The role of experts in constructing AI's social dimensions

英国政府近日发布了《国家人工智能战略》(OAI，2021)，为英国成为全球人工智能超级大国设定了路径。这一战略的一个关键要素是致力于建立人工智能研究基地。通常认为，那些被归类为人工智能研究人员的人是由计算机科学家组成的；例如，Krafft等人。(2020)在主要人工智能会议上根据自我认同或出版物使人工智能研究人员具有可操作性。这会导致社会科学和人文学科的学者被排除在外--那些接受过研究科学和技术的社会、文化、政治和伦理方面的培训的学者--他们可能会在不同的场所发表论文。如果将社会科学和人文学科排除在外，就会出现问题。如果将人工智能理解为与社会有关的应用背景，并且在概念和方法上被理解为社会在人工智能理论、方法和系统中是如何运作的(Roberge和Castelle，2021)。人工智能和社交之间的这种双重关系意味着，人工智能今天是一股代表和“发明社交”的力量(Marres等人，2018年)--它通过数据密集型应用(例如，在通信中)在阐明什么构成社交互动、社交行为者、社交利益等方面发挥着积极的作用。人工智能参与发明社交的前提提出了关于知识生产中部署的专业知识类型的问题：不仅“在哪里、何时、如何以及为谁开发人工智能”(Joyce等人，2021年)，而且谁在经验、方法、和概念性知识声称在人工智能和社会的交叉点。理解这一点的紧迫性体现在民族国家日益增长的雄心，即通过英国的“国家人工智能战略”等机制，利用人工智能来提高国际影响力。我的研究项目将采用实证的方法，研究当代英国人工智能研究中哪些类型的专业知识构成。具体地说，我的博士研究项目将以三个问题为指导：1.当今哪些类型的专家在科学、工程、社会科学和人文领域就人工智能的社会方面提出知识主张？2.在人工智能研究领域，认知相关性和权威性是如何建立的？3.什么样的知识主张在政策和媒体领域变得有影响力？为了回答这些问题，我将结合科学计量学和计算方法，包括文本分析和(视觉)网络分析，研究专家知识与人工智能辩论相关的三个领域。首先，我将研究学术研究文献，或正式的学术交流，以探索哪些学术领域声称在人工智能的社会维度中具有代表性和干预性。其次，政策文件将被分析为另一个次要领域，在这里，学术知识与人工智能的社会方面的相关性得到确立。第三，将收集Twitter数据，以丰富我的项目对围绕人工智能、社会及其生产的非正式公共专家话语的理解。我的项目旨在通过提供关于人工智能社会方面的学科和知识领域如何建立科学认知相关性和权威性的细微差别来推动科学和技术研究(STS)和知识社会学领域的发展。虽然目前计算机科学家可能在这三个领域占据主导地位，但了解其他类型的专家做出了哪些贡献(以及如何贡献)将有助于理解科学和政治领域的影响力和权力关系的动态，并支持对提高该领域社会科学和人文研究影响力的不同方式的思考。

The UK government recently published its "National AI Strategy" (OAI, 2021), setting out pathways to making Britain a global AI superpower. A key element of this strategy is the commitment to building up the AI research base. It is typically assumed that those categorized as AI researchers are composed of computer scientists; for example, Krafft et al. (2020) operationalized AI researchers based on self-identification or publications in major AI conferences. This has the effect of excluding scholars from social sciences and humanities-those trained to research social, cultural, political, and ethical aspects of science and technology-who might publish in different venues.The exclusion of social sciences and humanities becomes problematic if AI is understood as relating to society as a context of its application and, conceptually and methodologically, in terms of how the social is operationalized in AI theories, methods, and systems (Roberge and Castelle, 2021). This double relation between AI and the social implies that AI is today a force in representing and "inventing the social" (Marres et al., 2018) -it plays an active role in articulating, through data-intensive applications (e.g., in communications), what constitutes social interaction, social actors, social benefits, etc.The premise that AI is implicated in inventing the social raises the questions about the types of expertise deployed in knowledge production: not only "where, when, how, and for whom AI is developed" (Joyce et al., 2021), but also who makes empirical, methodological, and conceptual knowledge claims at the intersection of AI and the social. The urgency to understand this is exemplified by the growing ambitions of nation-states to capitalize on AI for international influence, through mechanisms like the UK's "National AI Strategy".My research project will take an empirical approach to researching what types of expertise constitute AI research in the contemporary UK. Specifically, my doctoral research project will be guided by three questions:1. Which types of experts today make knowledge claims about social aspects of AI across the sciences, engineering, social science and humanities?2. How is epistemic relevance and authority established in the AI research field?3. What kinds of knowledge claims become influential in policy and media domains?To answer these questions, I will combine scientometric and computational methods, including text analytics and (visual) network analysis to examine three arenas where expert knowledge becomes relevant to AI debates. First, I will examine academic research literature, or formal academic communication, to explore which academic fields claim representation and intervention in the social dimensions of AI. Second, policy documents will be analysed as an additional secondary arena where the relevance of academic knowledge in relation to social aspects of AI gets established. Third, Twitter data will be collected to enrich my project's understanding of informal public expert discourse around AI, the social and its production.My project aims to advance the fields of Science and Technology Studies (STS) and sociology of knowledge by providing a nuanced account of how scientific epistemic relevance and authority is established across disciplines and knowledge arenas on social aspects of AI. While at present computer scientists might be dominating all three arenas, understanding which other types of experts contribute (and how) will help to understand dynamics of influence and power relations across science and politics, and support reflection on different ways for making social sciences and humanities research in this area more impactful.