There's perhaps no technology more hyped than artificial intelligence. But how do we separate fact from fiction?
GovernmentCIO Magazine Editor-in-Chief Camille Tuutti spoke with Hila Mehr, a fellow at Harvard Kennedy School's Ash Center, who recently wrote about artificial intelligence and citizen engagement. Her report highlighted what this technology means to federal agencies and society. Hint: It probably won’t entail a robot uprising, but automating arduous processes and routine tasks, according to Mehr, who serves as IBM’s senior analyst for market development and insights.
GovernmentCIO Magazine: How would you describe the current status for AI?
Hila Mehr: Recent advancements in AI, and specifically machine learning, have brought AI more mainstream. While there are AI applications and use cases to point to across the private and public sector, it still faces many challenges — from logistical to technological, and from ethical issues to efficient and sufficient AI training and training data. There's a lot of hype around AI — and it is exciting — but it's not as common or easy to implement as the hype sometimes makes it seem.
GCIO Mag: You point out in your report that AI comes with many opportunities and challenges. Could you touch on the specific hurdles we see now and now those will evolve in the future?
Mehr: There's a lot of hurdles in the public sector when it comes to AI. While there are many more, I focused on six in the paper and how to navigate them: Making AI citizen-centric; getting citizen input in development and implementation; building upon existing data resources; acquiring, preparing and training data in the right way with a focus on privacy issues; mitigating ethical risk and avoiding AI decision-making; and augmenting employees in their work, not replacing them.
GCIO Mag: Where in government do you see the largest need for AI?
Mehr: Government has a great opportunity with AI when it comes to the unsexy but vital work that makes governments run — the administrative, procedural and data-intensive tasks. Deloitte estimates that automation of federal government employee tasks could save between 96.7 million and 1.2 billion hours annually, with potential savings between $3.3 billion and $41.1 billion, respectively.
GCIO Mag: How should agencies approach AI in the most strategic way? Any specific dos and don'ts?
Mehr: In my report, I outline six strategies — mentioned above — to navigate AI implementation in a strategic way. One of the most important places to start is to really question the goals for using AI in the first place. The question should not be "how will we use AI to solve a problem,” but “what problem are we trying to solve, why and how will we solve it?” If AI is the best means to achieve that goal, then it can be applied, otherwise it should not be forced.
Another thing to beware of is that given the ethical issues surrounding AI and continuing developments in machine learning techniques, AI should not be tasked with making critical government decisions about citizens. For example, the use of a risk-scoring system used in criminal sentencing and similar AI applications in the criminal justice system were found to be biased, with drastic repercussions for the citizens sentenced. These types of use cases should be avoided.
Finally, while any efforts to incorporate AI in the government should be approached as ways to augment human work, not to cut headcount, governments should also update fair labor practices in preparation for potential changes in workplaces where AI systems are in place.
GCIO Mag: Where, in government or outside, have you seen the most innovative uses of AI?
Mehr: One of the examples I mention in the report is a free chatbot lawyer app that helps refugees seeking asylum in the U.S. answer a series of questions to determine which application to fill out and whether they are eligible for protection. With the required information, the bot helps auto-populate the form and provides the applicant with instructions for next steps.
The bot has an added benefit of asking the questions in a straightforward way for forms that would otherwise be confusing for non-experts and non-native speakers.
I love this example because it's AI solving a real problem that a vulnerable community faces, and it makes a complicated task easier for them.
GCIO Mag: Whenever AI is discussed, people also bring up a future where human workers are replaced with automation. You argue for AI to augment workers, not to replace them. What would that machine-human pairing look like?
Mehr: A machine-human pairing could look like AI answering basic, route questions and handing off the more complicated questions that require a human-touch to the employee. Another example would be AI doing a time intensive data analysis, text review, or summary of information, freeing up human time to take those inputs and spend more time coming up with insights and recommendations and taking action on them.