Insight by Booz Allen Hamilton

Transportation is about to experience a paradigm shift. AI can help federal agencies prepare for it

This content is provided by Booz Allen Hamilton.

The COVID-19 pandemic has accelerated the adoption of new technologies by the federal workforce to a pace that few, if any, predicted. That includes artificial intelligence and machine learning, which many agencies are employing to streamline digital communications in order to help better serve constituents, and to make their own back of house business processes more efficient. And this trend isn’t likely to slow any time soon.

“AI will increasingly impact the global economy over time, improving worldwide economic growth by increasing workforce productivity, quality and efficiency,” said Kathleen Featheringham, director of artificial intelligence strategy at Booz Allen. “Researchers estimate labor productivity improvements in the United States could reach as high as 51% by 2031.”

One area that is particularly ripe for AI adoption is the transportation sector. Featheringham predicted that the application of AI to transportation systems will bring about a paradigm shift akin to the formation of the railroad industry in the early 1800s, or aviation in the early 1900s. Autonomous vehicles for personal and commercial use are just over the horizon, and will soon fill streets and skies across the country.

To prepare for this, federal agencies in the transportation sector need to be investing in AI and ML now, because these technologies will require the ability to collect, parse and operationalize data on a massive scale.

“We have found that there are three points that are key to successful AI adoption,” Featheringham said. “First, the vision for AI must include all parts of your organization, and not just IT. Second, it’s important to have an open meaningful conversation about the right applications for AI and the outcomes that are hoped to be achieved, ensuring that they are ethical auditable sustainable and reliable. And finally, don’t go it alone. It’s critical to include a diversity of stakeholders in this journey that will bring different perspectives, experiences, and backgrounds.”

That’s going to require agencies to begin by educating their workforces about AI, and not just in terms of technical knowledge. Federal employees and managers need to understand the value and outcomes of AI. Further, they will need to understand how expertise in their particular fields can be enhanced through the use of AI. For example, safety inspectors don’t need to develop applications that take advantage of AI or ML. But they will need to understand how it will impact their mission and allow them to make better decisions by leveraging larger datasets.

One thing agencies can do to begin cultivating these skills in their workforce is to think more broadly about the talent pipeline. As AI becomes more mainstream, the educational system will need to begin teaching students how it will apply to their future careers. And government can play a role in empowering them to do so.

“So what are AI and machine learning best suited for now? For detecting patterns, learning to make predictions and recommendations, and processing data directly while learning without a need to have every rule specifically defined,” Featheringham said. “Thinking about that from an outcome perspective, AI is useful for helping with operational efficiency, consistency, safety, and the identification of potential unknowns.”

That will be necessary as the transportation infrastructure transforms to support this new paradigm shift. For example, AI can help predict maintenance needs of our existing complex, aging transportation infrastructure, as well as fleets of autonomous vehicles. Data on system performance and operations can identify maintenance requirements and necessary upgrades before systems are in danger of failing, minimizing risk and reducing downtime.

But all of that requires data that can be trusted, transparent and ethical AI models that allow humans to continually evaluate the solutions generated. To that end, Featheringham described three tenets that should be central to any AI models.

First, AI teams should be meaningfully diverse and inclusive to ensure that society’s most critical problems get the attention they deserve. Second, every AI system should be designed for maximum accountability, explainability and auditability. AI isn’t meant to supplant humans, but to free them from mundane tasks so that they can devote themselves to critical thinking applications. That means humans always need to be able to verify the integrity of their operations, and correct them as needed. Finally, AI systems and applications need to be able to adapt with the field, reacting to new knowledge and advances from the scientific community.

“All of this requires us to move from a reactive use of data to a predictive use of information. Information needs to be readily available to inform better decision making and thereby making transportation, more efficient and safer,” Featheringham said. “Every mile traveled produces data that can be harnessed to improve safety, efficiency and mobility for all.”



Sign up for breaking news alerts