Computational Social Science for Public Policy

Background: Computational Social Science (CSS) utilizes large digital datasets and computational methods to study human behavior, raising ethical concerns about data privacy, informed consent, and potential misuse. Methods: This study employs a constructivist grounded theory approach, analyzing 15 in-depth interviews with CSS practitioners in Germany, Austria, and Switzerland. These countries share a European legal context regarding data privacy and hereby provide a comparable regulatory environment for examining ethical considerations. Results: Findings highlight key challenges in CSS research, including power imbalances with data providers, uncertainties around surveillance and data privacy (especially with longitudinal data), and limitations of current ethics frameworks. Researchers face tensions between established ethical principles and practical realities, often feeling disempowered and lacking support from ethics boards due to their limited CSS expertise. Regulatory ambiguity further discourages research due to fear of sanctions. Conclusions: To foster responsible CSS practices, this paper recommends establishing specialized ethics boards with CSS expertise. It also advocates for acknowledging CSS’s unique nature in research policy by developing tailored data guidelines and providing legal certainty through clear guidelines. Grounding recommendations in practitioners’ experiences, this study offers actionable steps to help enable ethical CSS research.

This chapter is concerned with setting up practical guardrails within the research activities and environments of Computational Social Science (CSS). It aims to provide CSS scholars, as well as policymakers and other stakeholders who apply CSS methods, with the critical and constructive means needed to ensure that their practices are ethical, trustworthy, and responsible. It begins by providing a taxonomy of the ethical challenges faced by researchers in the field of CSS. These are challenges related to (1) the treatment of research subjects, (2) the impacts of CSS research on affected individuals and communities, (3) the quality of CSS research and to its epistemological status, (4) research integrity, and (5) research equity. Taking these challenges as motivation for cultural transformation, it then argues for the incorporation of end-to-end habits of Responsible Research and Innovation (RRI) into CSS practices, focusing on the role that contextual considerations, anticipatory reflection, impact assessment, public engagement, and justifiable and well-documented action should play across the research lifecycle. In proposing the inclusion of habits of RRI in CSS practices, the chapter lays out several practical steps needed for ethical, trustworthy, and responsible CSS research activities. These include stakeholder engagement processes, research impact assessments, data lifecycle documentation, bias self-assessments, and transparent research reporting protocols.

Digital data on human behaviour and social interactions are a seemingly abundant and valuable resource of the twenty-first century, promising deep insights into the social processes generating them. Such data - commonly referred to as digital trace data, process-generated data, or digital behavioural data - are increasingly available to us, but the goal of unlocking their potential remains elusive despite the emergence of specialized fields such as computational social science, which has produced an ample body of research based on this resource. Achieving that goal requires revisiting the foundations of research based on this data type, a thorough understanding of its unique characteristics, overcoming access barriers, understanding data generating processes, and rethinking the role of theory at the intersection between computer science and social science. This chapter discusses the origin and characteristics of such data, as well as the interdisciplinary challenges researchers face in accessing, understanding, and using it for research.

We discuss how computational social scientists can achieve high standards of accessibility, transparency, replicability, and reproducibility by adhering to four open science principles: open practices, open data, open tools, and open access. For each principle, we discuss the underlying motivation, challenges, and potential solutions for computational social scientists. Open practices refer to honest and transparent specification of all steps in the data processing and data analysis. We recommend that computational social scientists share their materials and analysis scripts. Open data refers to data that are publicly available, and we urge computational social scientists to share their data whenever possible. When data cannot be shared, researchers ought to be explicit about the conditions of its availability. Open tools refer to software that is both free and has its source code available to others. We recommend that computational social scientists share their code, software, and educational materials whenever possible so that other people can learn computational techniques. Finally, open access is making the products of one’s research process publicly available. We encourage computational social scientists to use open-access outlets for the ultimate publication of work when feasible and to self-archive their research products regardless.

