# Signaling

The Handicap Principle is an Artifact
(with Simon Huttegger and Justin Bruner) forthcoming in Philosophy of Science
The handicap principle is one of the most influential ideas in evolutionary biology. It asserts that when there is conflict of interest in a signaling interaction signals must be costly in order to be reliable. We show how the handicap principle is a limiting case of honest signaling, which can also be sustained by other mechanisms. This fact has gone unnoticed because in evolutionary biology it is a common practice to distinguish between cues, indexes and fakable signals, where cues provide information but are not signals and indexes are signals that cannot be faked. We find that the dichotomy between indexes and fakable signals is an artifact of the existing signaling models. Our results suggest that one cannot adequately understand signaling behavior by focusing solely on cost. Under our reframing, cost becomes one -- and probably not the most important -- of a collection of factors preventing deception.
Methodology in Biological Game Theory
(with Simon Huttegger) forthcoming in the British Journal for the Philosophy of Science
Game theory has a prominent role in evolutionary biology, in particular in the ecological study of various phenomena ranging from conflict behavior to altruism to signaling and beyond. The two central methodological tools in biological game theory are the concepts of Nash equilibrium and Evolutionarily Stable Strategy (ESS). While both were inspired by a dynamic conception of evolution, these concepts are essentially static -- they only show that a population is uninvadable, but not that a population is likely to evolve. In this paper we argue that a static methodology can lead to misleading views about a dynamic evolutionary processes. Instead, we advocate a more pluralistic methodology, which includes both static and dynamic game theoretic tools. Such an approach provides a more complete picture of the evolution of strategic behavior.
Probe and Adjust in Information Transfer Games
(with Simon Huttegger and Brian Skyrms) forthcoming in Erkenntnis
We study a low-rationality learning dynamics called {\it probe and adjust}. Our emphasis is on its properties in games of information transfer such as the Lewis signaling game or the Bala-Goyal network game. These games fall into the class of {\it weakly better reply games}, in which, starting from any action profile, there is a weakly better reply path to a strict Nash equilibrium. We prove that probe and adjust will be close to strict Nash equilibria in this class of games with arbitrarily high probability. In addition, we compare these asymptotic properties to short-run behavior.
Separating directives and assertions using simple signaling games
The Journal of Philosophy 63(3): 158-169
Most contemporary accounts of meaning utilize some notion of the intentions of the speaker and or hearer. In this paper I develop an account of the distinction between directives and assertions which does not require that the speakers have intentions. Instead, this account relies on behavioral differences modeled as different strategies in an appropriately defined game.
Signaling Games: Dynamics of Evolution and Learning
(with Simon Huttegger) in Language, Games, and Evolution (Anton Benz, Christian Ebert, Gerhard Jaeger, and Robert van Rooij eds.) Springer
A survey of recent results pertaining to evolution of signaling in the Lewis signaling game.
Evolutionary Dynamics of Lewis Signaling Games
(with Simon Huttegger, Brian Skyrms, and Rory Smead) Synthese 172(1): 177-191.
Transfer of information between senders and receivers, of one kind or another, is essential to all life. David Lewis introduced a game theoretic model of the simplest case, where one sender and one receiver have pure common interest. How hard or easy is it for evolution to achieve information transfer in Lewis signaling?. The answers involve surprising subtleties. We discuss some if these in terms of evolutionary dynamics in both finite and infinite populations, with and without mutation.
The Role of Forgetting in the Evolution and Learning of Language
(with Jeffrey Barrett) Journal of Experimental and Theoretical Artificial Intelligence 21(4): 293-309.
Lewis signaling games illustrate how language might evolve from random behavior. The probability of evolving an optimal signaling language is, in part, a function of what learning strategy the agents use. Here we investigate three learning strategies, each of which allows agents to forget old experience. In each case, we find that forgetting increases the probability of evolving an optimal language. It does this by making it less likely that past partial success will continue to reinforce suboptimal practice. The learning strategies considered here show how forgetting past experience can promote learning in the context of games with suboptimal equilibria.
Talking to Neighbors: The Evolution of Regional Meaning
Philosophy of Science 74(5): 574-587.
In seeking to explain the evolution of social cooperation, many scholars are using increasingly complex game-theoretic models. These complexities often model readily observable features of human and animal populations. In the case of previous games analyzed in the literature, these modifications have had radical effects on the stability and efficiency properties of the models. We will analyze the effect of adding spatial structure to two communication games: the Lewis Sender-Receiver game and a modified Stag Hunt game. For the Stag Hunt, we find that the results depart strikingly from previous models. In all cases, the departures increase the explanatory value of the models for social phenomenon.
Between Cheap and Costly Signaling: The Evolution of Partially Honest Communivation
(with Simon Huttegger and Carl Bergstrom) Proceedings of the Royal Society of London B: Biological Sciences 208: 20121878.
Costly signaling theory has become a common explanation for honest communication when interests conflict. In this paper, we provide an alternative explanation for partially honest communication that does not require significant signal costs. We show that this alternative is at least as plausible as traditional cost signaling, and we suggest a number of experiments that might be used to distinguish the two theories.
Finding alternatives to handicap theory
Forthcoming in Biological Theory
The Handicap Principle represents a central theory in the biological understanding of signaling. This paper presents a number of alternative theories to the Handicap Principle and argues that some of these theories may provide a better explanation for the evolution and stability of honest communication.
Dynamic stability and basins of attraction in the Sir Philip Sidney game
(with Simon Huttegger) Proceedings of the Royal Society of London B: Biological Sciences 277, 1915-1922
We study the handicap principle in terms of the Sir Philip Sidney game. The handicap principle asserts that cost is required to allow for honest signalling in the face of conflicts of interest. We show that the significance of the handicap principle can be challenged from two new directions. Firstly, both the costly signalling equilibrium and certain states of no communication are stable under the replicator dynamics (i.e. standard evolutionary dynamics); however, the latter states are more likely in cases where honest signalling should apply. Secondly, we prove the existence and stability of polymorphisms where players mix between being honest and being deceptive and where signalling costs can be very low. Neither the polymorphisms nor the states of no communication are evolutionarily stable, but they turn out to be more important for standard evolutionary dynamics than the costly signalling equilibrium.

