We study the connection between communication network structure and an organization's collective adaptability to a shifting environment. Research has shown that network centralization-the degree to which communication flows disproportionately through one or more members of the organization rather than being more equally distributed-interferes with collective problem-solving by obstructing the integration of existing ideas, information, and solutions in the network. We hypothesize that the mechanisms responsible for that poor integration of ideas, information, and solutions would nevertheless prove beneficial for problems requiring adaptation to a shifting environment. We conducted a 1,620-subject randomized online laboratory experiment, testing the effect of seven network structures on problem-solving success. To simulate a shifting environment, we designed a murder mystery task and manipulated when each piece of information could be found: early information encouraged an inferior consensus, requiring a collective shift of solution after more information emerged. We find that when the communication network within an organization is more centralized, it achieves the benefits of connectivity (spread of novel better solutions) without the costs (getting stuck on an existing inferior solution). We also find, however, that these benefits of centralization only materialize in networks with two-way flow of information and not when information only flows from the center of the network outward (as can occur in hierarchical structures or digitally mediated communication). We draw on these findings to reconceptualize theory on the impact of centralization-and how it affects conformity pressure (lock-in) and awareness of diverse ideas (learning)-on collective problem-solving that demands adaptation.

How do social networks affect the diffusion of information? While previous research has mainly focused on the spread of specific messages, we study how the overall mix of information that diffuses through multiple intermediaries can become distorted or biased based on what categories of information people pass on versus filter out. We conducted randomized online laboratory experiments of diffusion through multi-step social networks. We find support for our pre-registered hypotheses that (1) the further someone is down a diffusion chain, the more the mix of information that they receive is biased toward rare categories of events because (2) information about rare categories of events is passed on disproportionately frequently. Our data is consistent with a preference for variety in what is shared, as well as a perceptual bias in favor of rare events. We name the disproportionate diffusion of rare categories of events "Man-Bites-Dog Contagion." Even when people intend to be accurate and informative, multi-step diffusion risks de-emphasizing the importance of empirically common categories of events and over-emphasizing the importance of empirically rare categories of events.

Promoting contribution of content is a key challenge for platforms that support the collective creation or transfer of knowledge. We study the role of forum size (number of people in a forum) on contribution of content per person with a field experiment on a massive open online course (MOOC). We find that larger forums elicit more contribution per person. The number of questions and other help-seeking threads posted per person was unchanged by size, but replies and other more conversational posts increased sharply. Most of the positive effect of size was in a subset of socially responsive subjects. The implication of social responsiveness driving our results is that the unequal distribution of contribution on online platforms is unlikely to be easily changed: if more contributions are elicited from infrequent contributors, the greatest contributors would contribute even more due to there being more to respond to.

?Que es mas productivo, que los empleados de la compania permanezcan continuamente colaborando con sus companeros o que trabajen de forma individual? ?En que momentos es mejor optar por la interaccion y en cuales es preferible aislarse? ?Cual seria el ritmo idoneo de colaboracion entre trabajadores? A partir de varios anos de investigacion, los autores del articulo han llegado a la conclusion de que, cuando se trata de resolver problemas complejos, interactuar mas no conlleva encontrar mejores soluciones, y que un exceso de colaboracion tiene sus costes, entre ellos, por ejemplo, que se elimina la diversidad de pensamiento que contribuye a crear las mejores soluciones. Por ello, proponen una serie de pautas para encontrar el ritmo adecuado en el trabajo.

This paper examines how centralization of organizational communication networks impacts problem solving performance on tasks requiring a shift from one inferior solution to a dissimilar, novel one—or, in the language of March, tasks that require exploration rather than just exploitation. Drawing on a 1,620-subject experiment using a new platform designed to run randomized experiments with network structure as the independent variable, we tested the effect of seven network structures—representing different levels of centralization—on problem solving success. To simulate a dynamic environment with shifting information, we designed a murder mystery task and manipulated when each piece of information could be found: early information encourages an incorrect consensus, requiring a collective shift of solution when more information emerged later. We find that when the communication network within an organization is more centralized, it achieves the benefits of social influence (learning) without the costs (herding). In centralized networks, peripheral nodes are relatively independent and less likely to get stuck at an earlier consensus solution (reduced herding). Central nodes then learn from members of the periphery and quickly spread the correct answer to others in the network (increased learning). We also find, however, that these benefits of centralization come with a major caveat: they only materialize in undirected networks, not in directed core-periphery structures that typify selforganized networks like digitally-mediated ones. We draw on these findings to reconceptualize existing theory on the impact of centralization on collective intelligence in problem solving that demands exploration and adaptation by an organizations members.

People influence each other when they interact to solve problems. Such social influence introduces both benefits (higher average solution quality due to exploitation of existing answers through social learning) and costs (lower maximum solution quality due to a reduction in individual exploration for novel answers) relative to independent problem solving. In contrast to prior work, which has focused on how the presence and network structure of social influence affect performance, here we investigate the effects of time. We show that when social influence is intermittent it provides the benefits of constant social influence without the costs. Human subjects solved the canonical traveling salesperson problem in groups of three, randomized into treatments with constant social influence, intermittent social influence, or no social influence. Groups in the intermittent social-influence treatment found the optimum solution frequently (like groups without influence) but had a high mean performance (like groups with constant influence); they learned from each other, while maintaining a high level of exploration. Solutions improved most on rounds with social influence after a period of separation. We also show that storing subjects' best solutions so that they could be reloaded and possibly modified in subsequent rounds-a ubiquitous feature of personal productivity software-is similar to constant social influence: It increases mean performance but decreases exploration.

