Winning the AI Revolution for American Diplomacy

WarontheRocks.png


By: Ryan Dukeman, November 25, 2020
Originally published in War on the Rocks

“Artificial intelligence is the future … for all humankind. … Whoever becomes the leader in this sphere will become the ruler of the world.”– Vladimir Putin

Artificial intelligence is the future … for all humankind. … Whoever becomes the leader in this sphere will become the ruler of the world.”

Vladimir Putin

Precognition algorithms. System-of-systems paralysis. An “all-seeing digital system of social control.” To nonexperts, these technologies may seem incomprehensible — the stuff of science fiction and 1984. But to authoritarian states’ military and political leadership, these are game-changing advances that could be the key to upending the international system. Perhaps most importantly, some of these technologies could be integrated sooner than one might think. According to one prominent journalist, in China and soon elsewhere, “the panopticon is already here.”

Big data and artificial intelligence (AI) are restructuring the geopolitical divide between nations that harness its potential and those that do not, transforming the nature of conflict, empowering asymmetric and nonstate actors, and yielding increasing returns for first movers. As oil factored significantly into the great geopolitical conflicts of the 20th century, so too will the politics of data and artificial intelligence quickly become the politics of international security, for better or for worse.

As the Department of Defense’s 2018 AI strategy notes, “AI refers to the ability of machines to perform tasks that normally require human intelligence — for example, recognizing patterns, learning from experience, drawing conclusions, making predictions, or taking action — whether digitally or as the smart software behind autonomous physical systems.” The United States is the global leader in such technologies for now, but its national security bureaucracy is poorly configured to capitalize on this advantage.

In particular, the government is sidelining the very agency that could both lead on the global policy challenges of AI and benefit from applying AI to its way of working. I’m talking, of course, about the State Department.

Yes, you read that right: For the United States to win the data wars, the State Department should become an interagency leader in AI.

While the AI and national security debate has predominantly focused on its applications in warfare, espionage, and homeland security (including election security), it has so far missed the critical relevance of AI for American diplomacy. This is a mistake, because in the coming decades AI will be essential both as a top-tier agenda item in strategic competition against digital authoritarianism and as a tool for more responsive, informed, and effective foreign policymaking.

AI is the next “gray rhino” security threat: a slow-building event that, when it finally hits, can be even more disruptive than conventionally recognized threats like interstate rivalry or terrorism. As such, it deserves an affirmative, cross-issue digital democracy agenda championed by the State Department. The department is uniquely positioned to lead on this agenda, leveraging the interdependent security, human rights, and economic dimensions of AI to advance governance in line with American values. Relatedly, the State Department can also help right the “institutional imbalance” of a militarized foreign policy and empower civilian diplomats by capitalizing on AI in its ways of working, including areas like strategic planning, crisis response, and other core diplomatic functions.

A Digital Democracy Agenda for the Next “Gray Rhino”

The ongoing pandemic has broadened the national security community’s conception of threats to include so-called gray rhinos. In addition to disease and the accelerating climate crisis, AI presents the third such threat, but is unfortunately under-resourced in current foreign policy budgeting. While the defense budget treats AI as a technological innovation for guarding against other threats, the national security and defense strategies do not treat it as a significant threat in its own right.

This is a mistake. If poorly managed, the economic, security, and human rights impacts of authoritarian-enabling AI pose a serious threat to the American way of life, and should be planned for and prioritized accordingly. First and foremost, frontier technologies constitute a digital battleground for near-peer competition between the United States, China, and Russia. China in particular has focused diplomatic and state-capitalist efforts on shaping a global approach to digital governance that allows for unmitigated domestic surveillance and state infringement on free speech. This effort is not just a mercantilist attempt to supplant American firms’ AI leadership with that of Chinese state-owned enterprises — it is also meant to shape an international order reflective of China’s abusive domestic system. These techno-authoritarian models of digital governance are in direct opposition to American values and pose a threat to U.S. leadership on issues of human rights, economic security, and protection of intellectual property. AI similarly enables disinformation campaigns and election interference (e.g., through generative networks used to create deepfakes), as well as more advanced, detection-evading cyber attacks, and thus should be internationally regulated as an anti-democratic force multiplier.

Yet so far, the United States has reacted defensively and ad hoc to competitors’ affirmative, strategic moves. Because of the Donald Trump administration’s hostility toward multilateralism, the United States has yet to take a leading role in defining and promulgating a Western vision for global AI cooperation. Rather than react piecemeal through issue-specific bilateral agreements, the United States should position its diplomacy to credibly advance an affirmative digital democracy agenda to shape AI in line with its interests and values.