# Social structure of science

Conservativism and the scientific state of nature
(with Erich Kummerfeld) forthcoming in British Journal for the Philosophy of Science
Those who comment on modern scientific institutions are often quick to praise institutional structures that leave scientists to their own devices. These comments reveal an underlying presumption that scientists do best when left alone -- when they operate in what we call the scientific state of nature. Through computer simulation, we challenge this presumption by illustrating an inefficiency that arises in the scientific state of nature. This inefficiency suggests that one cannot simply presume that science is most efficient when institutional control is absent. In some situations actively encouraging unpopular, risky science would improve scientific outcomes.
Network Epistemology: Communication in Epistemic Communities
Forthcoming in Philosophy Compass
Much of contemporary knowledge is generated by groups not single individuals. A natural question to ask is, what features make groups better or worse at generating knowledge? This paper surveys research that spans several disciplines which focuses on one aspect of epistemic communities: the way they communicate internally. This research has revealed that a wide number of different communication structures are best, but what is best in a given situation depends on particular details of the problem being confronted by the group.
The Independence Thesis: When Individual and Social Epistemology Diverge
(with Conor Mayo-Wilson and David Danks) Philosophy of Science 78(4): 657-677.
Several philosophers of science have argued that epistemically rational individuals might form epistemically irrational groups and that, conversely, rational groups might be composed of irrational individuals. We call the conjunction of these two claims the Independence Thesis, as they entail that methodological prescriptions for scientific communities and those for individual scientists are logically independent. We defend the inconsistency thesis by characterizing four criteria for epistemic rationality and then proving that, under said criteria, individuals will be judged rational when groups are not and vice versa. We then explain the implications of our results for descriptive history of science and normative epistemology.
Computer Simulation and Emergent Reliability in Science
Journal of Artificial Societies and Social Simulation 14 (4): 15
While the popular image of scientists portrays them as objective, dispassionate observers of nature, actual scientists rarely are. It is not really known to what extent these individual departures from the scientific ideal effects the reliability of the scientific community. This paper suggests a number of concrete projects which help to determine this relationship.
Wisdom of the Crowds vs. Groupthink: Learning in Groups and in Isolation
(with Conor Mayo-Wilson and David Danks) forthcoming in International Journal of Game Theory
We evaluate the asymptotic performance of boundedly-rational strategies in multi-armed bandit problems, where performance is measured in terms of the tendency (in the limit) to play optimal actions in either (i) isolation or (ii) networks of other learners. We show that, for many strategies commonly employed in economics, psychology, and machine learning, performance in isolation and performance in networks are essentially unrelated. Our results suggest that the performance of various, common boundedly-rational strategies depends crucially upon the social context (if any) in which such strategies are to be employed.
Social Network Structure and the Achievement of Consensus
Politics, Philosophy, and Economics 11: 26-44
Run the simulation
It is widely believed that bringing parties with differing opinions together to discuss their differences will help both in securing consensus and also in ensuring that this consensus closely approximates the truth. This paper investigates this presumption using two mathematical and computer simulation models. Ultimately, these models show that increased contact can be useful in securing both consensus and truth, but it is not always beneficial in this way. This suggests one should not, without qualification, support policies which increase interpersonal contact if one seeks to improve the epistemic performance of groups.
