Rain, Reflection and Results as Team THEMIS Meets at the IBM Innovation Lab in Paris-Saclay
- THEMIS 5.0
- 10 hours ago
- 4 min read
This week, the THEMIS 5.0 consortium gathered at the IBM Innovation Lab in Paris-Saclay for one of our final in-person project meetings. While the Paris weather greeted us with grey skies and persistent rain on day one, there was a certain symbolism in the sunshine that finally broke through on day two, a fitting reflection of our hard work that is steadily bringing greater clarity to the complex challenge of trustworthy AI.
Our meeting was opened by the Innovation Lab Director, Jean-Christophe Durand who introduced us to the cutting-edge work done within the building, and who reminded us of an old proverb:
Alone you can go fast. Together we can go further.
It was an appropriate way to begin discussions focused on integration, collaboration, and bringing together the many components of the THEMIS platform into a coherent and usable experience for future users.
Bringing the Pieces Together
The primary focus of the meeting was reviewing pilot testing activities and technical integration progress. Over the past years, partners have developed a range of tools, methodologies, and services designed to help people assess and improve trust in AI systems. Our challenge now is ensuring these components work seamlessly together.
Significant progress was demonstrated in the integration of multiple THEMIS tools into a single platform experience. With the next stage of integrated platform testing scheduled to begin in the coming weeks, our team spent considerable time reviewing workflows, user journeys, and how different components interact across the healthcare, ports, and media pilots.
Learning from Real Users
One of the most valuable aspects of pilot testing is discovering what users actually think, not what we researchers and developers assume they think.
A particularly interesting discussion centred around language and perception.
Within the healthcare pilot, we found that the use of the word morals generated an unexpected reaction. While the tool itself was not making moral judgements about users, some participants felt uncomfortable with terminology that appeared to suggest people could be categorised as moral or amoral. The lesson was clear: perception matters.
Even when a system is technically correct, users interpret language through their own experiences, expectations, and concerns. Building trustworthy AI is not only about algorithms and methodologies; it is also about communication, context, and user experience. As a result, we explored alternative terminology that better reflects the intent of the assessment without creating unintended interpretations.
Introducing the Super Decision Engine
A highlight of the meeting was IBM's presentation of the project's Super Decision Engine (SDE). The SDE has been designed to provide a more user-friendly way for people to interact with the various THEMIS tools and services. Rather than expecting users to understand which tool to use and when, the SDE acts as an intelligent orchestrator. Behind the scenes, dedicated large language models (LLMs) interact with specific THEMIS tools and capabilities. The primary interface used by the end user can then call upon these specialised agents as needed, refining requests and retrieving relevant information. This approach is particularly important because THEMIS is not being developed solely for AI developers and technical experts.
Our users include:
Healthcare professionals
Port operators
Journalists
Policymakers
Public sector organisations
These users need intuitive ways to assess AI trustworthiness without requiring detailed technical knowledge of the underlying tools. The Super Decision Engine represents an important step towards making sophisticated trustworthiness assessment capabilities accessible to a much broader audience.
Looking Ahead
With integrated platform testing about to begin, the project is entering an exciting phase. The coming weeks and months will provide valuable insights into how users interact with the full THEMIS platform and the SDE and how effectively our advanced methods, tools, and approaches support real-world decision-making.
As always, some of the most important lessons will likely come from unexpected places.
The discussions in Paris-Saclay reminded us that building trustworthy AI is as much about understanding people as it is about developing technology. And that is exactly what we aim to achieve with THEMIS,
More updates from the pilots, and the technology innovations, coming soon.
Finally, Thank You IBM Innovation Lab
A big thank you to the team at the IBM Innovation Lab in Paris-Saclay for hosting us and looking after us so well throughout the meeting. The success of a project is not only dependent on the expertise and dedication of the consortium partners. It also relies on the wider teams who create the environment that enables collaboration, creativity, and productive discussions.
From the meeting facilities and technical support to the logistics, hospitality, and warm welcome, the IBM Innovation Lab team helped create the conditions for a highly successful and enjoyable event.
As we move into the final phase of THEMIS and begin integrated platform testing, we are grateful for their support and contribution to bringing the consortium together.
Thank you from all the THEMIS partners.
Comments