Furhat is a back-projected humanoid social robot developed by the Swedish company Furhat Robotics. It uses projection-based facial rendering, speech synthesis and recognition, and programmed behavioral responses to support human–machine interaction. It has been deployed in sectors including education, healthcare, recruitment, and transport, and has been used in studies of social robotics at several universities.
Design and features Furhat is a humanoid robot composed of a physical bust with a translucent mask onto which facial features are projected. Facial expressions are back-projected onto this mask, allowing for the robot to change personas and emotional expressions without mechanical actuators. This projection-based system allows the display of different facial appearances and expressions without mechanical facial components.The system includes computer vision components for detecting and tracking faces within its field of view, and it can register multiple users at the same time. The robot produces visual cues such as facial expressions, head orientation, gaze direction, and eyebrow movement as part of its interaction capabilities. Its hardware comprises a movable neck structure, a wide-angle camera, stereo microphones, and a speaker. These features can detect user gaze, speech, and proximity. Developers can build conversational flows using a no-code or Software Development Kit (SDK) interface. The system has been referenced in academic studies related to human–robot interaction, including work examining gaze behavior, turn-taking, and interaction in group settings. Additional studies have considered variations in design and personalization in different application contexts.
Software and integration Furhat is used for prototype and application development, with researchers and developers able to update Furhat's code to test out verbal and non-verbal features. The robot operates using a modular software platform supported by an official SDK and developer tools provided by Furhat Robotics. Developers can build applications using a domain-specific language based on Kotlin, with built-in support for dialogue flows, intent recognition, and multimodal interaction. The SDK includes a simulator that mirrors the robot's behavior and facial animations, allowing for local testing before deployment. Furhat also provides a Remote API compatible with Python, C#, and JavaScript, enabling control of the robot from external systems over a network. Applications can be developed using integrated development environments such as IntelliJ IDEA and integrated with third-party services such as Amazon Polly and ElevenLabs. Furhat can also be integrated with generative AI models such as OpenAI's GPT-3.5. for creating conversational flows.
Applications Furhat featured at Bucharest Tech Week 2019 and has been used as a museum guide, for concierge services, and in other public engagement contexts. Furhat has also been incorporated in healthcare trials, with applications in autism therapy and mental health assessment. The most extensive application of the robot has been as a platform for academic research in human-robot interaction. Furhat has been used at:
KTH Royal Institute of Technology (Sweden) – early work in multimodal interaction and facial animation TU Graz (Austria) – natural language conversation and dialogue systems UK National Robotarium, Heriot-Watt University (UK) – research into conversational AI and human-robot interaction, including a public demonstrations as a GPT-powered receptionist, and a conversational robot used in a housing complex for older residents University of Manchester (UK) – research in cognitive robotics and healthcare communication University of Twente (Netherlands) – studies on bias and ethnic representation in human-robot interaction Uppsala University - studies and research in measuring children’s wellbeing and mental health with social robots, large language models-powered social robots in cybersecurity applications, co-designing a social robot facilitator to boost community engagement with type 2 diabetes prevention, and symbolic logic framework for situational awareness in mixed autonomy.
Named deployments Several Furhat-based robots have been used in applied settings. Examples include:
Tengai – Developed with the Swedish recruitment firm TNG, Tengai is a Furhat-based robot used in structured interview settings to reduce bias. It uses the Furhat platform but has a distinct personality and dialogue model tailored for recruitment. SEMMI – A multilingual travel assistant robot installed by Deutsche Bahn at German train stations. It engages with travelers in multiple languages and provides transportation services and travel advice. FRAnny – A localized version of SEMMI deployed at Frankfurt Airport. FRAnny provides personalized assistance and acts as concierge. PETRA – A conversational health-screening robot developed by Furhat Robotics, the Merck Group, and the design agency Prototyp. PETRA conducts pre-screening interactions with users to detect signs of diseases such as psoriasis and depression. Basil – A robotic barista employed in an old-age home to study human-robot interaction with older adults. BRILLO – A robotic architecture developed by researchers at the University of Naples that uses the Furhat head in combination with other robotic elements for bar-tending.
Public reception Heba Soffar of Science Online wrote that Furhat's mechanical abilities were limited by its lack of limbs and mobility and that its cost was higher compared to other social robots. Other commentators noted that hyperrealistic features—such as simulated empathy and personalized dialogue—may raise ethical concerns, particularly in emotionally sensitive contexts like hiring and healthcare. Studies have linked these reactions to the "uncanny valley" effect. Opinion is divided about whether Furhat succeeds to evade the "uncanny valley" effect, or whether users find it unsettling to engage with it. These discussions align with wider debates in AI ethics around transparency, overtrust, and anthropomorphism in social robots.
… excerpt ends here. Continue reading the full article.


