Dagstuhl-Seminar 27281
Human-Guided AI: The Role of Interactive Visualizations in Human-AI Teaming
( 11. Jul – 16. Jul, 2027 )
Permalink
Bitte benutzen Sie folgende Kurz-Url zum Verlinken dieser Seite:
https://www.dagstuhl.de/27281
Organisatoren
- Polo Chau (Georgia Institute of Technology - Atlanta, US)
- David S. Ebert (University of Arizona - Tucson, US)
- Mennatallah El-Assady (ETH Zürich, CH)
- Daniel A. Keim (Universität Konstanz, DE)
Kontakt
- Marsha Kleinbauer (für wissenschaftliche Fragen)
- Susanne Bach-Bernhard (für administrative Fragen)
Artificial intelligence is increasingly used across research and industry, making effective human-AI teaming ever more important. Yet, many AI systems rely on textual prompt-based interfaces with limited communi-cation bandwidth and little transparency into how outputs are produced.
This Dagstuhl Seminar will examine the role of interactive visualizations in improving human-AI teaming and discuss the field of human-guided AI. Visual interfaces can make AI systems more interpretable, under-standable, responsible, accountable, and fair – all crucial for increasing human trust. They can illuminate hidden processes through visual representations of model reasoning, confidence levels, potential biases, and data flow. Such transparency not only aids in debugging and refining prompts but also empowers users to make informed decisions based on AI outputs.
Such clarity is essential in domains where decisions have significant economic or societal impacts, where AI should augment rather than obscure human agency. Beyond technical benefits, effective human-AI teaming also depends on sociological and psychological factors, which must be considered alongside interface de-sign.
A central theme of this seminar is co-adaptive guidance between humans and AI systems. Humans bring contextual awareness, ethical considerations, and the ability to detect anomalies that AI systems might over-look. AI systems not only assist users by providing recommendations and insights but also adapt to users’ preferences, expertise, and goals over time. Interactive visualizations are a key medium for this exchange: they help AI communicate its reasoning and receive feedback from users to shape future system behavior.
The Dagstuhl Seminar aim to bring together researchers from relevant communities to discuss how visual interfaces can strengthen human-guided AI and enable more effective, transparent, and responsible human-AI collaboration.
Creative Commons BY 4.0
Polo Chau, David S. Ebert, Mennatallah El-Assady, and Daniel A. Keim
Verwandte Seminare
Klassifikation
- Graphics
- Human-Computer Interaction
- Machine Learning
Schlagworte
- Interactive Visualization
- Machine Learning
- Human-AI Teaming
- Human-Guided AI
- Understandability
- Responsibility
- and Trust
