How do social, linguistic, and technological factors influence processes of collective cognition such as collective creativity, problem-solving, judgment, knowledge generation, innovation, evaluation, attention, perception, and interpretation?
How do these collective cognition processes influence organizational and market outcomes?
How do these collective cognition processes influence organizational and market outcomes?
RESEARCH INTERESTS
My research investigates how social, linguistic, and technological factors influence processes of collective cognition, and how these processes then affect organizational and market outcomes. I take groups as the basic units of analysis and apply the principal metaphor of cognitive science—that of cognition as computation—to their operation, tracing their information processing activities in fine detail.
By taking advantage of recent advances in computational and natural language processing techniques, I theorize and empirically trace how collectives innovate, problem-solve, coordinate, evaluate, judge, and decide. I study these patterns of collective cognition in social systems ranging from small groups such as mountaineering expedition teams to large-scale social systems such as markets and scientific disciplines. In so doing, my research bridges multiple disciplinary domains, including organization theory, economic sociology, cognitive science, linguistics, and information theory.
PUBLICATIONS
Evans, James A., and Pedro Aceves. 2016. “Machine Translation: Mining Text for Social Theory.” Annual Review of Sociology.
More of the social world lives within electronic text than ever before, from collective activity on the web, social media, and instant messaging to online transactions, government intelligence, and digitized libraries. This supply of text has elicited demand for natural language processing and machine learning tools to filter, search, and translate text into valuable data. We survey some of the most exciting computational approaches to text analysis, highlighting both supervised methods that extend old theories to new data and unsupervised techniques that discover hidden regularities worth theorizing. We then review recent research that uses these tools to develop social insight by exploring (a) collective attention and reasoning through content from communication; (b) social relationships through the process of communication; and (c) social states, roles, and moves identified through heterogeneous signals within communication. We highlight social questions for which these advances could offer powerful new insight.
More of the social world lives within electronic text than ever before, from collective activity on the web, social media, and instant messaging to online transactions, government intelligence, and digitized libraries. This supply of text has elicited demand for natural language processing and machine learning tools to filter, search, and translate text into valuable data. We survey some of the most exciting computational approaches to text analysis, highlighting both supervised methods that extend old theories to new data and unsupervised techniques that discover hidden regularities worth theorizing. We then review recent research that uses these tools to develop social insight by exploring (a) collective attention and reasoning through content from communication; (b) social relationships through the process of communication; and (c) social states, roles, and moves identified through heterogeneous signals within communication. We highlight social questions for which these advances could offer powerful new insight.
DISSERTATION
Title: “The Linguistic Relativity of Collective Cognition and Group Performance”
Committee: James A. Evans (Chair), John Levi Martin, Amanda Sharkey, Sameer Srivastava
The long-researched linguistic relativity hypothesis predicts that the structure of a person’s language influences their cognition. While this hypothesis has only been pursued in the context of how language affects individual cognition, my dissertation brings this argument into sociological territory by asking how the structure of language influences group performance. The idea is that the structural characteristics of a language can constrain some patterns of communication and collective cognition, while enabling others. I ask: 1) Can differences in language structure affect the performance of groups? If so, 2) what mechanisms account for this performance difference? To answer these questions, I trace the social interaction and collective cognition effects of a novel language structure attribute. I create estimates of information density, the average amount of conceptual information contained within words of a language, across the world’s languages. I then use computational, archival, and experimental methodologies to trace how the information density of a language affects the nature of social interactions and collective cognition as well as the way information flows through those interactions, thereby affecting overall rates of group performance.
Committee: James A. Evans (Chair), John Levi Martin, Amanda Sharkey, Sameer Srivastava
- National Science Foundation DDRI Grant
- Winner, INFORMS/Organization Science Dissertation Proposal Competition, 2017
The long-researched linguistic relativity hypothesis predicts that the structure of a person’s language influences their cognition. While this hypothesis has only been pursued in the context of how language affects individual cognition, my dissertation brings this argument into sociological territory by asking how the structure of language influences group performance. The idea is that the structural characteristics of a language can constrain some patterns of communication and collective cognition, while enabling others. I ask: 1) Can differences in language structure affect the performance of groups? If so, 2) what mechanisms account for this performance difference? To answer these questions, I trace the social interaction and collective cognition effects of a novel language structure attribute. I create estimates of information density, the average amount of conceptual information contained within words of a language, across the world’s languages. I then use computational, archival, and experimental methodologies to trace how the information density of a language affects the nature of social interactions and collective cognition as well as the way information flows through those interactions, thereby affecting overall rates of group performance.
WORKING PAPERS
"Linguistic Relativity, Collective Cognition, and Team Performance"
Team performance research has established how variation in team structure or composition shapes performance outcomes. Both research perspectives posit that the flow of information among group members and between groups and their broader environment constitute primary mechanisms underlying collective performance. Language—the principal technology of information transfer—might be expected to play a critical role in group dynamics and performance. Nevertheless, virtually no research attention has examined the role that language structure plays in shaping group processes and outcomes. In this paper, I create a novel measure of language structure, information density, to examine whether differences in language structure lead to distinct performance outcomes. Language information density is the average amount of conceptual information contained within words of a language. I argue that higher information density facilitates group movement through conceptual space as groups converse, especially during tasks requiring effective search, creativity, problem-solving, and decision-making. Tracing the outcomes of 1,870 monolingual expeditions to the Himalayas from 1907 to 2015, I show that speaking languages with higher information density is associated with summiting a larger proportion of team members, and conditional on reaching the summit, doing so more quickly. I extend insights from this study to explore how and why other contexts of group and organizational activity are likely to be influenced by language information density.
