Responsible AI: Advanced Level certification by LabelIA for Ekimetrics

Issued by Labelia Labs, an independent association that fosters the development of trustworthy data science ecosystems, Ekimetrics was awarded the “Labelia – Responsible and Trustworthy AI” certificate, Advanced level, on May 25, 2023. It represents a very high level of maturity in Responsible AI practices across more than 35 criteria.

Emerging needs for the responsible use of AI

On June 14, 2023, the EU Parliament passed the AI Act, a significant milestone in AI regulation that concerns not only EU companies but also international companies doing business in the EU. The act aims to ensure the regulation of AI systems used in the EU in terms of safety, transparency, traceability, non-discrimination, and environmental aspects.

While it has yet to be passed into law in EU countries, businesses can anticipate that they will soon be required to submit risk assessments regarding their use of AI, to provide more information about how the AI tools are trained, and to disclose the nature of interactions between AI and users.

Initiatives that led us to the Advanced level certification

It is Ekimetrics’ strong conviction that data science processes and practices need to be responsible and trustworthy. To us, responsible AI is about transparency and a limited negative social and environmental impact. As part of our iterative improvements regarding the sustainable and ethical practice of Data Science and AI, we initiated the 3-phase audit process of obtaining LabelIA, an independent recognition of collaborative, responsible, and trustworthy data science. Some of the good practices, processes and tools that we have put in place include:

• Defining, documenting, making available and providing training around the 7 pillars of Responsible AI (Transparency, Security, Interpretability, Explainability, Vigilance, Robustness, Sustainability)
• Evaluating projects with a criticality checklist in the early stages: by asking questions about the project’s background and objectives, the checklist helps determine its potential impacts and the risks that will have to be monitored.
• Developing tools to assist consultants when they apply these standards: cheat sheets for developers, model card creation tools, energy consumption tracking tools, etc.

Obtaining this certification at the advanced level is a recognition that we are regulation-ready. Over the years, the approach we adopted around AI and data science is responsible by design, both internally with our Eki.People and process, and externally, with our clients.

We integrated carbon footprint measurement into our platform and the development of dedicated custom tools is currently a work in progress. We have also implemented several practices to reduce the consumption of data storage, datal handling and modeling. The relevant standards are available in our internal knowledge management system and the model card tool will soon be integrated in our core data science platform and available for all our projects.

We have now applied the best practices for Responsible AI in some of our missions. We helped a client in the beauty industry write their Responsible AI standards. We’ve also used SHAP and EBM to assist clients in the food industry to generate reliable forecasts and support their store location decisions.

We are proud to be able to accompany our clients on this transformative journey. We will continue to allow others to benefit from our know-how and offer our clients and partners the means to scale the responsible deployment of AI and technology in the long term.