Atlas of AI Risks

ai-impact-assessment-methodology
One of the AI uses featured in the Atlas of AI Risks is a seemingly low-stakes job recommendation system. While it helps connect job seekers to employment opportunities, potentially boosting their income, it can also create problems, like biased job ads. In 2022, show blue-collar job ads to younger men, excluding older workers and women.

AI is already quietly entering all areas of life—making decisions about who gets a job, which videos are recommended, and even who might be flagged by facial recognition systems. But as AI's presence grows, so do questions about its risks. Could facial recognition be eroding privacy? Can AI systems be trusted to make decisions that affect people's livelihoods? The Atlas of AI Risks aims to help the public navigate these pressing concerns.

The Atlas of AI Risks is an interactive tool that lets users explore over 350 real-world AI applications. Unlike traditional visualizations that focus only on technical details, the Atlas takes a wider view. For each AI application, the Atlas classifies it by risk levels under the EU AI Act, ranging from low to unacceptable risk. It then provides an impact assessment card that breaks down the benefits, societal concerns, and real-world incidents linked to that application.

The visualization was built through a rigorous, crowdsourced approach, involving two studies with over 180 participants, along with consultations with legal experts. The first study helped identify how to best communicate AI risks to non-technical users. The second study tested the Atlas with a diverse group, ensuring that it not only presented risks and benefits clearly but also made them visually engaging and easy to navigate.

In a world where AI influences everything from policing to banking, understanding its risks and benefits is critical for everyone—not just experts. The Atlas of AI Risks fills a critical gap by making this information accessible to non-technical users, fostering informed public discussions about AI governance and ethics.

The Atlas of AI Risks: Launch demo

Publications

  • The Atlas of AI Risks: Enhancing Public Understanding of AI Risks AAAI HCOMP 2024 PDF | Supplementary Materials
  • The Atlas of AI Incidents in Mobile Computing: Visualizing the Risks and Benefits of AI Gone Mobiles ACM Mobile HCI 2024 PDF
  • Good Intentions, Risky Inventions: A Method for Assessing the Risks and Benefits of AI in Mobile and Wearable Uses. ACM Mobile HCI 2024 PDF
  • Decoding Real-World AI Incidents. IEEE Computer Project website