Team performance research has established how variation in team structure or composition shapes performance outcomes. Both research perspectives posit that the flow of information among group members and between groups and their broader environment constitute primary mechanisms underlying collective performance. Language—the principal technology of information transfer—might be expected to play a critical role in group dynamics and performance. Nevertheless, virtually no research attention has examined the role that language structure plays in shaping group processes and outcomes. In this paper, I create a novel measure of language structure, information density, to examine whether differences in language structure lead to distinct performance outcomes. Language information density is the average amount of conceptual information contained within words of a language. I argue that higher information density facilitates group movement through conceptual space as groups converse, especially during tasks requiring effective search, creativity, problem-solving, and decision-making. Tracing the outcomes of 1,870 monolingual expeditions to the Himalayas from 1907 to 2015, I show that speaking languages with higher information density is associated with summiting a larger proportion of team members, and conditional on reaching the summit, doing so more quickly. I extend insights from this study to explore how and why other contexts of group and organizational activity are likely to be influenced by language information density.
"Amplifying and Dampening Social Influence: The Effect of Rating Scales on Social Valuation Dynamics"
This study investigates what happens to social influence dynamics and ultimate valuations when different rating scales are used to generate collective valuations such as product ratings. While prior research has emphasized the role of evaluator or product attributes in shaping collectively generated valuations, I problematize the role played by differences in the valuation tools used. By exploiting millions of beer ratings by thousands of evaluators on two distinct online platforms, I show that the same individuals rate the same beer differently on each platform. I find that when valuations are assigned through scales that are consistent across dimensions (e.g., ratings of 1-5 are used for all product dimensions—e.g., taste, aroma, overall, etc.), evaluators are more socially influenced by prior ratings and the rating system generates higher overall valuations. Alternatively, when ratings are assigned through scales that are inconsistent across dimensions (e.g., appearance is rated on a 1-5 scale, taste is rated on a 1-10 scale, etc.) evaluators are less subject to social influence and the system generates lower overall valuations. These results inform the study of social influence and the ever-prevalent role of (e)valuation in organizational and economic life.
This study investigates what happens to social influence dynamics and ultimate valuations when different rating scales are used to generate collective valuations such as product ratings. While prior research has emphasized the role of evaluator or product attributes in shaping collectively generated valuations, I problematize the role played by differences in the valuation tools used. By exploiting millions of beer ratings by thousands of evaluators on two distinct online platforms, I show that the same individuals rate the same beer differently on each platform. I find that when valuations are assigned through scales that are consistent across dimensions (e.g., ratings of 1-5 are used for all product dimensions—e.g., taste, aroma, overall, etc.), evaluators are more socially influenced by prior ratings and the rating system generates higher overall valuations. Alternatively, when ratings are assigned through scales that are inconsistent across dimensions (e.g., appearance is rated on a 1-5 scale, taste is rated on a 1-10 scale, etc.) evaluators are less subject to social influence and the system generates lower overall valuations. These results inform the study of social influence and the ever-prevalent role of (e)valuation in organizational and economic life.
“Inequality All The Way Down? The Structure of Attention in 20th Century Physics” (with James Evans, Jacob Foster, and Christopher Rhea)
Scientific activity takes place within a scaffolding of social structures, cultural schemas, and material resources. These structures, schemas, and resources interlock in such a way that different scientific disciplines end up exhibiting different distributions of attention. In this article, we explore these patterns of attention through a unique dataset that includes all journals published by the American Physical Society from 1950 through 2013. We use citations as a proxy for attention and characterize the distribution of attention in different subfields of physics with a Gini coefficient. We argue that the concentration of attention in each subfield is shaped by three factors. First, subfields in which there is a significant disconnect between theory and experiments (e.g., string theory) experience a more unequal distribution of attention compared to those in which theory and experiments are tightly coupled (e.g., material sciences). Second, subfields that rely on large-scale experiments with high capital requirements (e.g., experimental particle physics) exhibit a more unequal distribution of attention compared to subfields with lower capital requirements (e.g., optical physics). Third, subfields with large, comprehensive, and integrated theories (e.g., theoretical particle physics) experience higher inequality than subfields with small and disjointed theories (e.g., statistical, nonlinear, or soft-matter physics). We conclude by elaborating on the mechanisms through which these factors lead to the observed distributional outcomes. Understanding how attention is distributed across fields helps elucidate important biases that are part of the present-day scientific endeavor
Scientific activity takes place within a scaffolding of social structures, cultural schemas, and material resources. These structures, schemas, and resources interlock in such a way that different scientific disciplines end up exhibiting different distributions of attention. In this article, we explore these patterns of attention through a unique dataset that includes all journals published by the American Physical Society from 1950 through 2013. We use citations as a proxy for attention and characterize the distribution of attention in different subfields of physics with a Gini coefficient. We argue that the concentration of attention in each subfield is shaped by three factors. First, subfields in which there is a significant disconnect between theory and experiments (e.g., string theory) experience a more unequal distribution of attention compared to those in which theory and experiments are tightly coupled (e.g., material sciences). Second, subfields that rely on large-scale experiments with high capital requirements (e.g., experimental particle physics) exhibit a more unequal distribution of attention compared to subfields with lower capital requirements (e.g., optical physics). Third, subfields with large, comprehensive, and integrated theories (e.g., theoretical particle physics) experience higher inequality than subfields with small and disjointed theories (e.g., statistical, nonlinear, or soft-matter physics). We conclude by elaborating on the mechanisms through which these factors lead to the observed distributional outcomes. Understanding how attention is distributed across fields helps elucidate important biases that are part of the present-day scientific endeavor
