Text as Data is a remarkable milestone in demonstrating how analysis of large text-based corpora can enable original ways of theorizing in the social sciences. While the authors provide a thorough overview of algorithms and models, these features are not the book’s main attractions. The real value of this book stems from its insights into research principles for scholars’ use of textual data, enriched by numerous illustrations based on published studies. I believe the book will inspire organizational theorists to develop novel types of research questions and to deploy inventive approaches to examining social phenomena.
The book offers two other noteworthy contributions. First, “a central argument of this book is that the goal of text as data research differs from the goals of computer science work” (p. 5). In articulating this difference, the authors “reject the idea that models of text should be optimized to recover one true underlying, inherent organization in the texts—because, [they] argue, no one such organization exists” (p. 9). This statement embodies what the authors call the agnostic approach to textual analysis. Second, the authors repeatedly emphasize the importance of validation when working with textual data, because “there are few (if any) theorems that justify text analysis methods as applied to natural language” (p. 30). With that, they underscore that validation which maintains “humans in the loop” (p. 31) is fundamental for good research in the social sciences. This agnostic approach and the call for extensive validation and human interpretation draw an important line between social scientists and computing scientists. The distinction is crucial because it reminds organizational theorists that our primary focus should be theorizing and that computational tools, regardless of how shiny they might be, are just that: tools.
Text as Data is a remarkable milestone in demonstrating how analysis of large text-based corpora can enable original ways of theorizing in the social sciences. While the authors provide a thorough overview of algorithms and models, these features are not the book’s main attractions. The real value of this book stems from its insights into research principles for scholars’ use of textual data, enriched by numerous illustrations based on published studies. I believe the book will inspire organizational theorists to develop novel types of research questions and to deploy inventive approaches to examining social phenomena.
The book offers two other noteworthy contributions. First, “a central argument of this book is that the goal of text as data research differs from the goals of computer science work” (p. 5). In articulating this difference, the authors “reject the idea that models of text should be optimized to recover one true underlying, inherent organization in the texts—because, [they] argue, no one such organization exists” (p. 9). This statement embodies what the authors call the agnostic approach to textual analysis. Second, the authors repeatedly emphasize the importance of validation when working with textual data, because “there are few (if any) theorems that justify text analysis methods as applied to natural language” (p. 30). With that, they underscore that validation which maintains “humans in the loop” (p. 31) is fundamental for good research in the social sciences. This agnostic approach and the call for extensive validation and human interpretation draw an important line between social scientists and computing scientists. The distinction is crucial because it reminds organizational theorists that our primary focus should be theorizing and that computational tools, regardless of how shiny they might be, are just that: tools.- Reproduced

https://journals.sagepub.com/doi/full/10.1177/00018392241239340