Addressing the policy challenges and opportunities of “Big data”

Utilizing statistical physics and machine learning to discover collective behavior on temporal social networks

Social networking Web sites are amassing vast quantities of data and computational social science is providing tools to process this data. The combination of these two factors has significant implications for individuals and society. With announcements of growing data aggregation by both Google and Facebook, the need for consideration of these issues is becoming urgent. Just as Web 2.0 platforms put publishing in the hands of the masses, without adequate safeguards, computational social science may make surveillance, profiling, and targeting overly accessible.
 
 The academic study of computational social science explains the field as an interdisciplinary investigation of the social dynamics of society with the aid of advanced computational systems. Such investigation can operate at the macro level of global attitudes and trends, down to the personal level of an individual’s psychology. This paper uses the lenses of computation social science to consider the uses and dangers that may result from the data aggregation social media companies are perusing. We also consider the role ethics and regulation may play in protecting the public.

This chapter outlines the current state and future trajectory of computational social science (CSS), a growing field combining tools and techniques from computer science with theories and methods from the social sciences. This fusion of approaches offers enormous possibilities to expand understanding of human society at scale. We discuss the diverse types of data used in this area of research, including text, image, audio, video, network, location, and sensor data, as well as archival, self-reported, and qualitative data. The chapter offers an overview of key opportunities and obstacles faced by researchers in this field, highlighting structural, ethical, social, and technological factors. Of particular interest at this time is an evolving crisis in data access, as platforms have greatly constricted data availability in recent years. We emphasize the importance of computational methods for theory development and discuss their capacity to expand existing knowledge, adapting classic models for a new information environment.

The usefulness of research infrastructures (RIs) for digital humanities (DH) and computational social sciences (CSS) depends on their capability to support research work tasks. RIs can provide more effective support and services when they have a clear understanding of the work practices and specific tasks researchers are engaged in. This paper explores the ways of working of social sciences and humanities (SSH) scholars to develop resources that truly endorse their data-intensive research processes. Particularly, the study investigates how SSH scholars interact with digital tools and materials to both determine their information needs and how to better support them. A qualitative analysis of 21 semi-structured interviews with potential end-users of a national RI for the DH and CSS that is currently under development revealed three themes: digitizing SSH research, meeting the information needs related to DH/CSS, and supporting DH/CSS research. Based on our findings, we were able to create an understanding of the work task-based requirements of SSH scholars that can be used to inform designs for improved RIs. Suggestions for enhancing the sustainability of resources and services in SSH are also put forth, emphasizing the importance of understanding and accommodating the unique scholarly practices within these fields.

In this poster we describe a pilot study of searching social science literature for legacy corpora to evaluate text mining algorithms. The new emerging field of computational social science demands large amount of social science data to train and evaluate computational models. We argue that the legacy corpora that were annotated by social science researchers through traditional Qualitative Data Analysis (QDA) are ideal data sets to evaluate text mining methods, such as text categorization and clustering. As a pilot study, we searched articles that involve content analysis and discourse analysis in leading communication journals, and then contacted the authors regarding the availability of the annotated texts. Regretfully, nearly all of the corpora that we found were not adequately maintained, and many were no longer available, even though they were less than ten years old. This situation calls for more effort to better maintain and use legacy social science data for future computational social science research purpose.

Computational social science is regarded as a research method that provides an unprecedented breadth and depth of data to classify and analyze how human interaction occurs and provide a different dimension to studies of collective human behavior. Computational social science refers to computational methods and approaches to study the social sciences, particularly incorporating big data, social network analysis, social media content, and spatial data. The outcomes of computational approaches should help overcome the limitations of existing theory in the field. This chapter focuses on the term computational approach and discusses the opportunities and challenges in the social sciences.

