We partnered with 10 European AI businesses (based in or holding key operations in Europe/across the EU) to co-create and test an AI risk assessment framework on different AI applications. – We called that assessment “ADIA”, Automated Decision Impact Assessment.
Participating companies were asked to select an (in-house) AI/ML application that would produce effects or have an impact on people, to simulate the application of the ADIA framework on that particular application. We adopted a design sprint-inspired four week prototyping methodology, focusing on the participants’ journey when simulating risks and thereby exploring implications for policy understanding, policy effectiveness, and policy costs.
The results of this initial policy prototyping program clearly demonstrate that performing such assessments in practice is a valuable tool for companies to identify and mitigate risks from AI/ADM (automated decision-making) systems.
- Focus on procedure instead of prescription as a way to determine high risk AI applications
- Provide specific and detailed guidance on how to implement an ADIA process, and release it alongside the law
- Be as specific as possible in the definition of risks within regulatory scope
- Improve documentation of risk assessment and decision-making processes by justifying the selection of mitigation measures
- Develop a sound taxonomy of the different AI actors involved in risk assessment
- Specify, as much as possible, the set of values that may be impacted by AI/ADM and provide guidance on how they may be in tension with one another
- Don’t reinvent the wheel; combine new risk assessment processes with established ones to improve the overall approach.
- Leverage a procedural risk assessment approach to determine what is the right set of regulatory requirements to apply to organisations deploying AI applications (instead of applying all of them by default)
AI Impact Assessment: A Policy Prototyping Experiment
This report presents the findings and recommendations of the Open Loop’s policy prototyping program on AI Impact assessment, which was rolled out in Europe from September to November 2020.
As the report outlines, the results of Open Loop’s first policy prototyping experiment were very promising. Based on feedback from the companies we collaborated with, our prototype version of a law requiring AI risk assessments, combined with a playbook for how to implement it, was clearly valuable to the participants as a tool for identifying and mitigating risks from their AI applications that they may not have addressed otherwise.
The experiences of our partners highlighted how this sort of risk assessment approach can inform a more flexible, practicable, and innovative method to assessing and managing AI risks compared to more prescriptive policy approaches.