#SPATIALsolutions – Introducing the COMPASS framework

The Erasmus University of Rotterdam, one of the SPATIAL project partners, has created a tool called COMPASS.

The COMPASS framework enables organizations to critically evaluate both technical innovation potentials and societal impacts of AI systems to ensure a responsible and trustworthy AI system. COMPASS provides a roadmap for AI developers and auditors to navigate the complex landscape of AI systems and their impact on society. This framework offers flexibility for AI practitioners to tailor an evaluation process to the industries` specific needs and priorities through a self-assessment process. Using the SPATIAL design principles helps identify where to leverage necessary skills.

By incorporating all components of COMPASS into the life cycle of AI systems, trust, fairness, and societal benefits can be fostered in the AI era.

COMPASS stands for:

CONTEXT: Defining and determining the context of the AI system, including who is involved in its lifecycle and who is affected by the system.
OPENNESS AND TRANSPARENCY: Ensuring AI systems and algorithms are transparent and understandable for all stakeholders, considering access and documentation.

MEASURES: Iteratively developing mechanisms for evaluating AI systems so that they operate fairly and reliably.
PRIVACY POTENTIALS: Safeguarding user privacy and data protection throughout the AI life –cycle.
ACCOUNTABILITY: Highlighting trustworthiness and holding AI systems and developers accountable for their actions.
SECURITY and SAFETY: Implementing security measures and safety precautions throughout the AI life cycle to minimize potential attacks.
SUSTAINABILITY: Integrating mechanisms to maintain the reliability and performance of AI systems while providing sustainable and environmentally friendly AI solutions.

The adoption of COMPASS is intended as a self-reflexive evaluation process to develop reliable, bias-reduced processes throughout the AI lifecycle. Strengths and weaknesses in AI development and deployment become immediately visible to define areas for improvement or development. COMPASS allows an organisation to:

identify weaknesses at every stage of the life cycle,
determine potential reasons for these weaknesses (e.g. biased training data, inadequate tracking systems),
implement best practices through guidelines and resources.

COMPASS provides an actionable evaluation process that guides users towards effective, trustworthy AI development processes.