What would such an agenda look like? First, it would prioritize emerging digital dimensions of human rights and personal freedoms — the right to be forgotten, personal ownership of data, and transparent algorithmic justice in government and law enforcement, to name a few. Second, it would be multilateral to the core, as no one country can protect its citizens from malign applications of AI on its own. Recent calls for a “T-12” alliance of democratic and tech-leading states recognize the fundamentally transnational nature of AI, and the establishment of the Organization for Economic Cooperation and Development’s AI Policy Observatory is a promising start. Beyond the creation of common standards across countries representing at least a large plurality of global GDP, multilateralizing the AI agenda allows for easier pooling of research and regulatory innovations, so that free societies, rather than techno-authoritarian ones, remain leaders in the technologies themselves.

Some specific ideas worth considering include the creation of international standards for a “Personal Data Record” and interoperability between platforms, so that a consumer’s existing data does not anti-competitively tie them to their current social network or healthcare provider; a digital concert of liberal democracies with data rights standards extending the reach of currently regional approaches like the European Union’s General Data Protection Regulation; updating the laws of war to include transparent standards for conflict in cyberspace (a “Cyber Geneva Convention”); or a broader framework for global data and privacy governance to protect civil society, democracy, and human rights.

The State Department Should Lead on AI

Why is diplomacy — rather than defense or industrial policy — the right venue for advancing a digital democracy agenda, or something like it? Simply put, the multidimensional and transnational nature of AI requires an integrated, strategic approach, and diplomats are uniquely skilled in making policy across these issues more than the sum of their piecemeal parts. To channel former French Prime Minister Georges Clemenceau, AI is too important to be left to the generals — or the technologists — alone.

So far, most of the government’s focus on AI has been through research and development investments at the Defense Department, or through White House Office of Science and Technology Policy efforts to capitalize on it as an innovative economic windfall. Unfortunately, this approach — including leaving much of the intergovernmental norm-setting on AI to peer-to-peer military discussions — misses the strategic forest for the tactical trees. Issue-specific approaches that target global AI only as a military tool or an economic boon have two dangerous flaws: They largely ignore AI as a human rights issue, and forsake the potential for alliances and issue-linkage across policy areas. For example, existing efforts have failed to leverage U.S. technological leadership to advance global norms aligned with Western values, or to use foreign assistance to promote inclusive digital development at scale.

The State Department’s diplomats maintain a comparative advantage over interagency peers in remedying exactly these flaws. Diplomats are skilled in dealing with crosscutting global problems, and well suited to play a leading, orchestrating role in advancing a concerted American vision for AI across its economic, security, and human rights aspects. In Washington, at their posts, and in multilateral organizations, diplomats spend much of their days integrating peer agencies’ competing perspectives to advance a concerted U.S. agenda; leveraging instruments of national power such as foreign aid, military assistance, or visa regulations to advance disparate goals across different issues; and negotiating in bilateral and multilateral fora on economic, political, human rights, and military matters. The State Department’s sustained leadership in international organizations, which will play a major role in digital technology governance, is an additional underutilized advantage. The White House is wrong not to include the State Department as a “key agency” in its strategy on AI. In fact, the department may be the keystone agency — linking disparate, issue-specific efforts together so the result is more than the sum of its parts.

From Data Diplomacy to Data-Driven Diplomacy

Beyond its leadership potential on AI as an international policy issue, the State Department has three key advantages in applying AI technologies to its way of working, increasing its value in the interagency foreign policymaking process and reclaiming some of the influence it has ceded to the Defense Department and National Security Council in recent decades. First, the State Department has an exclusive, novel dataset of millions of diplomatic cables representing a century of collective knowledge of foreign governments and societies. Second, it has a unique overseas infrastructure for overt information collection in the form of embassies and diplomats. Third, it has one of the most highly skilled workforces in the federal government, with incoming Foreign Service officers having, on average, a master’s degree and a decade of work experience. By activating these assets and using AI as a force multiplier, the department can contribute more to the interagency policymaking process by bringing original and rigorous insights to national security debates. America’s oldest cabinet agency — often maligned as stuck in the past, pining for an imagined diplomatic golden age and averse to technological reforms — can reclaim leadership by investing in AI as a tool for scenario planning, crisis response, and innovative policy analysis. While frontier technologies will ultimately find their way into most parts of how the department operates, there are a few areas that are most important to prioritize in the near term to deliver immediate impact.

Predictive platforms, similar to those commercially available in the political risk industry today — and some being prototyped by peer agencies — are remarkable in their ability to use micro-level data (e.g., credit card purchases, social media posts, and search engine trends) to predict macro-level geopolitical events in ways no humans on their own could detect. Unfortunately, however, the State Department today is unique among foreign affairs agencies in that it uses almost no quantitative scenario modeling in its strategic planning, limiting the utility of its crisis prediction or strategic planning based too often on anecdote and experience without sufficiently testing hypotheses or assumptions.

Peer agencies can point the way to a better process. For crisis prediction, for example, the U.K. Foreign Office has built a platform to automatically translate and analyze hundreds of thousands of hyperlocal news sources to spot early warning signs of humanitarian crises and violent extremism. For longer-term forecasting, the Intelligence Advanced Research Projects Activity’s Aggregative Contingent Estimation program uses advanced analytics to aggregate not just data but the wisdom of crowds: The program elicits judgments of event probabilities from many intelligence analysts, and then uses algorithmic techniques to weight them on factors like past performance and “cognitive style,” providing more systematic estimates than any analyst could provide individually. Even in the policymaking process, AI can greatly empower human analysts by ingesting orders of magnitude more information and applying human-defined rules and principles more consistently. Doing so tests long-held assumptions, questions well-documented cognitive biases, investigates patterns of history rather than relying on simple analogies, and casts a wider aperture on sources of analysis like social media, satellite imagery, and user-generated video content.

Where would the data feeding these prediction engines come from? Fortunately, in the parlance of data science, the State Department sits on top of a “novel data asset” — a unique trove of data only it owns, but which it has yet to fully leverage. For decades, the collective knowledge of diplomats across the globe has been transcribed, shared, and stored in cables. U.S. diplomats have documented everything from the department’s grand strategy to individual meetings with foreign officials. This corpus of cables constitutes America’s diplomatic equivalent of a “knowledge graph,” but unlike those that drive the functionality behind Facebook and Google, it has never been fully utilized to help diplomats do their jobs. Much foreign policy analysis today involves manually traversing this knowledge graph by consuming hundreds of cables a week, yielding a policy process overly reliant on anecdote and predictably subject to human cognitive biases toward action, “credibility” preservation, and parochial conceptions of national interests. AI-powered capabilities such as enhanced search, natural language processing, and natural language generation can augment this type of work to help diplomats quickly make connections they would otherwise not have discovered, such as by detecting patterns of state behavior, analyzing past interactions with foreign leaders, and evaluating alternative courses of action.

This is not to say that AI is a magic bullet in the craft of foreign policymaking. Pattern recognition algorithms, for instance, can break down in the face of truly unprecedented events. States’ behavior often results from the aggregation of thousands of individual decisions made by idiosyncratic individuals in diverse settings, yielding wide error bars in human- and machine-generated forecasts alike. Last but not least, effective predictive engines depend on vast troves of high-quality training data that require upfront investment to parse, clean, and validate, challenges even the Defense Department — with its vast resources — has handled far from perfectly.

Yet on the whole, advanced analytics can significantly augment, not automate, traditional diplomatic tradecraft, putting the State Department in a position to make first-in-class contributions to the interagency policy process and allow its workforce to focus on higher-value tasks. To consider a relatively low-tech example, if annual publications like the Trafficking in Persons Report looked less like 300-page narrative PDFs and more like the Johns Hopkins University COVID-19 Dashboard, State Department analysts could spend more time explaining why trends and events happened rather than simply describing what took place. Beyond tech upgrades, though, retooling the State Department’s hiring, training, and education practices to augment diplomatic expertise with the insights of AI can multiply the value the department brings to foreign policymaking.

Pathways to Change

Unfortunately, the State Department’s current structure is not optimized for its potential to lead on AI as a top-tier policy priority and transformative way of working. As Jared Cohen and Richard Fontaine recently noted, advanced technology “is not simply a niche functional issue buried in a crowded foreign policy agenda; it is a central element of modern geopolitical competition.” Indeed, reorienting the State Department and its people for great-power competition in the realm of frontier technologies is a central thrust of the second quarter report from the National Security Commission on Artificial Intelligence, yet has received relatively little attention. Across the Foreign Service — traditionally generalist diplomats who rotate between overseas and Washington assignments — and Civil Service — typically Washington-based officials with relatively specialized training and longer assignments — reforms to mainstream AI, advanced analytics, and frontier-technology fluency are ambitious yet achievable.

