Catalogue of Tools & Metrics for Trustworthy AI
These tools and metrics are designed to help AI actors develop and use trustworthy AI systems and applications that respect human rights and are fair, transparent, explainable, robust, secure and safe.
Organisation(s): united nations / unece grmunited nations
Following close participation by the Human-Ai.Institute's Founding Director Prof. Markus Krebsz (www.krebsz.net) within a United Nations working group (WP.6/UNECE GRM) and presentation of the GCRP at a conference in early 2017 in Germany, the group felt there could be value in removing Finance references from the GCRP in order to making it more generically applicable leading to the establishment of the 'Universal Conduct Risk Paradigm (UCRP)' - inspired by the bank/FI-specific GCRP - but with a much greater focus on NGOs, Multinational and Government bodies, non-financial corporates and in fact any legal entity for whom people work. Subsequently, the UCRP was published on the UN/UNECE website for free/open source and is available here:
Given the dynamics in the AI, NeuroTech and Robotics ('AiNR-Tech') domains, we would expect this to evolve further. Given that each of the boxes under the HAiPECR header “Protections” represents a separate scoring criterium, we are trying to maintain a fine balance between going too granular whilst at the same time being specific enough to understand an organization’s exposure to Ethics, Conduct and Risk as part of it acting as a producer of such AiNR-Tech Systems.
Further, we anticipate developing a similar quantitative scoring model that deploys a qualitative survey approach, to assess how a AiNR-tech producer develops and implements its system throughout the R&D / implementation lifecycle. Over time, we are also aiming to collect the outcomes of those audit assessments, useful for developing a peer-review process across both industries and geographies. The paradigm therefore will provide a normalised baseline with the aim of better understanding organizations maturity in deploying the ethical principles highlighted at the top half of HAiPECR.
Although we anticipate that the HAiPECR scoring method [to be developed] will not specifically reflect both those foundational pillars, we are looking for ways to address some of those elements as part of the survey for the protective mechanisms that we will assess. At this juncture, this remains work in progress, and we are happy to share more once we have concluded amongst ourselves if and how to best to address this.
Finally, it is also worthwhile highlighting the intrinsic tension between AiNR-Tech system producers (typically corporations), AiNR-Tech system consumers (typically individuals and/or legal entities), regulatory bodies (typically mandated by national governments and sometimes by the international organization by way of treaties or embargos) and society at large. HAiPECR represents this through the 2 overlapping triangles at the bottom half of the paradigm as a timely reminder that it will remain difficult satisfying everyone given the typically misaligned objectives.
We much appreciate the opportunity to contribute to such an important challenge for humanity and please do reach out to the Human-AI.Institute if you have any questions.
 The Global Conduct Risk Paradigm (GCRP, for banks): https://tinyurl.com/GCRP-Krebsz
 The Universal Conduct Risk Paradigm (UCRP, non-banks): https://tinyurl.com/UCRP-Krebsz
 UN WP.6 Annual report, bottom of page 3: https://unece.org/sites/default/files/2022-08/ECE_CTCS_WP6_2022_05_GRM_E.pdf
About the tool
- digital ethics
- ai assessment
- conduct risk paradigm