Welcome to the third blog of the DISCOVER THE SPATIAL WORK series, where we dive into the SPATIAL work packages with our WP Leaders!

Let’s analyse WP3 – System architecture, consistency and accountability for AI, Validation and Testing, where Huber Flores, Associate Professor at the University of Tartu, WP3 leader, will present the work done, the impact generated by this WP and the plan for the coming months.

  • Main activities carried out in the first project period.

The main activities of WP3 are deployment and integration of the SPATIAL developments.  The envisioned SPATIAL platform consists of a front-end and a back-end. The front-end fosters human-in-the-loop supervision and control by stakeholders, whereas the front-end contains all the metrics and services that analyze AI models and their respective datasets. Currently, the back-end is completely deployed and functional. Stress and capacity testing also has been performed to determine the platform’s performance. The back-end provides metrics that can be used to derive trustworthy properties of AI models, such as fairness and privacy. The front-end provides a basic visualisation of metrics and explanations, which is expected to be improved during the rest of the project.

  • How is this WP linked with the others? How are the activities carried out in this WP beneficial for the other project tasks?

WP3 is a core technical work package with direct links to all other WPs of the project. WP3 achieves the platform’s deployment with the technological design choices and functionality defined in WP1 and WP2. WP3 also exploits development in WP4 in an iterative manner such that it is possible to understand how to integrate human-in-the-loop within the solution and the type of stakeholders the solution needs to consider. The SPATIAL platform is subsequently passed from WP3 to WP5, so our industrial vendor partners can evaluate their use cases and provide feedback.

  • What is the part of this WP that can be the most impactful?

There are several impactful outcomes that we can mention:

-Synergies from different partners have come together to build a platform that can derive multiple trustworthy properties from AI models.

-A human-in-the-loop approach has been adopted to foster control over AI developments, which are also in compliance with the AI Act.

-We are contributing towards developing trustworthy AI, an ongoing process involving collaboration among technologists, policymakers, ethicists, and different stakeholders.

-Our developments also contribute to open-source and educational content.

  • What are your plans for the remaining part of the project?

Since the back-end of SPATIAL is deployed and functional, our remaining efforts will be focused on improving the front-end and aiding industrial partners to integrate their use cases into the platform. Moreover, the UT team is also committed to refining and validating the SPATIAL concepts throughout the entire project, ensuring its successful completion.