Launching a Computational Social Welfare Lab: Reflecting on the Benefits and Obstacles Identified in the Grand Challenge

Background and Purpose: The GBIC Computational Social Welfare Lab aspires to be a world-class environment where interdisciplinary researchers and policy practitioners can combine their domain knowledge of social welfare with big data and advanced data science methods to address complex social problems and promote social wellbeing. However, finding resources to guide GBIC actions in this endeavor has been difficult. Published commentary about and recommendations for using big data for social good are available. This includes work by scholars that build the case for computational social science, scholars who reflect on the risks of wide-spread data sharing and merging, and best practices in computing. But empirical evidence about the effects of adopting big data on social welfare research or on the individuals and communities served by social work is thin. And a manual that articulates the steps for building a computational social welfare lab does not seem to exist. This paper aims to present the steps taken by GBIC to launching a computational social welfare lab and reflect on these activities in light of the recommendations in “Harnessing Big Data for Social Good.”

Methods. This paper reviews the literature on computational social science and uses a case study approach to describe and critique the activities used by GBIC to launch a lab.

Results: Since 2017, the GBIC Lab has (1) worked with government officials to identify and transfer data, (2) built a secure mechanism for data storage that can accommodate constant data acquisition and which produces continually growing longitudinal data resources, (3) established data privacy and confidentiality procedures that have been IRB approved, (4) developed and executed standardized and replicable workflows for data cleaning and merging including translation of all numeric and text data from Chinese to English, (5) created a library of computational code for building various models that can be quickly activated to mine lab data in response to inquiries from government officials, researchers, industry partners, and other collaborators, (6) acquired staff and equipment that supports efficient and effective lab functioning, (7) hosted collaborating scholars, including joint authorship of scholarly articles, and (8) provided classroom and lab-based data science training. Challenges have included (1) establishing data sharing agreements, especially in a setting of international cooperation, (2) managing delays with data sharing, (3) recognizing data errors, (4) interpreting data in the absence of data dictionaries and the common presence of unstandardized text fields, (5) understanding generalizability of findings, (6) recruiting and retaining qualified staff, and (7) meeting government timelines and needs that are incongruent with a research agenda,

Conclusions and Implications: The GBIC Computational Social Welfare Lab is a work in progress. The launch to date has included successes and missteps. This paper will be of interest to other researchers engaging in computational social welfare. And it hopes to begin a discussion within social welfare about the benefits, risk, and mechanics of embracing big data and data science via the establishment of computational labs.