Deputy Federal CIO Maria Roat explained today that Federal IT policymakers are devoting considerable effort toward how to share and otherwise re-use across government the applications of automation technologies that Federal agencies have been developing on their own.
During a keynote address at ACT-IAC’s Digital Transformation Summit, Roat said a range of automation technologies including robotic process automation, machine learning, and blockchain “are really demonstrating their promise in various functions across the government,” and meeting the pressure being felt by government to employ technology to reduce backlogs, promote transparency, and increase efficiency.
Good real-world examples of the benefits of those technologies include e-discovery applications that can locate 95 percent of relevant documents in the discovery phase of legal cases, and “allow frontline workers to dramatically reduce” time and energy spent on administrative work, she said.
Similarly, Roat cited recent deployments of drone aircraft to better track hurricanes, and work by the Environmental Protection Agency (EPA) to use machine learning to improve the processing of record schedules that now reside in the form of millions of paper files and uncategorized digital records.
“This is really a great example where one effort at an agency can potentially be applicable to other agencies across the Federal government,” she said.
“This is the kind of stuff that can get reused across the Federal government,” she continued. “It’s impactful, it hits the mission, and it significantly reduces backlogs. You get so many efficiencies with that, and it’s something that can be shared.”
Along those lines, Roat said the Federal CIO Council is leading agency efforts to inventory AI use cases, share use-case inventories with other agencies, and make those inventories available to the public.
“This work is just getting started so it’s certainly going to take some time to do it,” Roat said. “But this is really important work so that we can get Federal-wide, governmentwide sharing, continuing to view the federal government as an enterprise. This will continue to encourage the secure sharing of data across the federal government agencies.”
“I think sharing information across agencies is critically important so that you’re not reinventing the wheel all the time and you learn from each other,” Roat said. “And I think that is really, really important as we’re moving into AI use cases.”
‘Early Adopter’ Excitement
Roat – who became Deputy Federal CIO last year and as such is closely involved in governmentwide IT policy development – was effusive in her interest in the ability of automating technologies to increase high-value work, and on the experiences of early adopters of the technologies.
“This is where I think I’ve lived most of my career pushing the envelope, living in that early adopter space,” she said.
“When we’re talking about technology, that gets me excited,” she said. “Sitting where I sit as the deputy Federal CIO I get to see so much of what’s going on in the IT space across the government and it is just really cool to see all the adoption, all these new technologies, and how that computing power is allowing us to really take advantage of that and really benefit the mission of the Federal government.”
Hunting Unintended Bias
The deputy Federal CIO also included a note of caution in her remarks about unintended biases that can exist within the design of automated technologies, algorithms, and the data they rely on. “For the Federal government, we need to understand those algorithms, and we need to be able to explain those … because of those biases that could be underlying within those algorithms,” she said.
To that point, Roat explained work underway at the Department of Labor to ensure that human resources-related AI is not unintentionally biased against job seekers with disabilities when it uses video and audio analysis of candidates during job interviews.
“Think about this on unintentional biases – if someone’s facial attributes or mannerisms are different than what’s defined as the norm, they get no credit [in the AI evaluation], even if their traits and their skills could be beneficial to the job,” she said. “So without that reverse training for that AI model, the model would not be able to learn any characteristics demonstrated by people with disabilities who are later successful.”
“The project’s goal is to build a common language around equitable and inclusive AI, and, and they want to provide practical resources for AI tech implementers,” she said.
On the data front, Roat said “there’s so much going on” with the Federal Data Strategy and the Chief Data Officers (CDO) Council, and added, “we have to make sure we enhance” access to data and “make sure it’s usable Federal data for the models … we need to make sure that the data is good.”