Scope and Objectives
Networks and digital infrastructure, such as telecom, transport, energy and water, have been underpinning the functioning of a society and economy. With the fast development of Internet of Things (IoT), Industrial IoT (IIoT), sensors, edge/cloud computing, and artificial intelligence/machine learning (AI/ML) technologies, many traditional critical infrastructure systems, which were isolated somehow, are now exposed, and connected to the Internet, transforming to digital infrastructure. Such a transformation heavily depends on efficient data collection, effective data analysis, and intelligent decision making, but can gain lots of benefits including creating intelligent services such as smart grid and offering a range of advantages for cost savings and efficiencies on network operation and security management. As the evolution and transformation move on, future networks and digital infrastructure would turn into increasingly complex systems.
Many AI/ML models have been developed to improve the efficiency of operating and managing such network and digital infrastructure systems, enhance the performance of smart services they support, as well as build up the security, robustness, resilience and reliability of the infrastructure. These models have not been widely adopted by network and digital infrastructure operators due to their potential ethical issues, such as fairness, transparency and accountability. These ethical issues make AI/ML models untrusted on managing network and digital infrastructure. Operators don’t want to be surprised with any unexpected actions given by models under certain infrastructure states. Insurance companies are also reluctant to trust ‘black-box’ models without sensible explainaibility. In addition, AI/ML models are introduced to automate various tasks for managing networks and digital infrastructure. If we target 100% automation, we must let AI/ML models themselves make decisions autonomously and they may encounter various ethical issues similar with the trolly problem. AI/ML models must be designed with moral theory and ethical principles considered at the outset.
To have a wider acceptance and implementation of AI/ML models in managing future networks and digital infrastructure, we need systematically search and study the related ethical issues and develop ethical AI/ML solutions. This requires cross-disciplinary research at the intersection of computing, engineering, and social science. This SIG will focus on the technical challenges of developing ethically compliant AI/ML solutions for future networks and digital infrastructure.
Topics of Interest
The areas of interests include, but are not limited to, the following:
- Explainable AI for future networks and digital infrastructure
- Fairness, transparent and ethical AI for future networks and digital infrastructure
- AI accountability in future networks and digital infrastructure
- AI safety in future networks and digital infrastructure
- AI vulnerability in future networks and digital infrastructure
- Robustness of AI in future networks and digital infrastructure
- Verification of AI in future networks and digital infrastructure
- Adversarial attacks and defence in AI for future networks and digital infrastructure
- Trustworthy interactive AI for future networks and digital infrastructure
- Human-AI interaction in future networks and digital infrastructure
- Human-centric networking and communication
- Human-centric digital infrastructure
- Law, policy, regulation and standards for ethical AI implementation in future networks and digital infrastructure
- Case studies of ethical AI in future networks and digital infrastructure
Proposed activities
The SIG will aim to hold conference/workshop calls 2-3 times per year and organise at least one physical/virtual meeting every year during the proposed conference/workshop or a CNOM related conference (NOMS and CNSM).
In the early stages, the SIG will aim to write a white paper that captures the global vision on ethical AI for future networks and digital infrastructure, and a range of detailed scenarios that ethical AI would produce significant impact.
The SIG will subsequently form smaller groups that will target peer reviewed publications on specific technical issues such as fairness, transparency and accountability of AI in future networks and digital infrastructure. The SIG may consider writing joint research proposals based on the innovation results from their joint research works.
In addition to providing a collaboration platform for joint publications and/or research proposals, a key activity of the SIG will be to organise several events including: (i) yearly conference/workshop, (ii) keynote speeches, panels and tutorials at conferences/workshops, and (iii) journal special issues.
Chair:
Yulei Wu, University of Exeter, UK
Vice-Chair:
Noura Limam, University of Waterloo, Canada
Yan Zhang, University of Oslo, Norway
Tarek Salhi, Vodafone, UK
Website: