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. (shrink)

Because of its complexity, contemporary scientific research is almost always tackled by groups of scientists, each of which works in a different part of a given research domain. We believe that understanding scientific progress thus requires understanding this division of cognitive labor. To this end, we present a novel agent-based model of scientific research in which scientists divide their labor to explore an unknown epistemic landscape. Scientists aim to climb uphill in this landscape, where elevation represents the significance of the (...) results discovered by employing a research approach. We consider three different search strategies scientists can adopt for exploring the landscape. In the first, scientists work alone and do not let the discoveries of the community as a whole influence their actions. This is compared with two social research strategies, which we call the follower and maverick strategies. Followers are biased towards what others have already discovered, and we find that pure populations of these scientists do less well than scientists acting independently. However, pure populations of mavericks, who try to avoid research approaches that have already been taken, vastly outperform both of the other strategies. Finally, we show that in mixed populations, mavericks stimulate followers to greater levels of epistemic production, making polymorphic populations of mavericks and followers ideal in many research domains. (shrink)

Modelers often rely on robustness analysis, the search for predictions common to several independent models. Robustness analysis has been characterized and championed by Richard Levins and William Wimsatt, who see it as central to modern theoretical practice. The practice has also been severely criticized by Steven Orzack and Elliott Sober, who claim that it is a nonempirical form of confirmation, effective only under unusual circumstances. This paper addresses Orzack and Sober's criticisms by giving a new account of robustness analysis and (...) showing how the practice can identify robust theorems. Once the structure of robust theorems is clearly articulated, it can be shown that such theorems have a degree of confirmation, despite the lack of direct empirical evidence for their truth. (shrink)

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 tradeoff between the reliability of a community and the speed with which it reaches a correct conclusion. ‡The author would like to thank Brian Skyrms, Kyle Stanford, Jeffrey Barrett, Bruce Glymour, and the participants in the Social Dynamics Seminar at University of California–Irvine for their helpful comments. Generous financial support was provided by the School of Social Science and Institute for Mathematical Behavioral Sciences at UCI. †To contact the author, please write to: Department of Philosophy, Baker Hall 135, Carnegie Mellon University, Pittsburgh, PA 15213-3890; e-mail: kzollman@andrew.cmu.edu. (shrink)

This article examines two questions about scientists’ search for knowledge. First, which search strategies generate discoveries effectively? Second, is it advantageous to diversify search strategies? We argue pace Weisberg and Muldoon, “Epistemic Landscapes and the Division of Cognitive Labor”, that, on the first question, a search strategy that deliberately seeks novel research approaches need not be optimal. On the second question, we argue they have not shown epistemic reasons exist for the division of cognitive labor, identifying the errors that led (...) to their conclusions. Furthermore, we generalize the epistemic landscape model, showing that one should be skeptical about the benefits of social learning in epistemically complex environments. (shrink)

Some scientists are happy to follow in the footsteps of others; some like to explore novel approaches. It is tempting to think that herein lies an epistemic division of labor conducive to overall scientific progress: the latter point the way to fruitful areas of research, and the former more fully explore those areas. Weisberg and Muldoon’s model, however, suggests that it would be best if all scientists explored novel approaches. I argue that this is due to implausible modeling choices, and (...) I present an alternative ‘epistemic landscape’ model that demonstrates the alleged benefits from division of labor, with one restriction. (shrink)

The article presents an agent-based model of scientific interaction aimed at examining how different degrees of connectedness of scientists impact their efficiency in knowledge acquisition. The model is built on the basis of Zollman’s ABM by changing some of its idealizing assumptions that concern the representation of the central notions underlying the model: epistemic success of the rivalling scientific theories, scientific interaction and the assessment in view of which scientists choose theories to work on. Our results suggest that whether and (...) to what extent the degree of connectedness of a scientific community impacts its efficiency is a highly context-dependent matter since different conditions deem strikingly different results. More generally, we argue that simplicity of ABMs may come at a price: the requirement to run extensive robustness analysis before we can specify the adequate target phenomenon of the model.1 1Introduction2Zollman's 2010 Model3Static versus Dynamic Epistemic Success 3.1Introducing the notion of dynamic epistemic success3.2Implementation and results for the basic setup4Critical Interaction 4.1Introducing critique4.2Implementation and results5Inertia of Inquiry 5.1Introducing rational inertia5.2Implementation and results6Threshold Below Which Theories Are Equally Promising 6.1An inquiry that is even more difficult6.2Implementation and results7Discussion8Conclusion. (shrink)

