Study Focusing on Artificial Intelligence Applicability for ITS Identifies 12 Challenges and Offers Lessons and Insights to Help Mitigate These Challenges.
Nationwide, United States
Artificial Intelligence (AI) for Intelligence Transportation Systems (ITS) Challenges and Potential Solutions, Insights, and Lessons Learned
Summary Information
Artificial Intelligence (AI), including Machine Learning (ML), has the potential to create opportunities to bring transportation systems to more desirable levels in terms of safety, equity, reliability, accessibility, security, efficiency, and resilience. While recognizing the transformative potential of AI and ML, it's also crucial to be mindful of the challenges they present, such as data related issues, supporting technology, bias, security, privacy, ethics and equity, generalization, model drift, explainability, liability, talent/workforce availability, and stakeholder perception. In light of these observations, this study, led by USDOT's Intelligent Transportation Systems Joint Program Office (ITS JPO), focused on the implications of challenges surrounding AI implementation for ITS, offering insights that agencies could take into account to help mitigate these challenges.
- Choose fairness metrics to measure AI model bias and collect sufficient data to assess error statistics across demographic groups. The optimal model isn't always the one with the least negative impact. Strategies like resampling or reweighting data from protected groups and gathering diverse data can help alleviate bias. Additionally, fairness metrics reflecting organizational values and the AI system's potential risk groups can be useful.
- Implement a socio-technical systems strategy to reduce bias in AI systems. Forming diverse AI development teams and implementing inclusive stakeholder processes that consider cultural dynamics and societal impacts can help mitigate bias. It is also important to continuously monitor bias mitigation throughout the AI system lifecycle.
- Enhance the safety of AI-based ITS systems by employing intrusion and misbehavior detection systems. These systems are commonly used in applications such as phishing email classification or fraud alert systems and can be adopted to ITS systems. Understanding potential security threats from misuse of AI-based applications is essential to better forecast, prevent and mitigate the threats. Facilitating collaboration among various stakeholders to identify transportation cybersecurity vulnerabilities and best practices can also be beneficial.
- Provide privacy for sensitive data. Obscuring/encrypting sensitive data or prioritizing collection of non-sensitive data, as well as synthetic data would minimize data privacy issues. In addition, policies should be in place regarding data sharing techniques and privacy protocols.
- Create AI systems with ethics, equity, and transparency at the forefront. These important values should also be translated into engineering aspects of the AI systems, with supporting workforce getting proper training and education to meet future AI needs. As always, including a human-in-the-loop would be essential for making critical decisions.
- Enable explainability for the AI system. Understanding potential tradeoffs between interpretability and performance, while balancing explainability with security and privacy would be essential towards this purpose. Outputting multiple performance metrics, as well as visualizing results are always considered as desirable methods towards enhancing explainability.
- Partner closely with agency risk management teams to consider legal and compliance issues from the perspective of organizational experts. This includes assessing legal restrictions for data and AI algorithms, as well as maintaining human accountability by assigning responsibility for AI system outcomes on specific individuals and organizations.
- Promote diversity in the AI workforce. This would help overcome challenges related to limited resources. In addition, conducting periodic education and training for existing staff, new hires, and domain experts would be supplementary in keeping up with advances in AI.
