Artificial Intelligence is Here | Session 1, Part 1: “What is AI?”

This session introduces participants to the basic concepts of artificial intelligence (AI) and an understanding of how AI and machine learning works, as well as the nature of the risks and benefits it generates. We’ll take technical ideas and make them easily accessible for a non-technical audience, with the goal of laying the foundations for confident engagement by regulators and administrators in discussion and decision-making around AI. Key takeaways will include how machine learning differs from conventional computer programming—and why it presents new and different risks; what choices go into building an AI system—from choosing data sets to model architecture to performance metrics—and thus where oversight and regulation can intervene; and, finally, where value judgments are being made, and who is (or should be) making them. This session will be of interest to anyone interested in understanding more about AI and getting basic grounding to understand more applied topics. It is targeted at a non-technical audience.

Speaker:

Gillian K. Hadfield, Schwartz Reisman Chair in Technology and Society; professor of law and marketing, University of Toronto; Canada CIFAR AI Chair, Vector Institute.

Artificial Intelligence is Here | Session 1, Part 2: “What is AI?”

Governments face at least four fundamental challenges in decision making: making a large number of decisions; making policy-consistent decisions; making procedurally fair decisions; and learning from the outcomes of decisions. AI and machine learning have the potential to make major advances on all these challenges. Decisions which can be automated, or which can move to a recommendation more rapidly can allow for faster decision making. Similarly, automating decision processes has the potential to bring to bear data and other information that are chosen because of their consistency with policy goals. At the same time, we can eliminate the influence of irrelevant and potentially biased/biasing sources of information. Because machines do not get tired and are not naturally biased, they can under some circumstances make choices which are procedurally fairer. Automated decision making can incorporate much more information into any retrospection or audit. Learning is easier. There are, however, major risks to be considered, among them principal-agent problems, problems of biased data inputs, problems of explainability and black-boxing, and problems of consent and procedural unfairness, where citizens will not accept a decision made without a human in the loop. This talk presents a short framework for understanding how a policy or administrative might be a good candidate to implement AI or a machine learning process.

Speaker:

Peter Loewen, Director, Munk School of Global Affairs & Public Policy, University of Toronto; Associate Director, Schwartz Reisman Institute.

Artificial Intelligence is Here | Session 2: “Citizen consent and the use of AI in government”

Governments are increasingly exploring the use of the algorithms, machine learning, and artificial intelligence to improve government performance, increase the speed of decision making, and enhance the fairness of government decision making. Do citizens support this? Assuming that citizen support of the process of government decision making is necessary for institutional legitimacy, this talk explores citizens’ support for governments relying on algorithms for decision making. Drawing on data from more than a dozen countries, it identifies four distinct obstacles. First, citizens support various justifications for the use of algorithms, but no set of justifications strongly. Second, citizens evaluate (any) algorithmic innovation negatively versus the status quo. Third, citizens’ trust in algorithms develops independently of algorithmic performance. Fourth, opposition to algorithmic government is higher among those who fear the broader effects of automation and AI, suggesting a potential for a populist backlash against government use of algorithms for decision making.

Speaker:

Peter Loewen, Director, Munk School of Global Affairs & Public Policy, University of Toronto; Associate Director, Schwartz Reisman Institute.

Artificial Intelligence is Here | Session 3: “Deciding when and how to use AI in government”

This session dives into some case studies of how governments around the world have started to use AI—from facial recognition technology at the border to processing of tax or benefit claims to autonomous public transit systems. We draw out some lessons learned and opportunities that AI presents, if done well. Then we turn to some of the tools and regulatory frameworks that are being developed—and could be developed—to help guide the responsible and safe procurement and implementation of AI systems by governments. Key takeaways include the need to understand the challenges of risk assessment in this rapidly and constantly evolving complex technology, the importance of pilots and designing an agile regulatory framework, where general procurement and implementation policies are valuable and where sector- or application-specific approaches might make more sense, and the importance of outcome-based metrics and auditing for fairness and safety.

