What has been published about ethical and social science considerations regarding the pandemic outbreak response efforts?

Bogdana Rakova
10 min readApr 22, 2020

This summary introduces the results from preliminary analysis of the CORD-19 research dataset and aims to investigate the ethical and social science considerations regarding pandemic outbreak response efforts. In particular, we identify the research articles in the dataset which discuss the potential barriers and enablers for the uptake of public health measures for prevention and control. We also identify articles discussing the implications of public health measures to vulnerable groups such as front-line care providers, elderly, homeless communities, gig workers, and others. We hope that these findings are helpful in highlighting how others have been able to address social and ethical considerations during critical pandemic response efforts. In this way we aim to provide more resources to policy-makers, decision-makers, ethicists who might be part of a hospital ethics committee, or other individuals who need to navigate similar difficult decisions.

Findings

  • The scientific literature has put a lot of attention into studying the social and ethical concerns of public information campaigns, restrictions on internal movement such as city lockdown and quarantine, healthcare interventions and vaccines, and recently the considerations regarding contact tracing.
  • Much less attention has historically been given to the implications of public health measures deployed in certain sociocultural and socioeconomic contexts.

On March 16th, the White House published a call to action to the tech community on a new machine readable dataset of scholarly literature about COVID-19, SARS-CoV-2, and the Coronavirus group (CORD-19). The CORD-19 dataset is a free resource of over 52,000 scholarly articles, including over 41,000 with full text. The dataset has been made available by the Allen Institute for AI, Chan Zuckerberg Initiative (CZI), Georgetown University’s Center for Security and Emerging Technology (CSET), Microsoft, and the National Library of Medicine (NLM) at the National Institutes of Health.

AI research has evolved natural language processing (NLP) and natural language understanding (NLU) algorithms that power many of the conversational AI systems in use today. Many of them are rule-based, for example the queries a user asks to a chatbot are matched to predefined rules, the algorithm is then able to extract the intent of the query and use that information to return a satisfying answer. More complex algorithms utilizing Deep Learning and other methods have shown better results, however, current state-of-the-art systems are still a long way from engaging in truly natural everyday conversations with humans. The NLP/NLU challenges we faced while working on the CORD-19 dataset were similar but also much simpler than the challenges in working with free-form multi-turn language data. For example, we had to consider what techniques could allow us to address common-sense reasoning for understanding concepts as well as context modeling for relating past concepts.

The Oxford COVID-19 Government Response Tracker (OxCGRT) project by the Blavatnik School of Government, University of Oxford, provides systematic updates on the policy actions taken in response to the crisis. The data is continuously updated and publicly available. The OxCGRT policy response metrics framework includes 13 indicators listed below (read the OxCGRT working white paper here). Through our research analysis of the CORD-19 corpus, we aim to further identify the social and ethical concerns related to each of these indicators.

What are the barriers and enablers (blue) for the uptake of public health measures and their implications (red) given the variation in policy responses?

The data vizualization shows the number of CORD-19 references to each of the indicators where the similarity between conepts discussed in the dataset and the OxCGRT indicators was calculated using a Deep Learning model — Tensorflow’s Universal Sentence Encoder model.

In what follows, we highlight the top research references that were found to discuss the social and ethical considerations of each individual policy response indicator. Learn more about the technical details and explore the source code within our interactive notebook in the Kaggle platform.

School closing

Workplace closing

Cancel public events

Close public transport

Public info campaigns

Restrictions on internal movement

Internation air travel controls

Fiscal measures / What economic stimulus policies are adopted?

Monetary measures / What monetary policy interventions are adopted?

Emergency investment in health care

Investment in vaccines

Testing policy

Contact tracing

We are eager to keep working on this project, fine-tuning the results and making them more relevant to concrete policy action questions. Were any of these findings helpful for you and your work? What kind of analysis would you want to see on the CORD-19 dataset?

--

--

Bogdana Rakova

Senior Trustworthy AI Fellow at Mozilla Foundation, working on improving AI contestability, transparency, accountability, and human agency