How to Ethically Use AI in the Public Sector

You’ve probably heard how artificial intelligence (AI) is redefining the way we work. New technology is released daily, and it can be easy to become overwhelmed. 

At Code for Canada, we’re collaborating with partners and public servants to navigate this emerging landscape. This guide provides a primer on how we approach AI in our work.

What do we mean when we say “AI”?

These days, the term “Artificial Intelligence” is being used as a catch-all term to describe a wide range of technologies and tools. Here, we define different forms of AI and how they relate to one another. 

Artificial intelligence 

Artificial intelligence is the science of making software act and think like humans by training machines to go beyond pre-programmed rules. We interact with AI-powered tools daily — from flight price predictions to asking questions to models like ChatGPT or CoPilot. 

While recent advances in AI have made the field more familiar (and perhaps a bit daunting), Artificial Intelligence remains a tool that excels at specific tasks. The idea of general AI — a system that matches human intelligence across multiple fields — has yet to be realized.

Today's AI advances build upon past milestones. Let’s break them down so we can understand them better.

Machine learning 

One of the first major milestones in AI research was machine learning, where software is trained on large datasets to enable prediction and automation. 

Computers excel at quickly processing vast amounts of data and finding patterns in it. Machine learning has enabled advanced data processing that would have taken humans a long time to accomplish without these tools. 

Deep Learning is a subset of machine learning, which builds AI systems using neural networks, the development of artificial neurons that mimic how humans learn.

Large-Language Models (LLMs) 

The next breakthrough came with large-language models (LLMs), AI systems trained on text using deep learning. These systems are designed for Natural Language Processing (NLP), or the generation and processing of language. LLMs analyze text patterns and predict the next step. 

Generative AI

These GPT models are often referred to as Generative AI or an AI tool that can generate something new based on its training data. These are really exciting tools that bring AI capabilities to the consumer market, allowing us to summarize data, brainstorm ideas, or translate languages. While these tools are helpful and improving, they are still biased and can provide misinformation. 

How can AI be used ethically? 

We believe it is crucial to approach the use of AI using the same principles that govern the ethical use of other technologies. Building on the work of other dedicated civic technologists, Code for Canada has created our own set of principles for ethically working with AI. We encourage you to explore them and consider how they may apply to your work in the public sector. 

How can AI benefit the public sector?

Generative AI tools can make many office tasks easier, but the real value of artificial intelligence lies in using big data to solve problems. 

At its core, artificial intelligence involves building machines to mimic human abilities. Machine learning can automate repetitive tasks, allowing teams to focus on improving service delivery. 

With LLMs and GPT models, we can create reference tools to understand complicated material or reference documents. With NLP, we can decode handwriting and make reading text documents a lot easier. These are tools and not replacements for human ingenuity. 

At Code for Canada, we see big benefits to working with artificial intelligence within the public service. Let’s explore a few.

Enhancing decision-making

Artificial intelligence systems excel at working with data and can improve our decision-making process. While AI models won't eliminate the need for human judgment, they can help us to identify patterns, categorize data, and prioritize information more efficiently than manual methods. It is essential to always make sure there is a human review in the process.

Improving service delivery

We see potential in using artificial intelligence to help personalize public services while eliminating burdensome steps in the service delivery process. Any repetitive step for public servants or the public could be an opportunity to integrate an AI model. Processes that can be clearly defined and taught have the potential for AI integration.

Working with big data

If you’re working with large data sets, AI models can help operationalize and manage your data. Models can be designed to spot unusual patterns in big datasets that might need a closer look from humans. We can adjust these tools to look for different things, as long as we have enough data to work with.

Tackling challenges facing AI adoption in the public sector

We see huge potential for the public service to leverage AI systems to improve service delivery and tap into the wealth of available data. There is significant potential for AI services to provide benefits both to end users and internal teams. 

However, developing AI systems within the public service still presents challenges and barriers. With that in mind, Code for Canada is focused on two key areas.

Bridging knowledge gaps 

Developing AI systems and managing data is a difficult skill set to acquire and retain. Building and de-risking these systems without this talent can seem impossible. Even understanding the potential for integrating AI systems can seem out of reach. 

We see the knowledge and talent gaps as a major challenge in the adoption of AI systems within government. However, there are lots of resources online and through the Canadian School of Public Service that can help. Upskilling team members in AI development and data science will help bridge these gaps while exploring team-wide training on the use of AI can reveal its potential.

De-risking AI adoption 

Adopting new technology can feel like a daunting task, especially when working in government. Even though the private industry is moving quicker than ever to incorporate AI into all facets of our lives, we recommend stepping back and building an adoption plan. 

This starts with ensuring you have the skills available, the proper data, and a plan for developing AI systems with a human-centred approach, that puts the needs of your users at the forefront. It isn’t a race to adopt AI, it’s a journey of iterative development and team growth.

How Code for Canada can help

At Code for Canada, we believe in AI's power to transform how we work and interact with the public. We are committed to bringing an ethical human-centred approach that de-risks the development of AI systems. 

Collaborate with us today to explore the opportunities for AI systems within your organization. We provide team training on AI services and development to help bridge knowledge gaps and use our options analysis framework to de-risk the adoption of AI services.

Want to Learn More?

Interested in exploring how your team could leverage AI? Get in touch.