LMI Workshop Examines Data Ethics for Government

These were some of the questions tackled by panelists from academia, government, and industry at a workshop hosted by the LMI Research Institute (LRI). Operationalizing Data Ethics: Considerations for the AI/ML Development Lifecycle explored how data ethics principles may be implemented, enforced, and institutionalized to mitigate risks such as inherent discrimination and the misuse of data or analytical outcomes.

One consensus is that data governance frameworks must be adapted to remain relevant and applicable as AI/ML techniques bring about changes in analytical environments.

The work in this field, said Arlyn Burgess, associate director of operations and strategic initiatives at the University of Virginia School of Data Science (UVA SDS), “transcends traditional disciplinary boundaries to discover new insights, often by combining disparate datasets that would not likely be brought together otherwise.” She explained how helping students learn to view their work through an ethical lens is part of the school’s mission.

“Too often the ‘human in the loop’ is unable to overcome the inherent ethical challenges posed by AI algorithmic instruction,” said keynote speaker Dr. Jarrett Zigon, director of the UVA SDS Center for Data Ethics & Justice. He proposed that training data scientists to become ethically attuned individuals—enabling them to think ethically in the various contexts they encounter—would lead to more favorable outcomes than following a prescriptive set of rules alone.

Brant Horio, LMI’s director of data science, echoed the need for an approach that helps data scientists move beyond responding case by case. “Too often, it seems the response to data ethics concerns is that ‘it depends’,” he said. “There should be a conscious and deliberate data ethics perspective and approach that persists throughout the AI/ML lifecycle from ideation and development through monitoring and sustaining deployed solutions.”

Horio moderated a panel with Jessica Young from the National Security Commission on Artificial Intelligence; Davey Gibian, chief business officer at Calypso; and Joe Norton, LMI’s director of data visualization and product development.

During Q&A with the audience—predominantly senior decision-makers from federal agencies—panelists discussed how organizations can leverage culture to promote ethical frameworks and how analysts should annotate algorithm-produced recommendations.

“I think the takeaway here is that there’s not necessarily a reinvention that we have to go through as far as ethics and data,” said one workshop participant. “As the field continues to grow, we’re realizing we need to build these things in. We’re not reinventing the wheel. We’re just pausing to think about the ramifications and all the other pieces that go with the tradecraft of data science.”