Self assessment guide for city procurement and development of Urban AI systems
The Global Observatory of Urban AI has created a self assessment guide for city procurement and development of Urban AI and Automatic decision-making systems. This guide includes a set of questions linked to the Observatory’s 6 minimal ethical standards that city officials can go through to assess the ethical nature of AI systems being deployed in their cities.
Why do cities need ethical urban AI guidelines?
Cities are expanding their use of artificial intelligence, across the digital spectrum, about everything from transport to movement of people, to energy usage, security, infrastructure and weather. The potential of AI and the added value of its applications requires aggregation across sources to solve local challenges, and many cities are pursuing partnerships with public and private providers to develop initiatives that integrate these systems from multiple sources. Despite the opportunities AI brings, it potentially opens the door to negative effects on citizens’ rights and freedoms.
The Global Observatory of Urban AI works to produce policy standards to develop urban AI systems that put people at the core and develop a human-centric approach focused on improving citizens' lives. These self-assessment questions aim at offering a practical guide for local authorities to ensure that the algorithms used in the city remain aligned with the needs and interests of citizens in order to move forward digitally.
The principles in depth
The principle most raised in the documents analysed is that of fairness and non-discrimination. Due to the current debate the term ‘fairness’ is undertaking, non-discrimination appears as a more comprehensive concept to grasp algorithmic justice. In the case of the latest, providing solutions for indirect discrimination, and more concretely “disparate impact”, has been the focus on its literature, try to avoid any discrimination taking place when “neutral-sounding rules disproportionally affect a legally protected group”.
The principles of transparency and openness also figure, although they do so in the most elusive way due to the multiple understandings of the former. Explainability and interpretability appear as the factors that explain transparency from a technical aspect. On the other hand, seen in socio-political terms, this principle refers to the administrative openness of systems with the aim of generating legitimacy and trust that allows citizens to participate in their co-design.
The way artificial intelligence systems are arranged with human values, as well as error-free operation, understood as robustness, make safety and cybersecurity another key principle for cities. Whereas in the fairness principle, the misuse of these systems would result in citizens finding themselves in a situation of vulnerability due to the low level of cybersecurity provided by these systems.
While this Coalition stand by the principles of transparency and non-discrimination, as the Declaration states, another Coalitions´ principle figuring here is the protection of privacy. Those managing AI systems, whether local or national authorities or companies, must ensure that the systems not only do not violate the right to privacy in their application, but must also make proper use of the data, both in its protection and availability. To such a degree, personal data must be controlled by citizens and cannot be shared without their explicit consent, as cities such as Helsinki are already doing with their MyData project.
Sustainability has turned an ethical principle coming from the field of ecology into a driving vector in our society due to the major current problems we are facing, digital matters included. In the case of its application to AI, cities must ensure that it is done in the most respectful manner with ecosystems, but also taking advantage of the benefits provided to contribute to their environmental policies.
Finally, and related to the principles of transparency and openness, is that of accountability. The openness of these systems allows for better correction of system failures, as well as accountability to citizens by fulfilling the right of explanation.