The Global Observatory of Urban AI publishes the report “AI Ethics in policy and action: city governance of algorithmic-decision systems”
During Smart City Expo World Congress 2021, the Global Observatory of Urban AI has presented its first report on the ethical governance of algorithms in cities led by Andrea G. Rodríguez, project manager and researcher at CIDOB-Barcelona Centre for International Affairs. The report, which is entitled “AI Ethics in policy and action: city governance of algorithmic-decision systems”, is the first output of the deliverables that will be presented by The Global Observatory of Urban AI, which will be followed by an atlas of good AI practices in cities and a report on the use of facial recognition in public spaces.
Guiding Cities’ use of AI
Cities’ use of artificial intelligence (AI) systems is set to become the new normal. AI can help cities manage transportation, provide better water and energy services, improve the quality of the air or the speed of every-day bureaucracy. However, AI can have potential negative effects on the rights and freedoms of citizens that can be early identified and mitigated introducing the ethical lens on the design and use of AI in cities.
The field of ethics of artificial intelligence (AI) offers a set of guidelines about how to reduce the potential negative impact of AI systems to fundamental rights and freedoms, and how to ensure alignment with human values throughout the system’s lifecycle. However, there is no consensus about what role ethics should play while investing in algorithmic-decision systems nor what constitutes an ethical issue. Moreover, there is no international regulation that harmonises the ethical development of artificial intelligence systems.
Approaches to AI ethics
The report examines the most important documents dealing with the topic of AI ethics considering the geographical and industrial angles. Cities can learn from each other, as well as from the industry and the European, American, and Chinese approaches and adapt them to their necessities for their frontline applications affecting the life of citizens. In fact, there have been past approaches to the ethical application of disruptive technologies in the public space. Through the Sharing Cities Declaration cities agree to mitigate the negative impact of the platform economy while the Cities Coalition for Digital Rights has committed to a fair, ethical, and right-based digital transition.
While the European approach puts ethics at the centre of development, it is still risk based and focuses on high-risk applications—those that may cause harm and damage fundamental rights. The approach of the United States takes ethics as a complement to the technical fixes to misalignment and misbehaviour, and the Chinese approach offers the idea of “good AI” and sustainability but does not put people’s first but the good intentions of the developer. Section 1 of the report examines the approaches of cities to AI ethics and introduces the approaches to artificial intelligence of Microsoft and IBM, two important global players in the field of artificial intelligence closely connected with smart city solutions.
Minimal ethical standards for AI
Then, the report opens the analysis to a wider multidisciplinary portfolio and finds out that while there is no consensus of what should be considered “ethical”, there are a few topics that are commonly repeated in documents from governmental institutions, the business community, and think tanks and academia. Thus, the report proposes the consideration of these topics as minimal ethical standards and expands them offering new insights. The proposed framework considers that the following topics should be explored to mitigate the potential negative impact or AI:
- Fairness and non-discrimination
- Transparency and openness
- Safety and cybersecurity
- Privacy protection
- Sustainability and common good
The report also offers a self-assessment guide of compliance with the principles of the framework in the form of actionable questions that may also serve as a directory for capacity building.
To finish, it concludes that the lack of consensus and regulation about artificial intelligence is an opportunity for cities to discuss and adopt basic guidelines and incorporate them into their strategies creating an alternative model of governance for artificial intelligence systems. It nevertheless recommends periodical audits of the systems to enforce these principles and encourages cities to establish strong channels of communication with residents to identify possible variations in aligned behaviour.
Read the full report