by Fenwick McKelvey and Rob Hunt
In the coming weeks, we’ll be posting a summary of AI policy in Canada. For now, we’re happy to report on a productive meeting about the Treasury Board’s efforts to address the responsible use of AI in government.
On November 15, 2017, a group of professors, researchers, and practitioners were pleased to host Michael Karlin from the Treasury Board Secretariat and Tara Denham from Global Affairs Canada to discuss details of the Canadian government’s plans for implementing AI and to cultivate a broader conversation about AI policy in Canada.
Karlin’s in-progress white paper “Responsible Artificial Intelligence in the Government of Canada,” part of the Digital Disruption white paper series, served as the prompt for the meeting. Karlin is taking a novel approach to consultation, posting the paper as a Google document for the public to read and edit as well as making several stops across Canada to gather direct feedback about the paper’s recommendations. Only after this consultation will the paper be translated into formal standards for the federal public service. In addition to serving as a guide for implementing AI, Karlin hopes this flexible but productive style of consultation will become more common in Canadian government.
We were encouraged by the ease of participating in the Treasury Board Secretariat’s consultation process and their willingness to meet us at the university. Denham’s participation in the process was also welcome. As director of the Democracy Unit at the Office of Human Rights, Freedoms, and Inclusion at Global Affairs Canada, Denham seeks to ensure that Canada incorporates human rights considerations in its approach to AI in government, in the private sector, and in a global context. We remain hopeful that Canada will continue to foster such openness in policy formulation, allowing researchers as well as those affected by policy decisions to play a part in the early stages.
While appreciative of the process thus far, our group stressed the importance of committed inclusive consultation. Given the wide-ranging consequences of AI, the voices of indigenous people and union leaders must be sought out and their participation facilitated. Canada is currently experiencing what legal scholar Michael Geist has deemed a “consultation crisis” with cash-strapped public interest groups being asked to somehow contribute to a rapidly expanding number of important regulatory changes. Financial support is essential to ensure that people outside of academia or the corporate world can join these conversations. The CRTC, the more familiar regulatory institution for us, allows participants to charge for some of their expenses; this is a small but welcome acknowledgement that engaging in the policy development process can carry real costs for many participants.
Our conversation turned to the paper itself as Karlin walked us through the state of responsible AI at the Treasury Board Secretariat. The report considers a range of specific applications aimed at improving government service delivery, such as chatbots, while also attending to the risks of automation. In that regard, the report should be commended for what it avoids as much as what it includes.
Delving further into the complex ramifications of automation, we discussed ways that an automated decision making system could increase efficiency without undermining human oversight. There is a need to understand what Mary L. Gray and Siddharth Suri call “the paradox of automation’s last mile”—the phenomenon where automating something entails hiring humans in difficult, precarious, low-paying jobs to make those decisions that lie beyond automation’s grasp. As the history of women in computing has shown, what counts as routine is often the product of social division and class. These wider concerns haunt any specific application of AI.
We recognize that this white paper will only form one small piece of the AI puzzle in Canada, but overall the meeting represented a welcome step toward a more robust conversation about AI policy for Canada. We believe the Government of Canada has an opportunity to act as a leader in AI policy by establishing and following some guiding principles:
- Focus on assistance rather than automation—aid rather than replace workers.
- Recognize bias and the sources of training data—address potential harms from the start rather than seek to fix discriminatory outcomes afterward.
- Ensure explainability and accountability—make sure that AI improves public understanding of governmental processes and decision making.
By being a model citizen, the Government of Canada has an opportunity to lead by adopting strong, accountable, and public-minded policy toward AI. Furthermore, we hope this specific conversation about AI policy inspires a much broader discussion of these issues.