A context-aware trustworthy AI governance framework for smart city applications in Bangladesh

Abstract

In the pursuit of attaining smarter cities, the appointment of artificial intelligence (AI) has become ubiquitous and imminent. However, adoption has also raised concerns regarding accountability, fairness, transparency, and public trust, especially in developing countries where regulatory bodies are less equipped and not in pace with the advancement of technology. In Bangladesh, Digital Bangladesh and Smart Bangladesh 2041 initiatives have driven the development of smart cities, but it lacks a regulated governance framework to steer AI adoption in public sector urban applications in traffic management, disaster response, and surveillance. The paper presents a tailored framework for smart city applications in Bangladesh, named a 'Context-aware Trustworthy AI Governance Framework'. The principle adopted to create the framework has been derived through a systematic approach of comparative analysis of the present authoritative international AI governance instrument, such as the EU Artificial Intelligence Act, UNESCO's Recommendation on the Ethics of Artificial Intelligence, and finally, the EU High-level Expert Group's Trustworthy AI Guidelines. Five core principles have been recurring as qualitative thematic coding was applied in the aforementioned literature. The core principles are: explainability, reliability, fairness, data privacy, and human supervision. These core principles are then reformed through contextual constraints and feasibility to denote Bangladesh's socio-economic, infrastructural, and institutional maturity. The framework's suitability is shown through illustrative case studies which involve AI-based traffic congestion prediction, an early flood-warning system, and AI-assisted urban Surveillance. Furthermore, for validation of the concept, scenario-based pilot implementation has been used to evaluate accountability, exposure to ethical risk, and Governance readiness.