Does the Pandemic Reframe Political Data Science and AI in the Public Sector?

That startling event came in the form of the COVID-19 outbreak. The first high-level unifying global event to occur in the age of big data and machine learning, the coronavirus pandemic required urgent statistics-based analysis that would soon comprise an unprecedented international effort to quantify, analyze and interpret a new phenomenon.

With no other useful tools immediately available, data science consulting in politics has received a locus of interest, funding and intra-sector cooperation across the world, which seems unlikely to have occurred in any other circumstances, and which has endowed the field with a global showcase and a disruptive environment that is changing the previous circumspect approach to public sector AI — perhaps even for the long term.

The remaining four challenges from the OECD have been largely overcome in the current urgent spirit of global technological collaboration: the report recommends 'Promoting synergies across public sector organizations', as well as 'Jointly participating in international and/or regional projects', 'Stepping up collaboration with the private sector' and 'Experimenting and piloting' agile approaches to emerging technologies such as machine learning.

Taking into account platforms such as the American Artificial Intelligence Initiative, Germany's Artificial Intelligence Strategy, Britain's appointment of a Ministry for Artificial Intelligence, and various EU public sector AI projects, the report's core challenges  are characterized in terms of long-standing departmental rivalry; unrelated or incompatible data frameworks; a culture of protectionism among government sectors; a circumspect relationship between government and the private sector; and industry leaders' reluctance to pool their resources outside of the reciprocal benefits of open source development.

The coronavirus crisis has cut through nearly all these barriers at a stroke:

• Two of the world's largest pharmaceutical manufacturing rivals, GSK and Sanofi, have joined forces to develop a COVID-19 vaccine.
• Rolls Royce has gathered together government scientific advisors and notable IT industry leaders (such as IBM and Google Cloud) in the EMER2GENT initiative.
• Germany's government united 28,000 participants from the private and civic sector to organize a COVID-19 Hackathon that has produced 1,500 possible solutions to date.
• Italy's Presidency of Council of Ministers at the Department of Civil Protection has provided an open data statistical analysis dashboard.

• The United Nations Global Working Group on Big Data for Official Statistics (in association with the IMF and the World Bank) has developed a cloud-based ecosystem to support international statistics projects, with a stunning range of government and private sector collaborators from around the world.
• Industry rivals Apple and Google have created on an open-source contact tracing app framework with a common code-base for their respective platforms, now being implemented by various governments across the world.
• Pressure is intensifying for more unified federal oversight frameworks to cope with COVID-19 and future national emergencies, notably from the Center for American Progress (a clamor for cohesive strategy that has already seen congressional progress toward national AI regulation for self-driving cars in the USA).
• 24 universities from countries including Italy, UK, Germany, France, Switzerland, Austria, Spain, Slovenia, Romania, and Ireland are developing a machine learning model to unify and interpret global logistical and policy response to COVID-19.
• Facebook’s AI and data science experts have collaborated with Columbia University to produce new maps of COVID-19 hotspots for relief workers, with the data made available on UN OCHA's Humanitarian Data Exchange and Amazon’s AWS Open Data Sets.

The growing global agenda toward protectionism has been put aside, at least in terms of the need for cross-sector and international cooperation in data science and related artificial intelligence development.

Public/private collaborations that might have gone out to tender at length have been formed in weeks, sometimes days, on the most casual terms; committee-driven or political resistance has all but evaporated, as fully funded initiatives are forming organically and rapidly; and, generally, the public health mandate overrides nearly all other concerns.

The 2020 survey of AI use across the US Federal Government has found that machine learning is already implemented across a broad range of public areas, and especially in the police sector.

The Federal Data Action Plan also envisions a new era of cross-agency collaboration enabled by data science:

A November 2019 report by the OECD found that 36 of the 50 countries studied had active or pending strategies for integrating AI into public sector services and practices. Though these projects are too numerous to list, we can look at some current schemes:

• China's COVID-19 tracking system for citizens
• Pandemic detection in Australia with NLP
• Named Entity Feature Representation (NEFR) enabling unsupervised public health event detection in Germany
• The Paris Metro authority using facial recognition to determine if passengers are wearing PPE.

• A US-based early warning system for adverse drug reactions with NLP techniques
• Australia predicting patient readmission with convolutional neural networks
• The State of Washington using machine learning to analyze urban neighborhoods for levels of 311 (non-emergency) citizen calls to government
• Under a government scheme in New Delhi, AI analyzing vehicle movement patterns to manage traffic  
• The World Bank demonstrating the use of Latent Dirichlet Allocation (LDA) in NLP to correlate presidential speeches with civil insurgence (image below)

• AI-driven predictive scores are being used to plan election campaign fundraising in the US, as well as campaign tour strategy
• Machine learning and behavioral analytics are used to identify potentially fraudulent benefit claims in the UK and the US
• The US Bank uses ML for onboarding and fraud detection
• The state of Maryland has improved its tax fraud detection rate with analytics modelling
• Machine learning as a tool against election fraud is a notable area of study in the USA

Further examples of active public sector machine learning implementations abound in education, parking enforcement, the military, crime prevention, and an ever-increasing range of government departments.