Wisdom of Crowds vs. Groupthink: Learning in groups and in Isolation
(with Conor Mayo-Wilson and David Danks) Carnegie Mellon University Working Paper
We evaluate the asymptotic performance of boundedly-rational strategies in multi-armed bandit problems, where performance is measured in terms of the tendency (in the limit) to play optimal actions in either (i) isolation or (ii) networks of other learners. We show that, for many strategies commonly employed in economics, psychology, and machine learning, performance in isolation and performance in networks are essentially unrelated. Our results suggest that the appropriateness of various, common boundedly-rational strategies depends crucially upon the social context (if any) in which such strategies are to be employed.
Optimal Publishing Strategies
Episteme 6(2): 185-199.
Run the simulation
Journals regulate a significant portion of the communication between scientists. This paper devises an agent-based model of scientific practice and uses it to compare various strategies for selecting publications by journals. Surprisingly, it appears that the best selection method for journals is to publish relatively few papers and to select those papers it publishes at random from the available above threshold'' papers it receives. This strategy is most effective at maintaining an appropriate type of diversity which is needed to solve a particular type of scientific problem. This problem and the limitation of the model is discussed in detail.
The Epistemic Benefit of Transient Diversity
Erkenntnis 72(1):17-35
(source code)
There is growing interest in understanding and eliciting division of labor within groups of scientists. This paper illustrates the need for this division of labor through a historical example, and a formal model is presented to better analyze situations of this type. Analysis of this model reveals that a division of labor can be maintained in two different ways: by limiting information or by endowing the scientists with extreme beliefs. If both features are present however, cognitive diversity is maintained indefinitely, and as a result agents fail to converge to the truth. Beyond the mechanisms for creating diversity suggested here, this shows that the real epistemic goal is not diversity but transient diversity.
Social Structure and the Effects of Conformity
Synthese 172(3):317-340
Conformity is an often criticized feature of human belief formation. Although generally regarded as a negative influence on reliability, it has not been widely studied. This paper attempts to determine the epistemic effects of conformity by analyzing a mathematical model of this behavior. In addition to investigating the effect of conformity on the reliability of individuals and groups, this paper attempts to determine the optimal structure for conformity. That is, supposing that conformity is inevitable, what is the best way for conformity effects to occur? The paper finds that in some contexts conformity effects are reliability inducing and, more surprisingly even when it is counterproductive, not all methods for reducing its effect are helpful. These conclusions contribute to a larger discussion in social epistemology regarding the effect of social behavior on individual reliability.
The Communication Structure of Epistemic Communities
Philosophy of Science 74(5): 574-587
A revised version has been reprinted in (Dennis Whitcomb and Alvin Goldman, eds.) Social Epistemology Essential Readings Oxford University Press
Increasingly, epistemologists are becoming interested in social structures and their effect on epistemic enterprises, but little attention has been paid to the proper distribution of experimental results among scientists. This paper will analyze a model first suggested by two economists, which nicely captures one type of learning situation faced by scientists. The results of a computer simulation study of this model provide two interesting conclusions. First, in some contexts, a community of scientists is, as a whole, more reliable when its members are less aware of their colleagues' experimental results. Second, there is a robust trade-off between the reliability of a community and the speed with which it reaches a correct conclusion.
Network Epistemology
my dissertation

