Beyond Command and Control: How Should We Rethink Human-AI Collaboration

On 28 January 2026, the THEMIS 5.0 project contributed to the online webinar “From Command to Collaboration: Reimagining the Human–AI Relationship”, jointly organised with the EU-funded projects THEMIS 5.0, TANGO, AI4REALNET, HumAInE, PEER AI and COMFORTage. The event brought together more than 50 participants from research, policy, and practice communities across Europe.

AI-Cluster Infographic

The webinar examined how human–AI interaction is moving beyond command-and-control paradigms toward collaborative, human-centred models, while preserving clear boundaries of responsibility, accountability, and oversight. These discussions closely reflected European priorities on trustworthy, value-aligned AI.

THEMIS 5.0’s contribution

During the session, Dimitris Apostolou (ICCS/NTUA) presented the THEMIS perspective on human–AI collaboration through a risk- and values-based lens, drawing on insights from the project’s work across the media, port, and healthcare sectors. A central message of the THEMIS contribution was that collaboration with AI should not be mistaken for autonomy, agency, or moral capacity.

Screenshot from the THEMIS presentation

While AI systems are becoming increasingly capable and may appear more agentic in their behaviour, they do not possess intentions, values, or goals of their own. Instead, they operate within objectives, constraints, and assumptions defined by humans. From the THEMIS perspective, making this distinction explicit is essential for preserving clarity around human responsibility and accountability.

THEMIS emphasised that trustworthy human–AI collaboration depends on making human values explicit, supporting informed human judgment, and ensuring that AI systems remain assessable, contestable, and accountable throughout their lifecycle.

Trustworthiness beyond technical performance

Across the project presentations and open discussion, participants consistently distinguished between AI reliability and AI trustworthiness. While technical accuracy and robustness are necessary, trustworthiness was discussed as a broader, multidimensional concept, encompassing ethical alignment, transparency, governance arrangements, and user empowerment.

From the THEMIS viewpoint, trustworthiness emerges not only from system performance, but from the interaction between technical design, organisational practices, governance frameworks, and the values and expectations of users and stakeholders.

Oversight, design, and user understanding

Speakers also stressed the continued importance of human-in-the-loop and human-on-the-loop mechanisms, particularly as AI systems are deployed in increasingly complex and sensitive domains. Oversight was discussed as a continuous practice spanning design, deployment, monitoring, and evaluation.

The discussion highlighted that design choices and terminology can influence how users understand and engage with AI-supported decisions. In particular, participants noted that overly anthropomorphic descriptions may affect user expectations and perceptions of responsibility, reinforcing the need for careful and precise framing of AI capabilities in real-world systems.

Human-AI collaboration conclusion

The webinar reaffirmed the relevance of the THEMIS 5.0 approach: meaningful human–AI collaboration delivers value only when human responsibility remains central, visible, and enforceable. Rather than focusing on what AI systems can do independently, the discussion emphasised how AI can best support human judgment within clearly defined governance and accountability structures.

Exchanges with sister projects underscored the importance of interdisciplinary cooperation and human-centred design in developing AI systems that are both effective and aligned with European values, societal expectations, and emerging regulatory frameworks.

🎥 Watch the full webinar recording: https://youtu.be/5H7lK_l-SnY