In the first of this two-part blog series, we outlined the data engineering challenges impeding greater uptake of Artificial Intelligence (AI) in the public sector. These challenges are not the only roadblock to progress; Governance of the development and use of AI is also a complex and evolving landscape in itself for public sector departments to operate within.
In part two, we explore the important governance considerations associated with AI, including the ethical questions for practitioners, and the need to both protect the public while also increasing citizen confidence in AI-enabled services.
Many of the objections the public has to greater use of AI are based on misconceptions of what the technology is capable of. To dispense these concerns and restore trust in AI, the public sector acknowledges that it must work harder to educate citizens on the positive benefits to society of AI. And it has set out in the National AI Strategy (2022) the need to provide the right governance guardrails, including establishing an AI governance framework, and developing practical governance tools and standards for AI programmes.
So, how straight forward is this to achieve?
Drawing an ethical line in the sand
On the one hand, public scepticism of AI is understandable. Despite legal and regulatory constraints, there are ongoing worries that citizen data will be repurposed, shared and used by public sector bodies without consent, particularly within the context of leveraging large datasets for the purposes of training AI models.
There are also valid concerns associated with bias.
If AI services are only as good as the data they're trained on, what is being done to mitigate the conscious and unconscious bias of the individual(s) who originally logged the data? To what extent will biased decisions be made without any recourse by data subjects to intervene or object? If the public sector isn't marrying up demographic with operational data, assumptions may be erroneously made with a bias towards white males or other groups, for example.
Ethics and trust in AI, as covered in a recent research paper by Kainos and Tortoise, is a broad and complex topic and the public sector and its partners continue to deepen their understanding of how the technology will be used, and are working on relevant best practices, policy and legislation. The UK Government recently undertook public consultation and produced a new action plan in July, which provides update on the National AI Strategy and outlines what will happen over the coming 12 months. The AI Action Plan represents an honest update that there is still much more to be done.
Trust in AI is mission critical to Kainos. We have recently hired our first data ethicist and have, in partnership with public sector departments, developed an AI governance framework. The framework aims to safeguard the delivery of our AI-enabled digital services; in line with the National AI Strategy and with particular focus on explainability and bias considerations.
Dispelling myths
Winning over the public is important to the success and adoption of government AI projects going forward. A great deal of opposition is rooted in outdated or erroneous ideas about what AI is or what it can do. In reality, AI is simply a powerful pattern detection system that can help to make organisations more efficient and to deliver innovative customer-facing services.
This is the message that the public sector needs to promote: positive use cases, like the use of machine learning to risk assess MOT testing sites to identify fraud and inadequate standards; or the use of machine learning to compare precedent transfer deed documents to identify discrepancies as part of the new building development process.
For AI to transform the delivery of public services, it needs to first gain public acceptance. And that will only happen with greater transparency over its use and the development of clear governance rules.
From our experience, the public sector wants to be guided in the right direction on both. By asking the tough questions now, and laying the foundations for trustworthy AI, we will be ready when the time comes to ensure that AI can make a real difference.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.