How government agencies can better use data to make decisions
- Government agencies often run into speed bumps when trying to adopt data-driven decision models.
- Instead, they should use an AI-driven analytics system powered by a knowledge model to sort the data.
- It helps businesses make smarter decisions and achieve scalable results.
Digital transformation is driving new innovations in government. It relies on complete and accurate data, but agencies often run into slowdowns when trying to adopt data-driven decision models.
Indeed, it is not enough to integrate data sources and feed information into large-scale analytics systems to become a data-driven agency. Agencies and their data scientists need to make sure they don’t drown in a flood of data they can’t use, said Forrest Hare, developer of cyber operations solutions at SAIC.
“It’s easy to get data overload and decision paralysis when you’re too good at integrating your data,” he added. Agencies need to find an easy, repeatable, and scalable way to distill the numbers into meaningful knowledge.
“Fast data-driven decisions require automated analysis,” Hare said. “Without it, you need lengthy, personalized analyzes for every decision.”
When thinking of automated meaning creation – also known as machine-based understanding – data scientists often look to artificial intelligence (AI). Its ability to make patterns appear in large amounts of data makes it possible to navigate through a large amount of incoming information, extracting nuggets of useful information.
Pointing a neural network at a large amount of data can generate some patterns, but that won’t help the meaning-making part, Hare said. Without an understanding of this data, AI will bring up patterns, but nothing more. He can’t deduce their real-world implications because he doesn’t understand what they mean. He only sees them as collections of zeros and ones.
The components of a data-driven culture
To make sense of all this data, we need to pair it with understanding – and a good culture starts with two things: a knowledge model and data management.
Our databases contain the assets we want to mine, but our understanding of that data relies on human subject matter experts who have spent years gaining implicit knowledge in their areas of specialization. In most cases, this knowledge stays locked in their heads. The good news is that we can unlock its value by extracting it through a knowledge model and formally representing it.
A knowledge model encodes the knowledge gathered about a particular field or discipline by mapping the elements it contains and how they relate to each other. We need to produce these models in formats that machines can read, allowing them to absorb and use some of this human understanding.
An AI-based analysis system driven by a knowledge model can infer things about the models it sees. It can answer questions about domain-specific entities such as hospital readmission rates of patients or the reliability of intelligence sources. AI can use the knowledge model to extract the right information from the data.
The other essential part of a data-driven culture is data management.
“We need data security through effective governance that leaves our data organized and cleansed,” says Hare. “This leads to better use of AI, producing decisions that are not biased.”
How to achieve your data-driven goals
Reaching this utopia of machine-based understanding takes work. This means instilling a culture of openness in your agency, encouraging the use of open standards and protocols for better integration. This will allow you to integrate agency data solutions with domain-specific knowledge models to encourage better understanding.
This open culture is also based on collaboration between users and technology experts. “Collaboration leads to better results in digital transformation,” Hare said. “The end users of the data know best how they want to use it and that should determine how you structure the knowledge models. “
One way to encourage collaboration is to use appropriate data sharing and discussion tools, but these alone will not be enough. Buy-in from senior management on the operations side is crucial if employees are to work together to build and use knowledge models.
SAIC has its own methodology for creating knowledge models called COSIKnE. He began to use this framework to transform the way elements of the US military look at their data. Any agency can benefit from encoding human understanding into their knowledge framework, according to Hare, and by improving the use and governance of data through close collaboration. On the journey to becoming a data-driven agency, that just might mean the difference between success and failure.
Find out how to make your agency more data-driven.
This post was created by Insider Studios with SAIC.