Home > Pathways
Five Pathways to Better AI
Where there are crowds, a crisis is never far away. One thunderstorm, for example, and panic can erupt on an otherwise calm summer beach. In such moments, it’s essential for crowd managers to have a clear, real-time understanding of the situation—and to act swiftly. Artificial intelligence (AI) is becoming indispensable in this context. AI Compass is developing AI tools that are not only powerful, but also safe, transparent, and responsible.
To make this possible, we follow five interconnected pathways:

1. Design Guidelines for AI
Drawing on the latest EU legislation—the AI Act—we are developing new guidelines for the design of AI tools, including real-world use cases. These tools must be explainable, accountable, ethical, and lawful. We are also introducing a review mechanism to ensure ongoing compliance with all relevant standards.

2. Protocols for Information Provision
The information delivered by decision support systems must be tailored to the user. It should be not only understandable but also context-aware: what is relevant right now, in this situation? We are creating protocols and tools that ensure users get the right information at the right time.

3. Framework for Human-AI Collaboration
Coordinating often decentralized crowd management teams is already complex. The introduction of (semi-)autonomous AI systems adds a new layer of challenge. We are developing a collaboration framework that enables effective interaction between humans and AI systems—with humans always in charge.

4. Tools for Co-Creation
Crowd management involves many different stakeholders: from police and municipalities to local businesses and residents. To ensure that AI tools address the needs of all, we are building tools for co-creation, joint learning, critical reflection, and structured feedback. This helps foster trust, acceptance, and long-term support for AI-assisted solutions.

5. Improved AI Tools
Throughout the project, we are developing new, data-driven AI tools that provide real-time insights into crowd dynamics using data from sensors, video, and other sources. These tools will also support predictive capabilities, helping to anticipate developments in and around crowded areas. Their design will continuously incorporate input from the other four pathways.