Using Innovative Data Analytics to Improve Government Services

Sanith Wijesinghe is a lead in the Modeling, Simulation, Experimentation, and Analytics Innovation Center and one of MITRE’s newest Technical Fellows. He works with the design of computational algorithms and has extensive experience developing and operating trading systems for the capital markets industry. He brings economic and game theory to bear on key challenges and innovation aspects of science and technology and the Great Power Competition.  

Q: Sanith, can you say a bit about yourself and what you do at MITRE?

I lead MITRE’s Agile Connected Government (ACG) research portfolio, within our independent research program. Over the last couple of years ACG has focused on developing novel analytics, processes, and frameworks to help government agencies improve their customer service offerings.

Delivering customer services tailored to the unique needs and circumstances of citizens is a high priority for our sponsors. However, there continues to be a mismatch between the desired outcomes and the systems that have been implemented.

The public has come to expect the same level of customer service standards from government as offered by private industry. Unfortunately, flat or declining budgets have kept agencies from acquiring the required infrastructure and modern technology to remain on par.

Our sponsors see this as a real challenge, and my team looks for ways to help them meet that challenge.

Q: You said you can use “novel analytics” to improve government services. Can you explain that?

Government agencies are experimenting with new kinds of data analytics, machine learning, and artificial intelligence techniques to determine ways to improve their customer service. Examples include using automated chatbots and voice agents to minimize call wait times and statistical adjudication algorithms to accelerate benefits application reviews.

The next step is to deliver anticipatory services. These are services that are proactive, personalized, and consider a citizen’s circumstance and background. To do that, however, agencies must capture, merge, and update customer service data elements accurately and quickly.

In the private sector, customer service data capture is occurring at increasingly faster speeds and in larger volumes. But it lacks the privacy and security protections necessary to protect the public interest. Public sector agencies must chart a different course to build citizen trust and avoid potential backlashes.

Q: Are there ways the government can improve tax administration with analytics?

Analytics offer the opportunity for tax administrators to do two things. First, they can improve overall mission value through increased efficiency and reduced costs of service delivery. Second, they can increase value to customers through improved efficacy of service interactions.

By identifying patterns and trends in service use, segmented by taxpayer demographics, tax administrators can better forecast service demand spikes, identify opportunities for early customized outreach, and better optimize internal resource allocation.

For example, simulation-based approaches can explore how taxpayers might respond to a new tax policy and drive insight about areas of potential noncompliance.

Q: You’ve mentioned machine learning and simulation techniques. Can you explain the similarities and differences?

Machine learning describes a class of techniques that use automated means to determine associations and correlations between underlying data elements. In recent decades, the advent of big data and scalable, cheap computing have accelerated machine learning in a variety of disciplines, ranging from computer vision to natural language processing, and even drug discovery.

Simulation, on the other hand, belongs to a class of techniques that don’t require, as a pre-requisite, large quantities of underlying data. Instead, simulation begins by prescribing rules that define the actions of specific entities—such as taxpayers and auditors—and then explores different future states by evolving behaviors over time.

Both approaches are useful to discover patterns and anomalies that wouldn’t otherwise be “guessed” by subject matter experts.  

Q: What role is there for using simulation and simulated data in your work?

Simulation and simulated data are great ways to conduct “what-if” experiments. Prior to deploying a new service offering and conducting field experiments, for example, simulation-based techniques would help in the research phase to explore potential risks and unintended outcomes.

These early explorations can help better define the parameters for field experiments in a more real-world setting. Simulation can thereby help reduce costs and accelerate time to overall solution deployment.

Q: Are there other areas of government where analytics can lead to better services?

An important area where the government can expand the use of analytics is monitoring and analyzing public risks to safety. In particular, expanding early-warning systems for citizens to better avoid or mitigate negative outcomes from natural and human disasters, epidemics, and financial shocks will be an increasingly important service to consider.

Separately, the use of analytics can, by itself, lead to detrimental outcomes due to potential bias and lack of transparency. If appropriately formulated, however, analytics might also help minimize existing social inequity and injustice.

These latter efforts are the focus of current research in my group’s innovation area, and I welcome feedback and collaboration on how MITRE can best help our sponsors.

—interview conducted by Gregory Michaelidis