This article is written by Dr Shainaz Firfiray, Senior Lecturer, University of Aberdeen.

  • The problem: Issues about data privacy and algorithmic bias in AI are occupying centre stage along with a focus on how to ensure accountability as work is becoming increasingly automated.
  • Why it matters: It is critical that we find a set of strategies that would establish trust among public servants and citizens that the AI used in public services is aligned with ethical principles. Additionally, it is important that we acknowledge the risks involved in AI use if we want to reap its potential benefits.
  • The solution: A set of strategies that will offer guidance on how to build public trust, respect human rights and the rule of law and ensure transparency and responsible disclosure around AI systems.

We are currently witnessing a surge in the adoption of AI technologies within public services. Despite their popularity, there is relatively little guidance on issues of privacy, safety, data acquisition and ethical implications of AI.

Recent breaches of public and private IT systems have received widespread coverage. Issues related to data security can get exacerbated through the use of AI applications, which create highly connected and interdependent systems.

There are several trade-off judgements that need to be made when using AI in public services.

The services offered by AI are also frequently in conflict with the societal ideals of fairness as algorithms that excel in making the most accurate decisions may systematically discriminate against certain minorities. There are several trade-off judgements that need to be made when using AI in public services. For instance, judging when it is acceptable to use an algorithm that performs badly for a specific minority if it is on average accurate for the population. Or determining what types of decisions and skills should always be left to humans, and where to reject AI applications. Or deciding how data might be used to improve the quality and efficiency of public services while respecting the privacy and dignity of citizens. However, the main dilemma before public servants is how these trade-off judgements should be made.

Utilitarian lens and a cost-benefit analysis

Traditionally, a cost-benefit analysis drawing on a utilitarian lens has been applied to make trade-off judgements as it allows people to reach decisions on the basis of whether the benefits of adopting a specific technology outweigh the costs. These procedures are widely used in industry and governance because they are seen as objective and provide clear guidelines for reaching a decision. While a cost-benefit analysis may appeal to some public servants, by itself it is not ideal for exploring such trade-offs. It can yield some helpful data which may feed into the decision-making process, but it does not recognise that values are difficult to quantify and that numbers on their own may mask some controversial value judgements.

When different groups deliberate and express what matters to them, it is possible to reconcile conflicting perspectives and reach a compromise.

Resolution of such dilemmas can take various forms. Below, I propose strategies that will give voice to relevant stakeholders and help us reach decisions that are aligned with the rights and privileges of those affected by AI technologies.

Public engagement and dialogue

Public engagement is critical for addressing trade-off questions in a way that can be justified to all members of society, especially when trade-offs arise because of conflicting values and perspectives of different groups. However, when different groups deliberate and express what matters to them, it is possible to reconcile conflicting perspectives and reach a compromise. There are a range of methods that can be deployed to foster public engagement including large-scale public surveys, focus groups with an educational element to gather informed views of citizens and iterative public dialogue where technology experts and citizens collaborate to propose recommendations for policymakers.

Reinforcing cybersecurity

Cybersecurity issues may play out in complex ways as AI systems might get compromised in ways that are hard to understand and could result in major security lapses. Additionally, with an increase in the use of chatbots in public services, citizens could become potential targets for bots designed by hackers and scammers who could gain access to sensitive information. Hence, security and privacy concerns of citizens should be at the forefront of any AI initiative.

Strengthening the AI ecosystem

There are significant barriers to adopting AI technologies in public services as they are risk-averse and often sceptical of experimenting with emerging innovations. Public services are more comfortable following rather than leading the way in adopting AI technologies. However, this wait-and-watch approach is unhelpful, given the dynamic pace of change within the field of AI. Effective use of AI will require public services to reframe their attitudes towards risk. They need to recognise that AI initiatives come with a certain set of risks, but these risks can be managed effectively over time. However, the reluctance to engage with new AI technologies may severely limit future opportunities that enable us to tackle major societal problems.

Gaining public trust

AI systems used in public services should be monitored on a regular basis. There are frequent reports in the media on the misuse of AI or the accidental negative outcomes attributed to AI. However, developing rules and processes for monitoring algorithms is a challenging task. Companies that develop AI applications are often unwilling to share details on how their systems work, and in the absence of this information, it is challenging to control these systems or ensure they function ethically. Many researchers have argued that existing laws are limited in their ability to protect citizens against biased algorithms. With an increase in the use of AI systems, it is crucial that public services show a willingness to open their systems for inspection to ensure citizens understand how these systems function and are able to trust their decisions and outcomes.

To sum up, AI has ethical considerations that are dissimilar to those of other technologies. Concerns over privacy and disclosure of sensitive information to the public, the irresponsible use of shared data, the incompatibility of algorithmic versus human value-based judgements and the biases in decision-making are well-documented. In order to promote better lives through AI, we need to ensure that all AI initiatives are able to foster a society that treats people with dignity and compassion, respects the rule of law and acts transparently.

👋 You can write an article like this one! Share your thoughts with a community of public servants. Learn more

(Image Credit: Pexels)