This article is written by Wesley Hau (Chair, Ontario’s Agency Open Data Community of Practice), Amy Bihari (Open data and emerging technology lead with Ontario's Digital Service) and Dalia Hashim (Policy advisor to the Chief Data and Digital Officer of Ontario)

Over the last year or so, many of us in the public service have begun to encounter new algorithmic technologies and have started to discuss their place in our programs and services.

This new suite of tools and tech bring new promises to our collective work, but at what cost? In Wisconsin state courts, they implemented a black box algorithm to facilitate sentencing, but one of the unintended consequences was unchecked racial inequality. The reach and impact of these little calculations can be huge, amplifying everything by moving faster and more efficiently — but toward better, faster services or mass inequality engendered in our technology and decision making?

We tend to trust technology because we think it's neutral, unbiased unlike us, but the truth is, more often than not we simply embed our biases into the technology while maintaining its illusion of neutrality. Often misleading us into thinking we’ve created more positive outcomes when we’ve simply systematised the same ones.

That’s where we find ourselves, on the edge of a brave new world. How do we wield these new tools responsibly? How can we define and guide use that incorporates ideals of fairness, trust, equity, justice, social good, economic good, environmental good?

Embracing the power of Open

We are a group of Canadian Public Servants (mainly in Ontario, from various agencies and ministries) working to lay a foundation for responsible and ethical algorithmic and artificial Intelligence (AI) use by government.

Our goal is to embed Open principles in the way the public sector approaches algorithms and data-driven technology like AI. To do this, we are working together as often and as collaboratively as we can to navigate this complex space and serve all our communities well.

In Ontario we have a strong Open Data program requiring all ministries and agencies to make their data assets as open as possible. This foundation of Open by Default has allowed Ontario to share data and data-driven tools (including code) quickly and easily during the current Covid-19 pandemic enabling unprecedented collaboration and community-led innovation during these difficult times.

This may not sound like a big thing, but in government this is not normal — you are expected to work within your team, unit or division and in some cases are discouraged from reaching out to a colleague outside of your group without express permission

Open Data programs are based on principles like transparency, accessibility and collaboration, so it was natural for those of us closest to open data in our organisations to reach out to each other, to figure out how we could work better together in our uses of data-driven technology. Something as simple as creating a community of practice where we can talk with each other about the various ways we could be using these new technologies, provide examples of when we’ve shared data in the open and how it’s led to positive impacts, can go a long way in changing organisational culture within government.

This may not sound like a big thing, but in government this is not normal — you are expected to work within your team, unit or division and in some cases are discouraged from reaching out to a colleague outside of your group without express permission. These boundaries between government teams and organisations reflect the past: past needs, past limitations, past goals. Chances are, the same things are happening in your organisation, and we hope our reflections can help you break down the silos between the teams that we urgently need to work together to create a more sustainable approach to AI in government.

In the course of a few weeks, it went from a community of 5 to a community more than 50

However, changing the way we work is not as simple as buying or building a new collaborative space or tool. People need to be empowered to come together whether in virtual or physical spaces and be free to speak candidly about challenges and safely share new ideas (and hopes and fears). This is the type of approach we are taking, creating safe, open spaces to share thoughts, challenges, questions and resources as we start to navigate new approaches, tools and technologies like Machine Learning (ML), Artificial Intelligence (AI), and other algorithmic uses.

Through our community of practice we’ve been able to learn about new pilots that might be underway using AI, reasons why teams might have considered AI, but then steered away from it, or instances where a project/service benefited from the addition of data driven technologies. All of these stories, case studies and learnings have also helped inform and shape our approach to what ethical use of AI looks like in Ontario that meets the needs of Ontarians and public servants alike.

Creating space for collaboration

Our journey started with a cup of tea.

One afternoon a few of us met at a cafe in Toronto after exchanging emails and discovering that we were all wading into the murky waters of government ML/AI use. As we chatted about our respective projects, concerns and communities we quickly ran out of time, but wanted to keep the conversation going. We wanted to loop in more allies, share files, links, ideas — we wanted more time together at the cafe.

That same week we created a shared Slack workspace, set up some channels (chat rooms) for our favourite topics and started inviting our friends. As people joined, we encouraged them to add their agency or organisation to their profile to help highlight the cross team communication that we were building. In the course of a few weeks, it went from a community of 5 to a community more than 50. This helped build confidence that more than one person or team or organisation thought this was a good idea. It also provided signage for who might be a good person to reach out to for insight or advice — I see Amy is a data lead at the Ontario digital service she might be able to tell me about the Provincial perspective on adopting data standards (or connect me with someone does!).

Now we had our virtual cafe.

Our collaboration efforts do not end within government, that’s just the beginning

Over the next few months we shared draft documents, presentation slides and links to learning events. We visited each other’s committees and working groups via MS teams or Zoom to share our journeys and learnings on topics like digital literacy, algorithmic transparency and data governance. No formal collaborations were struck, so no approvals were navigated, but collaboration was happening. More and more people were visiting the shared Slack space they had heard about, and more alignment began to appear in our presentations and reports concerning responsible AI systems.

Going boldly... together

Currently, we are using our growing network to explore and develop work concerning AI literacy and AI guidance for Ontario. We are hoping by establishing common understandings of uses, issues and goals we can create the conditions for wider collaboration and perspectives in our future standards, practices and policies concerning AI.

As AI begins to be considered or adopted by all facets of government, across all sectors and continents, there is an opportunity to develop stronger protections against inequality, bias, and discrimination in all our programs and services. By approaching ML/AI with lens of transparency, ethics and justice from the start we can work differently and produce different outcomes, outcomes that are better for more communities.

We strongly believe transparency and collaborative participation is setting us as public servants on an inclusive path to a data-enhanced future. We plan to continue to make room for more allies from different spaces with different experiences to connect and shape our collective understanding of responsible, safe and beneficial use of data-driven technology like algorithms and ML/AI.

Ethical, innovative and transparent AI needs to be a collaborative effort across many stakeholders, including yourself. We want to hear from AI practitioners, data users, businesses that rely on AI and other data driven technologies. Our collaboration efforts do not end within government, that’s just the beginning. You can reach out to us through our GitHub projects or email to share your thoughts, advice, join our Slack environment or meet for a (virtual) cup of tea.

Provide us with your feedback on our current work, point us to questions that you may be grappling with as it relates to AI, or simply tell us what your experience with AI technology has been so far. We’d love to hear from you. — Wesley Hau, Amy Bihari and Dalia Hashim

(Picture credit: Unsplash)