• Culture, Theory, Data:An Introduction
• Ted Underwood (bio), Laura McGrath (bio), Richard Jean So (bio), and Chad Wellmon (bio)

Culture. Theory. Data. In that group of three terms, the obvious interloper is data—a word not widely used in the humanities until this century. Not that humanists ever lacked data. Our sources and archives are data in the oldest sense of the word. Daniel Rosenberg's careful history of data concludes that it has long been "a rhetorical concept," meaning simply "that which is given prior to argument."¹ Of course, evidence is never given in an absolute sense. As Lisa Gitelman and Virginia Jackson have reminded us, every number in a table is collected and constructed by human hands.² But these constructions are converted into data when someone provisionally accepts them as given for the sake of a particular argument. Historical sources play a similar rhetorical role, and Rosenberg shows that the seventeenth century discussed them in a similar way, writing about "historical data" and even "scriptural data."³

By the late eighteenth century, however, data was specializing to contexts like "medicine, finance, natural history, and geography."⁴ By the twentieth century, it had become strongly associated with the sciences and with systematic, numeric recordkeeping. Economists certainly had data. But historians believed they had data only when they were working with a document like a ledger or a census. The word felt alien enough that it was often invoked to mark a divide between the humanities and social sciences. In 1979, for instance, Tony Judt pushed back against social-scientific work in history by remarking that such projects "resort to quantified and quantifiable data to compensate for the lack of an argument and the glaring absence of conceptual insight."⁵ If "conceptual insight" is akin to theory, Judt saw data as a poor substitute.

By the 1970s, data was becoming a metonym not only for the evidentiary procedures of science but for computers in particular. Lawrence Stone, criticizing quantitative work in the humanities around the same time as Judt, envisioned a scene where "squads of diligent assistants assemble data, encode it, programme it, and pass it through the maw [End Page 519] of the computer."⁶ Computers may seem less alien today than they did in 1979 (at any rate, they less commonly have a "maw"), but they are still strongly associated with the word data. So including data in the title of this special issue is probably enough to suggest that its essays will discuss the role that computational methods and digital media play in contemporary research on culture. Which they do.

On the other hand, as computers become increasingly embedded in daily life, data has expanded to cover a much wider range of things than it did forty years ago. The word still tends to imply that the information it describes is numeric or can be turned into numbers. But these days, what can't be turned into numbers? Even our recipes and home movies are transmitted digitally. Moreover, contemporary statistical models are no longer restricted to overtly quantitative problems with a small number of variables.⁷ Machine learning can create variables as needed to model images, social networks, and unstructured text. Most of us regularly use models created this way to answer questions and find documents. Studying culture with data is no longer a specialized practice requiring "squads of diligent assistants" who "encode" variables for analysis. It is something most scholars already do, consciously or not.

For that reason, the title of this issue is not something like "Culture, Theory, Digital Humanities." Instead of advancing a subfield, we hope to explore theoretical questions that may be relevant to anyone studying culture, whether they use computational models explicitly in their research or "only" to translate unfamiliar languages, navigate social networks, and find sources. In shaping the event that brought these authors together (a symposium in Charlottesville, Virginia in May 2022), we invited people who work in media studies, information science, sociology, and anthropology, as well as historians and literary scholars. Literary history was often the center of our conversation, as you might expect from the title of this journal. But we came to...

• 文化、理论、数据：简介
• Ted Underwood（生物）、Laura McGrath（生物）、Richard Jean So（生物）和 Chad Wellmon（生物）

文化。_ 理论。_ 数据。_ 在那组三个术语中，明显的闯入者是数据——这个词直到本世纪才在人文学科中广泛使用。并不是说人文主义者从不缺乏数据。我们的资源和档案是最古老意义上的数据。丹尼尔·罗森伯格 (Daniel Rosenberg) 对数据的仔细研究得出的结论是，它长期以来一直是“一个修辞概念”，意思只是“在争论之前给出的东西”。¹当然，从来没有绝对意义上的证据。正如 Lisa Gitelman 和 Virginia Jackson 提醒我们的那样，表格中的每个数字都是由人手收集和构建的。^2个但是，当有人出于特定论证的缘故暂时接受它们时，这些结构就会转换为数据。历史资料起着类似的修辞作用，罗森伯格表明 17 世纪以类似的方式讨论它们，写下“历史数据”甚至“圣经数据”。^3个

然而，到 18 世纪后期，数据专门用于“医学、金融、自然历史和地理”等背景。⁴到二十世纪，它已与科学和系统的数字记录保持密切相关。经济学家当然有数据。但历史学家认为，只有在处理分类账或人口普查等文件时，他们才有数据。这个词感觉很陌生，以至于它经常被用来标记人文科学和社会科学之间的分界线。例如，在 1979 年，托尼·朱特 (Tony Judt) 反对历史上的社会科学研究，他评论说此类项目“诉诸量化和可量化的数据，以弥补缺乏论证和明显缺乏概念洞察力”。^5个如果“概念洞察力”类似于理论，那么朱特将数据视为糟糕的替代品。

到 20 世纪 70 年代，数据不仅成为科学证据程序的代名词，尤其是计算机的代名词。劳伦斯·斯通 (Lawrence Stone) 与朱特 (Judt) 大约同时批评人文学科的定量工作，他设想了这样一个场景：“一队勤奋的助手收集数据，对其进行编码、编程，然后将其传递给计算机 [End Page 519] 。” ⁶与 1979 年相比，今天的计算机可能看起来不那么陌生（无论如何，它们不常有“花胶”），但它们仍然与数据这个词密切相关。所以包括数据本期特刊的标题可能足以表明其文章将讨论计算方法和数字媒体在当代文化研究中的作用。他们这样做。

另一方面，随着计算机越来越多地融入日常生活，数据已经扩展到比四十年前更广泛的范围。这个词仍然倾向于暗示它描述的信息是数字的或者可以变成数字。但是现在，什么不能变成数字呢？甚至我们的食谱和家庭电影都是以数字方式传输的。此外，当代统计模型不再局限于具有少量变量的明显定量问题。⁷机器学习可以根据需要创建变量来对图像、社交网络和非结构化文本进行建模。我们大多数人经常使用以这种方式创建的模型来回答问题和查找文档。用数据研究文化不再是需要“勤奋的助手团队”对变量进行“编码”以进行分析的专业实践。这是大多数学者有意或无意地已经在做的事情。

因此，本期的标题不是“文化、理论、数字人文”之类的。我们希望探索可能与任何研究文化的人相关的理论问题，而不是推进一个子领域，无论他们是在研究中明确使用计算模型，还是“仅”用于翻译不熟悉的语言、浏览社交网络和寻找资源。在塑造将这些作者聚集在一起的活动（2022 年 5 月在弗吉尼亚州夏洛茨维尔举行的研讨会）时，我们邀请了从事媒体研究、信息科学、社会学和人类学工作的人，以及历史学家和文学学者。文学史通常是我们谈话的中心，正如您从这本杂志的标题中所预料的那样。但是我们来到了...