationships remains a puzzle. Recently it has been hypothesized that altruism may be a byproduct of ``mentalizing'', the process of understanding and predicting the mental states of others. Another idea is based on sexual selection: altruism is a costly signal of good genes.  The paper shows that these two arguments are stronger when combined in that altruists who can mentalize have a greater advantage over non-altruists when they can signal their type, even though these signals are costly. Further, once such an equilibrium is established, altruists will not be supplanted by mutants who have similar mentalizing abilities but who lack altruism.

4:00pm: Break

4:30pm: Kevin Zollman, Carnegie Mellon University: "The Pygmalion Game: The Border Between Signals and Signs"

Abstract: Traditionally biologists have distinguished between signs and signals.  Signs -- like "smoke means fire" -- are those correlations that are not subject to natural selection and therefore present no evolutionary mystery.  Signals -- like conventional signals in language -- could easily have failed to convey information and so present some sort evolutionary question.  In this paper, we consider a game that has equilibria that lie on the edge between signs and signals.  In addition to illustrating a conceptual problem, this game provides a plausible explanation for some interesting cases of animal behavior.  This shows that the distinction between signs and signals may not be as neat as originally supposed.

6:pm: Dinner

Sunday, March 17, 2013

9:00am: Breafast

10:00am: Carl Bergstrom, University of Washington: "Defensive complexity in antagonistic coevolution"

Abstract: One strategy for winning a coevolutionary struggle is to evolve rapidly. Most of the literature on host-pathogen coevolution focuses on this phenomenon, and looks for consequent evidence of coevolutionary arms races. An alternative strategy, less often considered in the literature, is to deter rapid evolutionary change by the opponent. To study how this can be done, we construct an evolutionary game between a controller that must process information, and an adversary that can tamper with this information processing. In this game, a species can foil its antagonist by processing information in a way that is hard for the antagonist to manipulate. We show that the structure of the information processing system induces a fitness landscape on which the adversary population evolves. Complex processing logic can carve long, deep fitness valleys that slow adaptive evolution in the adversary population. We suggest that this type of defensive complexity on the part of the vertebrate adaptive immune system may be an important element of coevolutionary dynamics between pathogens and their vertebrate hosts.

11:30am: Elliott Wagner, University of Amsterdam: "Communication in finite populations with divergent interests"

Abstract: When speakers' and hearers' interests are sufficiently misaligned, communication may not be possible at Nash equilibria.  After all, why would the receiver listen to signals if the sender has a temptation to deceive? And why would the speaker signal informatively if she can gain through exploiting the receiver’s behavior?  Roughly speaking, this wisdom is confirmed through standard infinite population models of learning and evolution in games.  Infinite population models do not have limit points that are not Nash equilibria, and therefore these models do not predict the emergence of communication in signaling games with sufficiently misaligned interests.  Models of finite populations, however, sometimes exhibit different long-run behavior; due to random chance (which is washed out in infinite population models), non-equilibrium strategies can occasionally become fixed in finite populations.  Using techniques developed by Fudenberg and Imhoff (2006), it is demonstrated in this paper that finite populations may have a high probability of sustaining communication even when interests are so divergent that separating strategies are not Nash equilibria of the underlying signaling game.  The Sir Philip Sidney game is used as an example illustrating this phenomenon.

1:00pm Lunch

For further information, please contact Patty Jones, or 949-824-1520.


© UC Irvine School of Social Sciences - 3151 Social Sciences Plaza, Irvine, CA 92697-5100 - 949.824.2766