Advancing human-AI collaboration: dialog validation takes shape

Validation plays a central role in DIALOG's mission to redefine human-AI collaboration in air traffic management. On 6 November, the SESAR JU DIALOG project consortium convened virtually for a crucial workshop to refine its validation strategy. The focus was on aligning project objectives, defining short- and long-term outcomes, and planning the next steps to ensure the achievement of dialog's ambitious goals.

Co-funded within the framework of Horizon Europe, the project brings together the following organisations: Direction des Services de la Navigation Aérienne (DSNA), Office National d'Études et de Recherches Aérospatiales (ONERA), DFS Deutsche Flugsicherung GmbH (DFS), Stichting Radboud Universiteit (RU), SINTEF, and Deep Blue.

#aboutdialog: Spanning 30 months, from September 2024 to February 2027, DIALOG (deciphering intents of air traffic controllers, workload assessment and gaze analysis to enable their efficient and trustworthy collaboration with AI) aims to create an AI-based teamwork assistant capable of understanding air traffic controllers' (ATCOs) needs and providing precise, timely assistance. By enhancing human-AI collaboration, DIALOG seeks to improve operational efficiency and safety in air traffic management.

Why validation matters

Validation is not just a procedural milestone-it is fundamental to ensuring that DIALOG delivers on its promise of enhancing operational efficiency, reducing workload peaks, and building trust between humans and AI. The workshop provided an opportunity to address key questions:

• What are the most critical validation objectives?
• How can short-term progress guide the development of the teamwork assistant?
• What should long-term success look like?

These discussions set the foundation for ensuring the system meets real-world demands, paving the way for innovative human-AI collaboration in aviation.

Short- and long-term goals

The short-term outcomes for DIALOG are already taking shape. In the coming months, the consortium will focus on validating the core capabilities of the teamwork assistant, with an emphasis on improving recognition rates and streamlining task allocation. Partners will measure recognition accuracy, aiming for an 80% recognition rate of ATCOs' requests and refining real-time monitoring of ATCO workload and attention. These results are critical for determining the assistant's effectiveness in high-pressure environments, directly impacting operational efficiency.

Long-term goals are aligned with the project's larger vision: improving operational capacity, reducing workload during peak times, and enhancing the safety of air traffic control operations. Through ongoing refinement, the teamwork assistant will be tested in real-world scenarios, with a focus on building trust and transparency-ensuring the AI system is intuitive, easy to use, and an asset to ATCOs managing complex air traffic situations.

looking ahead: final validation and key deliverables

As the project progresses, DIALOG is set to undergo final validation at the conclusion of the testing phase. This will focus on evaluating overall system performance, including key factors such as efficiency gains, user acceptance, and the impact on workload reduction. By the end of the project, the goal is clear: deliver an AI-powered teamwork assistant that significantly reduces ATCOs' workload during peak times, boosts operational efficiency, and ensures a safe, transparent, and collaborative working environment between humans and AI.

Through these targeted short- and long-term goals, DIALOG aims for a future where air traffic management is not only more efficient but also smarter, more responsive, and more collaborative between humans and AI in the system.

Public link

Please use the above public link if you want to share this noodl on another website.