BRIEF SESSION DESCRIPTION:Learn to prepare students for interdisciplinary, multidisciplinary, and transdisciplinary work. This panel will explore and discuss the nature of crossdisciplinary work students may experience in their academic and professional careers, and introduce how experienced teachers of interdisciplinary courses engage their students in multi-faceted research here at Mason.________________________________________________________________FULL ABSTRACT: Most big problems facing humanity (give examples to the audience: Climate Change, understanding the Brain and Consciousness, Computational Social Science, and the future challenges as humanity moves into space) are multifaceted, interdisciplinary, multidisciplinary, and transdisciplinary (define these for audience). Also, as in any field of scholarship that involves basic research, the key developments that move a field forward typically come from unexpected directions (e.g. the internet, materials science, etc.). Furthermore, research in these highly interdisciplinary realms typically involves working in large teams of highly varied experts. How do you teach and train students to be aware of the large context in which they will be working in the future and at the same time give the students sufficient depth of understanding in a specific field so that they can make significant contributions to such projects?A‚Â This session will be a short presentation (15 mins) providing definitions and an overview of the nature of interdisciplinary, multidisciplinary, and transdisciplinary work, followed by a panel discussion on the opportunities and challenges of training students in interdisciplinary fields of research. The panel will include people who have substantial experience in teaching interdisciplinary courses and engaging in such multi-faceted research at Mason.Panelists include: Michael Summers (Astrobiology ASTR 301: Teaching the most interdisciplinary of all of the sciences: The origin, evolution, and future of life in the universe)For more information about teaching in the disciplines:A‚Â https://stearnscenter.gmu.edu/teaching/teaching-in-the-disciplinesFor session spacific materials, please request directly from the presenter Mike Summers:A‚Â msummers@gmu.edu

The introductory chapter provides an overview of the scope of the two-volume Handbook of Computational Social Science (CSS), highlighting its key features and contributions. The introductory chapter shows how the two volumes provide the foundation for what CSS is today and points to where it is heading, outlining key debates in the field, showcasing novel statistical modeling and machine learning methods, and also drawing on specific case studies to demonstrate the opportunities and challenges presented in CSS approaches. The introductory chapter also demonstrates the contributions of the two volumes to key debates in CSS, including ethics, theory, artificial intelligence, the role of digitization of life, and data quality. A key goal of the introductory chapter is to provide a concise review of all contributions to the two-volume Handbook, discussing how each chapter contributes not only to the consolidation of computational social science as a field but also to indicate new directions for the future. We end with a set of conclusions and suggestions for future research directions.

Although with some reluctance, social sciences now seem to have accepted the challenge deriving from the growing digitisation of communication and the consequent flow of data on the web. There are actually various empirical studies that use the digital traces left by the myriads of interactions that occur through social media and e-commerce platforms, and this trend also concerns the research in the PCST field. However, the opportunity offered by the digitisation of traditional mass media communication — the newspapers in particular — is much less exploited. Building on the experience of the TIPS project, this paper discusses the advantages and the limits of computational social science on PCST using newspapers as the main source of data. Some methodological issues are also addressed, in order to suggest a more aware use of such data and the several computational tools available for analysing them.

With ad hoc models of individuals that often fail to replicate, social science has been unable to generalize theory into physics models of social reality that benefit the new field of robotic teams. For example, Shannon’s information theory and the social sciences, including economics, assume that the individual observation of individual behavior records the true perceptions of the actual behaviors that have occurred, including for the self-reports of self-observed behavior. In computational social science, this phenomenon has been informally labeled as the “god’s eye view”, indicating that the “computer” within which computational thoughts and actions occur knows immediately whatever thought or action a computational agent holds or takes. In the social sciences this phenomenon allows social scientists to assume that self-reported behavior is actual behavior (e.g., but if true, deception would not exist), justifying the replacement of models of physical behavior with cognitive models. We claim that this assumption is unsupported by the evidence, including the laboratory finding agreeing with religious beliefs that cooperation provides for the best social good. At the heart of these rational, but false, models, interdependence is seen as a constraint (information theory) or experimental confound (cognitive models) that must be overcome to confirm a-theoretical models based on methodological individualism (MI). By replacing MI with quantum-like models, we have found that only a competition among teams establishes social reality; that the sum of non-affiliated individuals, as neutrals, determines the team that best captures reality; and that best performing teams maximize their search of the environment for solutions to the problems they were designed to solve, while the poorer performers search for better mates (e.g., business mergers) or seek to jettison weaker mates (e.g., business spinoffs). We extend these findings with relative entropy to show that poorly performing teams are improperly fitted. We conclude that, like entanglement at the atomic level, interdependence at the social level is the primary resource that ordinary humans exploit to innovate and promote social welfare.