Can robots truly engage in meaningful conversations with humans?
Researchers at KTH Royal Institute of Technology, are breaking new ground in Human-Robot Interaction. Dr. Bahar Irfan, Assist. Prof. Sanna Kuoppamäki, and Prof. Gabriel Skantze, latest studies tackle two critical challenges: making AI-driven companion robots more supportive for older adults and teaching robots the art of turn-taking in conversations. These pioneering works have now been recognised with two prestigious awards—highlighting their impact on the future of robotics and AI.
Can Robots Truly Hold a Conversation?
Can a robot talk with you as naturally as a human? This is one of the biggest challenges in human-robot interaction—one that researchers at KTH Royal Institute of Technology are tackling head-on. Their latest studies focus on two key challenges: how large language models (LLMs) can make AI-driven companion robots more suitable to the needs and preferences of older adults and how robots can master turn-taking for seamless dialogue.
Smarter Companion Robots for Older Adults
How can robots help combat loneliness and social isolation among older adults? The paper Recommendations for Designing Conversational Companion Robots with Older Adults through Foundation Models investigates older adults’ expectations on companion robots and how foundation models, such as LLMs and multimodal systems, can be used to allow for open-ended conversations in everyday life. A co-design study with 28 older adults revealed that they expect a companion robot to adapt to social context, remember past interactions, personalise responses, protect privacy, and show empathy.
"Out of 261 articles published last year, 11 papers were chosen. This paper was personally chosen for its significant impact in our field," says Prof. Kostas Kyriakopoulos, Field Chief Editor of Frontiers in Robotics and AI.
Teaching Robots When to Speak (and When to Listen)
At the 20th International Conference on Human-Robot Interaction (HRI) 2025 in Melbourne, the team presented groundbreaking research on how robots can predict conversational flow, earning them the Best Paper Award. Using AI models like TurnGPT and Voice Activity Projection (VAP), developed at KTH, they trained robots to anticipate conversational turns, reducing awkward pauses and interruptions. In tests, participants overwhelmingly preferred the AI-enhanced system.
"By predicting conversation flow, we make robots better listeners and speakers—just like a good human conversation partner," says Bahar Irfan.
Honouring Excellence in Human-Robot Interaction
Dr. Bahar Irfan, Prof. Gabriel Skantze, and Assist. Prof. Sanna Kuoppamäki have received two prestigious awards for their pioneering contributions. Their work is shaping the future of human-robot interaction, with potential applications beyond elderly care—spanning healthcare, customer service, and AI-driven assistants.
“We are honoured to see our research in Conversational AI and Human-Robot Interaction at KTH gain such significant international recognition,” says Gabriel Skantze.
Best Paper Award at HRI 2025 – for advancing AI-driven turn-taking in robots. Bahar Irfan, Gabriel Skantze.
Frontiers in Robotics and AI Editor’s Pick 2024 – for designing better companion robots for older adults with foundation models. Bahar Irfan, Sanna Kuoppamäki, and Gabriel Skantze.
Contact:
Text Maria Malmqvist