Skip to content

Programme

KnowRisk

Utilising advanced technologies to protect against supply chain risks and disruption

Supply chain disruption can have wide reaching impacts. For example, the COVID-19 pandemic made consumers keenly aware of this problem as the sudden changes brought about by the lockdown led to empty shelves at the retailers. There is urgent need to explore how advanced digital technologies can facilitate the smooth flow of goods.

Digital Catapult is part of a consortium that aims to reduce the risk and impact of supply chain disruption through the KnowRisk project. Utilising advanced technologies such as artificial intelligence (AI), distributed ledger technology (DLT) and geospatial intelligence (GEOINT), the project helps to assess and mitigate risks in a timely manner, acting as a proof of concept for future innovation.

The project involves ethics and policy interventions as well as AI based risk assessment:

Ethics: The KnowRisk project is the first opportunity to deliver specific AI ethics work as part of a CR&D project at Digital Catapult. The work is split into two parts:

  • An update of the Ethics Framework for a multi-stakeholder context and the application of the updated framework to the collaborative creation of an ethics roadmap for the consortium.
  • The development of Applied AI Ethics tools alongside our AI development work to increase transparency, namely model score cards for model reporting adapted for federated learning, and model robustness in the form of mitigations for adversarial attack and model corruption.

Policy: Digital Catapult will lead policy engagement sessions with government departments, regulatory bodies and businesses from the insurance, audit, construction and food and drink sectors. These sessions will inform policy development support. Findings and recommendations based on these policy sessions will be disseminated in a report, to inform both the KnowRisk project and the wider industries involved.

Risk assessment: With a need to produce an accurate and timely view of risks for insured assets such as construction sites, warehouses and office buildings, Digital Catapult will create a machine learning (ML) application to extract risks and mitigations from natural language insurance risk reports. The outcomes will be embedded in a federated learning (FL) system allowing data utilisation from participants without revealing confidential information. This has been released as dc_federated, an open source library for FL. The library is maintained by the DC AI and Engineering teams and is currently in beta version. It supports consortium level (~1000 workers) federated learning, with secure communication, worker authentication and worker management capabilities. The library can be viewed here. The project aims for this application to act as a prototype for alternative future risk score calculations, allowing for the inclusion of current information retrieved from sites.

The findings from the KnowRisk project will be compiled in a report including future recommendations.