# Methodology in game theory

The Limits of ESS Methodology
(with Simon Huttegger) in (Samir Okasha and Ken Binmore, eds.) Evolution and Rationality: Decisions, Cooperation, and Strategic Behavior
In this paper we show that there are certain limits as to what applications of Maynard Smith's concept of evolutionarily stable strategy (ESS) can tell us about evolutionary processes. We shall argue that ESS is very similar in spirit to a particular branch of rational choice game theory, namely, the literature on refinements of Nash equilibrium. In the first place, ESS can also be viewed as a Nash equilibrium refinement. At a deeper level, ESS shares a common structure with other rational choice equilibrium refinements. An equilibrium is evaluated according to whether it persists under specific kinds of perturbations. In the case of ESS, these perturbations are mutations. However, from a dynamical point of view, focusing exclusively on perturbations of equilibria provides only a partial account of the system under consideration. We will show that this has important consequences when it comes to analyzing game-theoretic models of evolutionary processes. In particular, there are non-ESS states which are significant for evolutionary dynamics.
Methodology in Biological Game Theory
(with Simon Huttegger) forthcoming in the British Journal for the Philosophy of Science
Game theory has a prominent role in evolutionary biology, in particular in the ecological study of various phenomena ranging from conflict behavior to altruism to signaling and beyond. The two central methodological tools in biological game theory are the concepts of Nash equilibrium and Evolutionarily Stable Strategy (ESS). While both were inspired by a dynamic conception of evolution, these concepts are essentially static -- they only show that a population is uninvadable, but not that a population is likely to evolve. In this paper we argue that a static methodology can lead to misleading views about a dynamic evolutionary processes. Instead, we advocate a more pluralistic methodology, which includes both static and dynamic game theoretic tools. Such an approach provides a more complete picture of the evolution of strategic behavior.

# Evolution of Learning

The Stability of Strategic Plasticity
(with Rory Smead) CMU Philosophy department working paper
Recent research into the evolution of higher cognition has piqued an interest in the effect of natural selection on the ability of creatures to respond to their environment (behavioral plasticity). It is believed that environmental variation is required for plasticity to evolve in cases where the ability to be plastic is costly. We investigate one form of environmental variation: frequency dependent selection. Using tools in game theory, we investigate a few models of plasticity and outline the cases where selection would be expected to maintain it. Ultimately we conclude that frequency dependent selection is likely insufficient to maintain plasticity given reasonable assumptions about its costs. This result is very similar to one aspect of the well-discussed Baldwin effect, where plasticity is first selected for and then later selected against. We show how in these models one would expect plasticity to grow in the population and then be later reduced. Ultimately we conclude that if one is to account for the evolution of behavioral plasticity in this way, one must appeal to a very particular sort of external environmental variation.

Plasticity and language: an example of the Baldwin effect?
(with Rory Smead) Philosophical Studies 147(1): 7-21.
In recent years, many scholars have suggested that the Baldwin effect may play an important role in the evolution of language. However, the Baldwin effect is a multifaceted and controversial process and the assessment of its connection with language is difficult without a formal model. This paper provides a first step in this direction. We examine a game-theoretic model of the interaction between plasticity (represented by Herrnstein reinforcement learning) and evolution in the context of a simple language game. Additionally, we describe three distinct aspects of the Baldwin effect: the Simpson-Baldwin effect, the Baldwin expediting effect and the Baldwin optimizing effect. We find that a simple model of the evolution of language lends theoretical plausibility to the existence of the Simpson-Baldwin and the Baldwin optimizing effects in this arena, but not the Baldwin expediting effect.