First, the State Department should accelerate structural changes that elevate the importance of AI and related frontier-technology issues across security, economic, and human rights siloes. Currently, cyber policy (broadly defined) is handled by a special coordinator for cyber issues in the Office of the Secretary of State. Last June, the department proposed a new Bureau for Cybersecurity and Emerging Technologies. However, progress in creating this new bureau has stalled, with House Democrats placing a hold on its approval and the Government Accountability Office noting faulting State for not including key interagency stakeholders in its design. Getting a Bureau for Cybersecurity and Emerging Technologies properly established as a permanent home for diplomacy on advanced technological issues is a start, but the new bureau will be housed under the undersecretary for arms control and international security, potentially sidelining the economic and human rights aspects of AI. Today, for example, cybersecurity issues are handled by the coordinator for cyber issues, while “critical communications infrastructure” issues are overseen by the Bureau of Economic and Business Affairs, bifurcating interlocking topics along bureaucratic lines.

No bureaucratic structure is perfect, but to remedy this, the State Department should create a permanent office for digital human rights in the Bureau of Democracy, Human Rights, and Labor, and promote frequent staff-level coordination across units like the Bureau for Cybersecurity and Emerging Technologies and its economic and human rights peers. Beyond a single focal point, though, the State Department should integrate or mainstream AI and frontier-technology diplomacy throughout its work. Recent proposals to do so include creating a “science cone” in the Foreign Service, or having dedicated technology policy officers in each major regional and functional bureau.

Beyond structuring the State Department to lead on AI as a policy area, training and professional education reforms can better prepare the diplomatic workforce to reap the benefit of AI as an analytic skill. Ultimately, the State Department’s most important asset is its people, and this workforce should see the use of advanced analytics as a benefit and not a hindrance. When incorporating new technologies into the work of the State Department, senior officials should be hyperfocused on empowering America’s world-class professional diplomats, thereby saving time, delivering novel insights, and serving as a foreign policy force multiplier. Simultaneously, foreign service officers need to be afforded training opportunities to understand the value of such tools to their day-to-day activities and be incentivized professionally to gain new skills that allow them to leverage technologies to make better decisions. Much as nuclear physicists were deeply embedded in America’s Iran deal negotiating team to integrate the finer points of scientific specialty with diplomatic expertise, 21st-century U.S. diplomats will need at least “conversational analytics” to integrate the data-scientific and geopolitical issues at the core of strategic competition between liberal democracy and surveillance authoritarianism.

Historically, this type of specialization cuts against the cultural grain at the State Department, which prides itself on adaptable generalism and learning on the job, particularly in the Foreign Service. Yet as several recent reports on the future of diplomacy have noted, “breaking away from ‘born, not made’” will be critical to equipping diplomats for effective interagency leadership on transnational issues of growing importance, and, as I’ve argued in the Foreign Service Journal, is more in keeping with a history of diplomatic professionalization than is typically assumed. AI and other growing specialist demands can help push the department to better incorporate professional education into diplomats’ career trajectory, meeting the moment by investing in its core asset: its people.

No such transformation is possible without congressional resourcing and executive sponsorship in the State Department. With some changes and investments, the State Department is well positioned to promulgate a democratic digital agenda that is aligned with American values, and to use investments in AI to transform its diplomatic and foreign policymaking functions. When discussing AI and innovation in government, it is common to fetishize technology and insist on its centrality to transformation. However, technological innovations should come with accompanying culture change, and leaders across the department should define and communicate the value of these innovations to the mission of diplomacy.

Skepticism towards technology is as old as the diplomatic profession itself: Upon receiving his first telegraph message in the 1860s, British Prime Minister Lord Palmerston, for example, exclaimed, “My God, this is the end of diplomacy!” Yet when used properly, AI can empower (rather than replace) career diplomacy like never before, putting civilian foreign policymakers back into the driver’s seat at home and overseas through leadership on a major “gray rhino” security threat and improved decision-making.

Ryan Dukeman (@RyanDukeman) is a senior fellow at FP21 and a Ph.D. student at Princeton University, where he researches institutional reform in U.S. foreign policy agencies. He previously helped found the U.S. State Department’s Center for Analytics. 

He is grateful to Garrett Berntsen for extensive feedback and collaboration on earlier drafts of this article. The views presented here are the author’s alone.

Previous
Previous

Why We Should Care About Politically Appointed Diplomats

Next
Next

fp21’s Inaugural Report: Less Art, More Science