The online interactive magazine of the Association for the Advancement of Artificial Intelligence

By Frank Stein

AI adoption in Government/Public Sector faces unique challenges and opportunities, including a higher standard for transparency, fairness, explainability, and operation without unintended consequences. Keynotes and panels are summarized below, and innovative contributions to addressing these needs are published in the conference proceedings.

This is the fifth year that we have run the symposium, and this year’s symposium was the largest in terms of paper submissions and attendance. The symposium benefited from the major government actions on AI including the Presidential Executive Order, resulting actions, and the NSF AI Research Institutes solicitation. AAAI and CCC also released their 20 year Community Roadmap for AI Research. This year there was a particularly strong emphasis on building AI that could be trusted and accepted by the users.

We started with a Keynote by Maj. Adam Taliaferro, US Army Futures Command: Findings on Operationalizing AI for Multi-Domain Operations. This talk provided an overview of MDO and ongoing learning and experimentation from across Army Futures and Concepts Center to operationalize AI and address MDO problems. He also highlighted the Trust issue – – the Commander owns the Risk — he must know how much to trust the AI and understand the confidence around the AI recommendation.

Aleksandra Mojsilovic, IBM Fellow, IBM Research, provided a keynote entitled: Towards AI Systems We Can Trust. While AI holds the promise of delivering valuable insights and knowledge, broad adoption of AI systems will rely heavily on the ability to trust their outputs. Fairness, robustness, explainability, and accountability will be the underpinnings of trusted AI, and we must learn how to build, evaluate and monitor for trust. Dr. Mojsilovic presented her team’s work on Fact Sheets that would provide transparency on how the models were developed and limitations of the models.

On Friday, Kathleen Carley, CMU/CASOS, provided a keynote on AI and Social Cybersecurity. Social cybersecurity is an emerging computational social science area concerned with keeping the internet open for public discourse without fear of undue manipulation and with a reduction of disinformation. This talk discussed the strengths and limits of AI as a method of value in this area. Dr. Carley provided illustrative examples from various online influence campaigns, and then discussed her work using network analysis to understand how social media is being manipulated.
The final Keynote was provided by Yolanda Gil, AAAI President on A 20 year Community Roadmap for AI Research in the US. Dr. Gil described the trends towards Integrated Intelligence, Meaningful Interaction with Computer, and Self-Aware Learning as 3 key areas for advancement. She also highlighted current issues that must be overcome including International Competition for talent, and the need for long term investments similar to what was done for the Apollo Mission to the Moon.

We had 2 lively Panel Discussions. The first was on Coordinating Vulnerabilities in AI Systems, moderated by Jonathan Spring, SEI/CMU. This panel discussed how might the public sector’s approach to vulnerability management change with the addition of AI systems.
The second panel Towards Open Source and Open Standards Approaches for Trusted AI Workflows was moderated by Jim Spohrer, IBM. This panel reviewed the open source data and AI landscape. The panel discussed the need for an AI Reference Architecture which could be validated against various government AI use cases.

The participants agreed that they would like to attend future symposia to share experiences and address some of the challenges posed this year.

Frank Stein (IBM) served as Chair of this symposium. The Program Committee included: Mihai Boicu (GMU), Lashon Booker (MITRE), Michael Garris (NIST), Zach Kurtz (SEI/CMU), Tien Pham (US Army), Alun Preece (Cardiff U), and Jim Spohrer (IBM). For more, see session papers in AAAI Press Technical Report FS-19-05 at .

Frank Stein is the Director of IBM’s A3 Center in the US Public Sector and Federal Market. He can be reached at