Lujain Ibrahim, Luc Rocher and Ana Valdivia
The adoption of AI applications has increased user interactions through advanced interfaces, which significantly shape user behavior and perceptions. However, evaluations of AI systems often overlook the impact of anthropomorphic, deceptive, and immersive interfaces. The new conceptual model, Design-Enhanced Control of AI systems (DECAI), addresses this gap by using principles from control systems theory to assess AI interface designs. DECAI reveals the broader impacts of interface design, as demonstrated in a case study on conversational language models, and highlights the importance of fairness, transparency, and trust in AI interactions.











