Social media have great potential to support diverse information sharing, but there is widespread concern that platforms like Twitter do not result in communication between those who hold contradictory viewpoints. Because users can choose whom to follow, prior research suggests that social media users exist in "echo chambers" or become polarized. We seek evidence of this in a complete cross section of hyperlinks posted on Twitter, using previously validated measures of the political slant of news sources to study information diversity. Contrary to prediction, we find that the average account posts links to more politically moderate news sources than the ones they receive in their own feed. However, members of a tiny network core do exhibit cross-sectional evidence of polarization and are responsible for the majority of tweets received overall due to their popularity and activity, which could explain the widespread perception of polarization on social media.

Where do new markets come from? I construct a network model in which national markets are nodes and flows of recorded music between them are links and conduct a longitudinal analysis of the global pattern of trade in the period 1976 to 2010. I hypothesize that new export markets are developed through a process of transitive closure in the network of international trade. When two countries' markets experience the same social influences, it brings them close enough together for new homophilous ties to be formed. The implication is that consumption of foreign products helps, not hurts, home-market producers develop overseas markets, but only in those countries that have a history of consuming the same foreign products that were consumed in the home market. Selling in a market changes what is valued in that market, and new market formation is a consequence of having social influences in common.

MOOCs offer valuable learning experiences to students from all around the world. In addition to providing filmed lectures, readings, and problem sets, many MOOCs allow students to ask and answer questions about course materials with each other through interactive user forums. However, in current MOOCs, only 3 to 5 percent of those students interact in the user forum (Breslow 2013, Rosé et al. 2014) and more than 90 percent of students stop attending the course altogether (Jordan 2014). According to prior studies, this low level of social engagement in MOOCs may lead to student attrition and low performance (Ren et al. 2007). Hence, a natural question that arises then is, how can we promote interaction among students in MOOC discussion forums in order to reduce students' attrition and raise their performance? In this paper, we conduct a field experiment on the edX platform to identify factors that promote student engagement in MOOC discussion forums. Researchers have discovered that the number of people interacting in one online location (e.g. group, community or virtual classroom size) is a key characteristic mediating user engagement (Butler et. al 2014), and most prior works have shown that users in a smaller size group participate more per person. However, contrary to prior research, our results show that the students in larger size cohorts interact more per person and that this greater interaction in turn increases student retention and performance.

Using data from a novel laboratory experiment on complex problem solving in which we varied the structure of 16-person networks, we investigate how an organization’s network structure shapes the performance of problem-solving tasks. Problem solving, we argue, involves both exploration for information and exploration for solutions. Our results show that network clustering has opposite effects for these two important and complementary forms of exploration. Dense clustering encourages members of a network to generate more diverse information but discourages them from generating diverse theories; that is, clustering promotes exploration in information space but decreases exploration in solution space. Previous research, generally focusing on only one of those two spaces at a time, has produced an inconsistent understanding of the value of network clustering. By adopting an experimental platform on which information was measured separately from solutions, we bring disparate results under a single theoretical roof and clarify the effects of network clustering on problem-solving behavior and performance. The finding both provides a sharper tool for structuring organizations for knowledge work and reveals challenges inherent in manipulating network structure to enhance performance, as the communication structure that helps one determinant of successful problem solving may harm the other.

To what extent do social media like Twitter support diversity of ideas and communication between those who hold contradictory viewpoints? Despite the evident potential for social media in this regard, two prominent theories predict that such diverse discourse may be very limited. We review the ideas of

We introduce a new statistic, 'spectral goodness of fit' (SGOF) to measure how well a network model explains the structure of an observed network. SGOF provides an absolute measure of fit, analogous to the standard R-squared in linear regression. Additionally, as it takes advantage of the properties of the spectrum of the graph Laplacian, it is suitable for comparing network models of diverse functional forms, including both fitted statistical models and algorithmic generative models of networks. After introducing, defining, and providing guidance for interpreting SGOF, we illustrate the properties of the statistic with a number of examples and comparisons to existing techniques. We show that such a spectral approach to assessing model fit fills gaps left by earlier methods and can be widely applied.

Using data from a large laboratory experiment on problem solving in which we varied the structure of 16-person networks we investigate how an organization’s network structure may be constructed to optimize performance in complex problem- solving tasks. Problem solving involves both search for information and search for theories to make sense of that information. We show that the effect of network structure is opposite for these two equally important forms of search. Dense clustering encourages members of a network to generate more diverse information, but it also has the power to discourage the generation of diverse theories: clustering promotes exploration in information space, but decreases exploration in solution space. Previous research, tending to focus on only one of those two spaces, had produced inconsistent conclusions about the value of network clustering. By adopting an experimental platform on which information was measured separately from solutions, we were able to reconcile past contradictions and clarify the effects of network clustering on performance. The finding both provides a sharper tool for structuring organizations for knowledge work and reveals the challenges inherent in manipulating network structure to enhance performance, as the communication structure that helps one aspect of problem solving may harm the other.

By lowering search costs, making more products available, and providing access to free content, information technology (IT) has changed patterns of market concentration in media industries such as recorded music. In this paper, I argue that to understand how IT has changed media markets we need to go beyond considering market concentration in terms of the distribution of sales by product, and also consider the distribution of sales by group of products: between-group concentration. With cross-country panel data, I show that IT had different effects on sales that depended strongly on the between-group concentration in each country at the time of the internet shock. I develop and test the hypothesis that the formal industry fared relatively better in markets with more between-group concentration than in markets with less. I conclude with a discussion of lessons and further questions implied by the results.