Opportunities and moral restrictions on the use of AI in the public sector

With Jannic Stolzenbach Jensen, Corporate Entrepreneur and Lead Business Developer, NNIT

Artificial intelligence (AI) has moved into the public sector, and the Danish authorities are exploring how the technology can become a valuable tool. According to NNIT's Jannic Stolzenbach Jensen, transparency and human supervision will be crucial in ensuring that AI can improve welfare.

In recent years, the growth and spread of AI have accelerated significantly, and the technology has long since found its way into our everyday lives, for example in advertisements, image recognition, chat-bots and translation programs.

The public sector has also started to explore how AI can streamline and improve the lives of citizens. And this is the right thing to do. Because if the public sector is to represent Denmark as a front-runner for responsible use of AI, it is time to dust off old-fashioned legacy systems and embrace the massive potentials in AI technology.

But how can the public sector innovate using new and, to a certain extent, still unknown, technology without putting sensitive personal data, due legal process and human life at stake?

In this blog, Jannic Stolzenbach Jensen, Corporate Entrepreneur and Lead Business Developer at NNIT looks more closely at the opportunities and moral dilemmas entailed in the use of AI in the public sector.

Transparent data, good decisions

Allowing AI to make decisions that affect people's lives raises important ethical and legal questions, such as:

How can we ensure just decisions? That a disease is diagnosed correctly? That tax revenues are managed appropriately? And what data are these decisions based on?

According to Jannic Stolzenbach Jensen, transparency will be a key requirement, if AI is to be unleashed within the public sector.

"AI and generative language models are generally trained on a wide range of data from the internet, including books, articles, websites and other public sources, so the technology has a broad and comprehensive understanding of facts, culture, language and opinions. However, the quality of the content generated strongly depends on the quality of the data it has been trained on", says Jannic Stolzenbach Jensen.

"If AI is to make decisions in the public sector, it’s important to demonstrate data transparency and to know the provenance of the data on which decisions are based. Fortunately, Denmark has relatively high data quality, so it isn’t necessary to have someone to approve all the data used. However, the origin of the data, as well as how they are collected, stored and used within the framework of the law, must be transparent", he continues.

One of the latest initiatives within secure and ethically responsible use of AI is the European Commission's proposed AI regulation. This regulation recognizes the enormous potential of the technology and its uses in many different areas, but it also stresses that AI may pose jeopardize the public interest and civil rights if it is not used correctly.

"It’ll never be possible to rely on AI for all decisions. And thank goodness for that! For example, the EU has decided that IT must never be permitted to make decisions on and advise about self-injection. If the decision can be harmful to human life, it is not permitted to use AI. It’s a simple, but important rule.", stresses Jannic Stolzenbach Jensen:

A helping hand in public administration

With ethical and security precautions in place, there are numerous examples of how AI can make a positive difference for citizens using the public sector in Denmark. This includes the healthcare sector.

"AI can improve diagnosis, patient treatment and health-data management. For example in scanned images, where, on the basis of a quality score, AI can suggest an informed diagnosis,” says Jannic Stolzenbach Jensen.

He also highlights that public administration is one of the areas in which AI can automate and take over the many administrative procedures and processes performed by the public sector in Denmark every day.

"AI can be used to write and send standard replies in a given process, for example in public benefits administration, in which generative large language models can optimize and streamline case processing", he continues.

AI can make case processing cheaper and easier for administrative employees in the public administration, but there are also benefits for citizens.

"Citizens have many channels of contact with the public sector, and we can see huge potential in using AI to work across the many public silos to create more cohesive user pathways because the systems communicate with each other. And there’s all the letters from the public sector, where AI can adapt the content so that the young and the elderly in particular can understand the content more easily, ultimately improving life as a citizen in Denmark.” explains Jannic Stolzenbach Jensen.

Common for all the examples above is that AI is rarely alone on the task, Jannic explains "In by far most cases, AI will be used as a tool to assist and support employees in the public sector. It’s important not to underestimate the value of human intelligence, which is much more able than artificial intelligence to sense, analyze, assess and validate. I think a long time will pass before decision-making authority is delegated to AI and no longer lies with people".

3 tips: How to choose the right AI partner

To sum up, Jannic Stolzenbach Jensen has three tips for what public authorities should focus on when selecting their AI suppliers and partners:

1. Transparency: Choose a partner who understands the value of data transparency and who won’t compromise on security. A partner who can build and train AI models securely and appropriately, and who understands the value of innovation within the framework of the law.
   
2. Ethics: When entering into a collaboration on AI, it is important to choose a partner with well-defined ethical principles and strong opinions on the use of AI. Such a partner should not only understand the technical side of AI, but also social, ethical and human aspects. This entails a deep understanding of how AI technologies affect real people and society.
   
3. Professional consultancy: Choose a partner who won’t only deliver solutions, but who will also act as a critical and constructive sparring partner. Such a partner should challenge your assumptions, encourage innovation and help you to identify both opportunities and risks in your AI initiatives. By offering professional consultancy and sparring, a partner can help you to develop more reliable, innovative and effective AI strategies. This requires a partner with not only technical expertise, but also with the ability to understand and get involved in your specific context, needs and objectives. Such a partner will help ensure that the AI solutions you develop are not only technologically advanced, but are also adapted and relevant to your unique situation and goals.