Speakers:

Gillian K. Hadfield, Schwartz Reisman Chair in Technology and Society; professor of law and marketing, University of Toronto; Canada CIFAR AI Chair, Vector Institute.

Peter Loewen, Director, Munk School of Global Affairs & Public Policy, University of Toronto; Associate Director, Schwartz Reisman Institute.

Artificial Intelligence is Here | Session 4: “How AI is transforming the economy”

AI is what economists call a “general purpose technology.” These technologies can have dramatic effects on the economy. In this session, we discuss how machine intelligence lowers the cost of prediction, enabling new types of decision-making. This creates opportunities for businesses and other organizations while changing the skills required to succeed. We talk about what kinds of transformations we are seeing, and can expect to see, in the economy. And we discuss the implications for thinking about economic policy—both to mitigate potential harms from AI deployment and to accelerate the economic benefits AI offers.

Speaker:

Avi Goldfarb, Rotman Chair in Artificial Intelligence and Healthcare, University of Toronto; Chief Data Scientist, Creative Destruction Lab; Research Lead, Schwartz Reisman Institute for Technology and Society.

Artificial Intelligence is Here | Session 5: “What's all this talk about bias, fairness, and transparency?”

In the last few years, conversations around the risks from AI have exploded. We’re hearing about racist, sexist, and in-scrutable algorithms taking over decisions in domains that run the gamut from hiring to criminal justice. We’re learning about how AI-driven content on social media platforms is fostering conspiracy theories and spreading hate and distrust. We’re worrying about how much corporations—and governments—are tracking our movements in public and online and whether our faces belong to us anymore. In this session, we give an overview of the sets of issues that are currently capturing attention in public debates and identify the issues we see on the horizon, including how to ensure that entities, including governments, that rely on AI to make decisions can be transparent and accountable and how to foster the technologies we’ll need to help regulate AI.

Speaker:

Gillian K. Hadfield, Schwartz Reisman Chair in Technology and Society; professor of law and marketing, University of Toronto; Canada CIFAR AI Chair, Vector Institute.

Artificial Intelligence is Here | Session 6: “The global effort to regulate AI”

The global focus on regulating AI has sharpened dramatically over the past few years. The EU has led with comprehensive proposed legislation, and global standard-setting bodies have been working hard to develop guidance for businesses and governments aiming to commit to responsible US of AI. Around the globe, governments are introducing bills to address risks ranging from automated decision-making to digital addiction fed by AI-driven social media platforms. This session maps this global landscape and discusses the core challenges for governments seeking to design, and comply with, new AI regulatory regimes.

Speaker:

Phil Dawson, Policy Lead, Schwartz Reisman Institute for Technology and Society.

Artificial Intelligence is Here | Session 7: “AI, machine learning, and foreign policy”

Governments are increasingly relying on the use of AI and machine learning in foreign and military policy. In this session, we explore the security and intelligence implications of AI and machine learning for democratic governments. How should governments use AI and machine learning for intelligence gathering and decision making? How should they think about the risk management of using existing national and international data to advance national security? And how are AI and machine learning changing the balance and exercise of power?

Speaker:

Janice Stein, Professor of Political Science, University of Toronto; Founding Director, Munk School of Global Affairs & Public Policy

Artificial Intelligence is Here | Session 8: “The future of AI in government”

This concluding session provides an overview of the course’s key insights concerning the benefits and challenges associated with the use of artificial intelligence (AI), its regulation, and the broad societal implications of the widespread adoption of AI. It also offers insights regarding the future of use of AI in government, including how AI will affect the work of public servants and the differences in how AI will be used by democratic vs. autocratic states.

Speaker:

Peter Loewen, Director, Munk School of Global Affairs & Public Policy, University of Toronto; Associate Director, Schwartz Reisman Institute.