Agencies ‘flying blind’ without more data
Connecting state and local government leaders
A two-year analysis warns that policy makers lack the data to inform their responses to workforce changes caused by key technologies.
The federal government needs better tools and datasets to evaluate the impact of advances in artificial intelligence and other emerging technologies on the public-sector work force.
That’s one of the critical findings of a recent report on “Information Technology and the U.S. Workforce,” by the National Academy of Sciences’ Committee on Information Technology, Automation and the U.S. Workforce.
A two-year analysis by the group warned that policy makers are “flying blind” without data to inform their responses to workforce changes caused by AI, driverless cars and other emerging technologies.
“The top-level problem is that technical advances, especially in information technology, are causing big changes in the workforce and will disrupt the employment of a lot of people in ways that policy makers should be aware of,” said Tom Mitchell, an author of the report and member of the Machine Learning Department at Carnegie Mellon University.
“The biggest surprise to me in the two-year study that we did was that the government was essentially lacking visibility into what’s going on,” Mitchell said. “It doesn’t have enough data.”
Policymakers don’t have the means to address even basic questions, he said, such as what jobs are being most affected by automation. “We don’t have the data to answer this basic question,” he said, “and so it’s difficult to make policy under these conditions.”
Historically the government relies on the availability of in-house data it collects as a part of agency missions. “The Bureau of Labor Statistics collects and publishes unemployment and other types of data,” Mitchell said, “but that data alone doesn’t answer the questions that need to be answered here.”
Instead, government policymakers should consider mixing in data resources now being produced, stored and put to use by private firms, Mitchell said.
“The good news,” he said, is that over the past 15 or 20 years, private organizations -- non-government organizations in the U.S. and beyond -- have learned the lesson of collecting lots of data and using it to improve their own decision-making by putting more and more data online.”
“The path toward a solution is for the government to adopt new strategies for partnering with non-government organizations that have a lot of data and finding ways to combine that data with data from other governmental organizations to build a much more effective information store for guiding policy,” Mitchell said.
One type of organization with enough data to help address these questions are universities, Mitchell said. “They have a lot of information, including data on what kind of skills students who are about to graduate have.”
Beyond universities, viable private-sector data sources include LinkedIn, Monster.com and Burning Glass, all with huge holdings of data pertaining to resumes and job postings. “These companies get a tremendous amount of information about exactly which jobs are being sought and what kind of jobs are being posted.”
“Government really needs more insight into which technologies are causing changes in the workforces,” Mitchell said. It must be able to identify the changes, the kinds of jobs most affected, the skills workers will need in a more automated economy.
This data is available, he argued. “A call for proper metrics or tracking AI technologies by building a comprehensive AI index would provide objective data on various AI developments,” he said.
Information measuring the demand for the underlying technologies is also part of the puzzle.
“In the AI field, you might get quantitative data about how many industrial robots are actually getting purchased. You would also want data on the supply of talent in AI, he added. “How many students are graduating with competence in AI, and how many of them are being demanded on the hiring side?
Mitchell suggested government data scientists run small-scale experiments, the way the private sector does, to gauge technology’s effect on the workforce.
“Run experiments and find out what really does happen,” he said. And then instead of making decisions in “a smoke-filled room,” government could use grounded evidence, the results of experiments and much more rational procedures in crafting policy, Mitchell said. “If one point comes through I hope that’s it,” he said.