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. 1 Introduction2 Scientists and Bandits3 Choosing an ϵ4 Structure of Communication5 Discussion. (shrink)

When should a scientific community be cognitively diverse? This article presents a model for studying how the heterogeneity of learning heuristics used by scientist agents affects the epistemic efficiency of a scientific community. By extending the epistemic landscapes modeling approach introduced by Weisberg and Muldoon, the article casts light on the micro-mechanisms mediating cognitive diversity, coordination, and problem-solving efficiency. The results suggest that social learning and cognitive diversity produce epistemic benefits only when the epistemic community is faced with problems of (...) sufficient difficulty. (shrink)

In this paper we examine the epistemic value of highly idealized agent-based models of social aspects of scientific inquiry. On the one hand, we argue that taking the results of such simulations as informative of actual scientific inquiry is unwarranted, at least for the class of models proposed in recent literature. Moreover, we argue that a weaker approach, which takes these models as providing only “how-possibly” explanations, does not help to improve their epistemic value. On the other hand, we suggest (...) that if ABMs of science underwent two types of robustness analysis, they could indeed have a clear epistemic function, namely by providing evidence for philosophical and historical hypotheses. In this sense, ABMs can obtain evidential and explanatory properties and thus be a useful tool for integrated history and philosophy of science. We illustrate our point with an example of a model—building on the work by Kevin Zollman—which we apply to a concrete historical case study. (shrink)

The aim of this paper is to offer an account of epistemic justification suitable for the context of theory pursuit, that is, for the context in which new scientific ideas, possibly incompatible with the already established theories, emerge and are pursued by scientists. We will frame our account paradigmatically on the basis of one of the influential systems of epistemic justification: Laurence Bonjour’s coherence theory of justification. The idea underlying our approach is to develop a set of criteria which indicate (...) that the pursued system is promising of contributing to the epistemic goal of robustness of scientific knowledge and of developing into a candidate for acceptance. In order to realize this we will (a) adjust the scope of Bonjour’s standards—consistency, inferential density, and explanatory power, and (b) complement them by the requirement of a programmatic character. In this way we allow for the evaluation of the “potential coherence” of the given epistemic system. (shrink)

A scientific community can be modeled as a collection of epistemic agents attempting to answer questions, in part by communicating about their hypotheses and results. We can treat the pathways of scientific communication as a network. When we do, it becomes clear that the interaction between the structure of the network and the nature of the question under investigation affects epistemic desiderata, including accuracy and speed to community consensus. Here we build on previous work, both our own and others’, in (...) order to get a firmer grasp on precisely which features of scientific communities interact with which features of scientific questions in order to influence epistemic outcomes. (shrink)

Figuring prominently in their decisions regarding which theories to pursue are scientists' appeals to the promise or lack of promise of those theories. Yet philosophy of science has had little to say about how one is to assess theory promise. This essay identifies several indices that might be consulted to determine whether or not a theory is promising and worthy of pursuit. Various historical examples of appeals to such indices are introduced.

