The paper "A.I. Robustness: A Human-Centered Perspective on Technological Challenges and Opportunities", written by Andrea Tocchetti and his colleagues, looks at different challenges in ensuring the reliability and safety of artificial intelligence systems. Despite the remarkable capabilities of AI, issues of robustness—meaning how well these systems perform in unpredictable situations—remain a significant barrier to widespread use.

The authors look more closely at the term 'robustness' by reviewing existing literature and developing a structured framework to categorise various approaches and methodologies. They identify three main areas of focus: methods to improve robustness within the machine learning process, specific applications of robust AI systems, and techniques for assessing their reliability.

A notable aspect of the paper is the emphasis on the need for a human-centered approach in AI development, which is highly relevant in the context of the AI-PROGNOSIS ecosystem. The authors argue that human insights and collaboration are essential for enhancing AI robustness, rather than relying solely on algorithm-driven solutions. They find that the current frameworks often lack specific guidance on how to effectively integrate human knowledge and experience into the development process.

By reviewing over 380 studies, the authors reveal that much of the research is concentrated on computer vision, while other domains, such as natural language processing and cybersecurity, are underexplored. The paper calls for a broader, multidisciplinary approach that considers human factors and real-world applications, ultimately advocating for research that integrates these important social elements into AI systems to better address their robustness.