# The Evolution of Cooperation

Evolutionary Considerations in the Framing of Social Norms
(with Brian Skyrms) Politics, Philosophy, and Economics 9(3): 265-273
In this article, we aim to illustrate evolutionary explanations for the emergence of framing effects, discussed in detail in Cristina Bicchieri’s The Grammar of Society. We show how framing effects might evolve which coalesce two economically distinct interactions into a single one, leading to apparently irrational behavior in each individual interaction. Here we consider the now well-known example of the ultimatum game, and show how this ‘irrational’ behavior might result from a single norm which governs behavior in multiple games. We also show how framing effects might result in radically different play in strategically identical situations. We consider the Hawk-Dove game (the game of chicken) and also the Nash bargaining game. Here arbitrary tags or signals might result in one party doing better than another.
Explaining Fairness in a Complex Environment
Politics, Philosophy, and Economics 7(1): 81-97.
This paper presents the evolutionary dynamics of three games: Nash Bargaining game, the Ultimatum Game, and a hybrid of the two. One might expect that the probability that some behavior evolves in an environment with two games would be near the probability that the same behavior evolves in either game alone. This is not the case for the Ultimatum and Nash Bargaining Games. Fair behavior is more likely to evolve in a combined game than in either game taken individually. This result confirms a conjecture that the complexity of our actual environment provides an explanation for the evolution of fair behavior.
Talking to Neighbors: The Evolution of Regional Meaning
Philosophy of Science 74(5): 574-587
In seeking to explain the evolution of social cooperation, many scholars are using increasingly complex game-theoretic models. These complexities often model readily observable features of human and animal populations. In the case of previous games analyzed in the literature, these modifications have had radical effects on the stability and efficiency properties of the models. We will analyze the effect of adding spatial structure to two communication games: the Lewis Sender-Receiver game and a modified Stag Hunt game. For the Stag Hunt, we find that the results depart strikingly from previous models. In all cases, the departures increase the explanatory value of the models for social phenomenon.

# Ethics

Dignity and the value of rejecting profitable but insulting offers
(with Timos Anthanasiou and Alex John London) forthcoming in Mind
In this paper we distinguish two competing conceptions of dignity, one recognizably Hobbesian and one recognizably Kantian. We provide a formal model of how decision makers committed to these conceptions of dignity might reason when engaged in an economic transaction that is it not inherently insulting, but in which it is possible for the dignity of the agent to be called into question. This is a modified version of the ultimatum game. We then use this model to illustrate ways in which the Kantian evaluative standpoint enjoys a kind of internal stability that the Hobbesian framework lacks. Our interpersonal argument shows that, under certain conditions, Hobbesians prefer to cultivate Kantian commitments in others and promote the presence of Kantians in the population. Our intrapersonal argument shows that agents who are conflicted between Kantian and Hobbesian commitments have powerful reasons not to resolve this commitment in favour of Hobbesian values. Our emulation argument illustrates that in repeated versions of the ultimatum game, the Hobbesian chooses to behave like a Kantian, including publicly repudiating her Hobbesian commitments. Here again, however, the Hobbesian is able to achieve a desired benefit only on the condition that there are genuine Kantians in the population. Finally, our social planning argument explores the reasons that a community of Hobbesians would opt to enshrine a Kantian conception of dignity into law. The paper concludes with some remarks about the policy implications of this work.
Research Subjects at the Auction Block: Problems for a Procedural Approach to Justice in International Research
(with Alex John London) Hastings Center Report 40(4): 34-45.
The "fair benefits" approach to international research is designed to produce results that all can agree are fair without taking a stand on divisive questions of justice. But its appealing veneer of collaboration masks ambiguities at both a conceptual and an operational level. An attempt to put it into practice would look a lot like an auction, leaving little reason to think the outcomes will satisfy even minimal conditions of fairness.

Kevin is an associate professor of philosophy at Carnegie Mellon University in Pittsburgh, PA. His research interests include philosophy of science, philosophy of biology, game theory, and decision theory.

#### Contact

Kevin Zollman
Baker Hall 135
Department of Philosophy
Carnegie Mellon University
Pittsburgh, PA 15213-3890

office phone: 412-268-8493

email: kzollman@andrew.cmu.edu