DC1. How to implement justice in AI for road safety?

Host: TUD & RHDHV

Objectives:

• To investigate the conditions to achieve both safety and justice on the road through AI
• To embed non-discrimination and inclusiveness in the design of AI technologies for road safety
• To disentangle the role of explainability for justice in AI from the user and the regulatory agency and policy-maker perspectives

Explored from a theoretical perspective within the recruitment at TUD and from a practical perspective (operationalisation on existing tools e.g., the Strategic Road Safety Policy tool, the Flowtrack smart traffic lights system) within the recruitment at RHDHV.

Expected results:

• A new theoretical framework to assess justice on the road, through data justice and attention to VRUs
• A set of recommendations for promoting justice by design, e.g., achieving a better inclusivity of vulnerable or excluded groups (e.g., pedestrians, cyclists, LMICs, women, and low-income households)
• A set of guidelines to design fair data protocols (e.g., data protection, user profiling) and algorithms (e.g., explainability, goal-oriented planning) in the operation stage of AI in road safety

Planned secondment(s):

• TNO-NL, Purpose: to explore the existing experience of TNO on goal-oriented planning for ethical and safe AI in autonomous systems on the road and in other domains.
• RHDHV-RSA, Purpose: international exposure and explore ethical issues of AI in a non-European LMIC (Republic of South Africa). Preliminary agreement with the RHDHV-RSA office.