A small consortium of philosophers has begun work on the implications of epistemic networks (Zollman 2008 and forthcoming; Grim 2006, 2007; Weisberg and Muldoon forthcoming), building on theoretical work in economics, computer science, and engineering (Bala and Goyal 1998, Kleinberg 2001; Amaral et. al., 2004) and on some experimental work in social psychology (Mason, Jones, and Goldstone, 2008). This paper outlines core philosophical results and extends those results to the specific question of thresholds. Epistemic maximization of certain types does show (...) clear threshold effects. Intriguingly, however, those effects appear to be importantly independent from more familiar threshold effects in networks. (shrink)

Pluralism about scientific method is more-or-less accepted, but the consequences have yet to be drawn out. Scientists adopt different methods in response to different epistemic situations: depending on the system they are interested in, the resources at their disposal, and so forth. If it is right that different methods are appropriate in different situations, then mismatches between methods and situations are possible. This is most likely to occur due to method bias: when we prefer a particular kind of method, despite (...) that method clashing with evidential context or our aims. To explore these ideas, we sketch a kind of method pluralism which turns on two properties of evidence, before using agent-based models to examine the relationship between methods, epistemic situations, and bias. Based on our results, we suggest that although method bias can undermine the efficiency of a scientific community, it can also be productive through preserving a diversity of evidence. We consider circumstances where method bias could be particularly egregious, and those where it is a potential virtue, and argue that consideration of method bias reveals that community standards deserve a central place in the epistemology of science. (shrink)

argumentation has been shown to be a powerful tool within many fields such as artificial intelligence, logic and legal reasoning. In this paper we enhance Dung’s well-known abstract argumentation framework with explanatory capabilities. We show that an explanatory argumentation framework (EAF) obtained in this way is a useful tool for the modeling of scientific debates. On the one hand, EAFs allow for the representation of explanatory and justificatory arguments constituting rivaling scientific views. On the other hand, different procedures for selecting (...) arguments, corresponding to different methodological and epistemic requirements of theory evaluation, can be formulated in view of our framework. (shrink)

Multiarm bandit problems have been used to model the selection of competing scientific theories by boundedly rational agents. In this paper, I define a variable-arm bandit problem, which allows the set of scientific theories to vary over time. I show that Roth-Erev reinforcement learning, which solves multiarm bandit problems in the limit, cannot solve this problem in a reasonable time. However, social learning via preferential attachment combined with individual reinforcement learning which discounts the past, does.

This paper examines lessons obtained by means of simulations in the form of agent-based models about the norms that are to guide disagreeing scientists. I focus on two types of epistemic and methodological norms: norms that guide one’s attitude towards one’s own theory, and norms that guide one’s attitude towards the opponent’s theory. Concerning I look into ABMs that have been designed to examine the context of peer disagreement. Here I challenge the conclusion that the given ABMs provide a support (...) for the so-called Steadfast Norm, according to which one is epistemically justified in remaining steadfast in their beliefs in face of disagreeing peers. I argue that the proposed models at best provide evidence for a weaker norm, which concerns methodological steadfastness. Concerning I look into ABMs aimed at examining epistemic effects of scientific interaction. Here I argue that the models provide diverging suggestions and that the link between each ABM and the type of represented inquiry is still missing. Moreover, I examine alternative strategies of arguing in favor of the benefits of scientific interaction, relevant for contemporary discussions on scientific pluralism. (shrink)

Much recent discussion in social epistemology has focussed on the question of whether peers can rationally sustain a disagreement. A growing number of social epistemologists hold that the answer is negative. We point to considerations from the history of science that favor rather the opposite answer. However, we also explain how the other position can appear intuitively attractive.

Throughout the first half of the twentieth century the research on peptic ulcer disease focused on two rivaling hypothesis: the “acidity” and the “bacterial” one. According to the received view, the latter was dismissed during the 1950s only to be revived with Warren’s and Marshall’s discovery of Helicobacter pylori in the 1980s. In this paper we investigate why the bacterial hypothesis was largely abandoned in the 1950s, and whether there were good epistemic reasons for its dismissal. Of special interest for (...) our research question is Palmer’s 1954 large-scale study, which challenged the bacterial hypothesis with serious counter-evidence, and which by many scholars is considered as the shifting point in the research on PUD. However, we show that: The perceived refutatory impact of Palmer’s study was disproportionate to its methodological rigor. This undermines its perceived status as a crucial experiment against the bacterial hypothesis. In view of this and other considerations we argue that the bacterial hypothesis was worthy of pursuit in the 1950s. (shrink)