Real-Time Interaction Quality Evaluation and Adaptive Strategies for Socially Interactive Agents
This PhD project at Xi’an Jiaotong-Liverpool University (XJTLU) focuses on developing a real-time, multi-dimensional framework for evaluating interaction quality between humans and Socially Interactive Agents (SIAs), including digital humans, embodied conversational agents, and virtual avatars. These agents are increasingly deployed in education, healthcare, entertainment, and service contexts, but current evaluation methods rely on post-interaction questionnaires and subjective feedback, which cannot capture dynamic changes during live interaction or provide immediate feedback for agent adaptation.
The project aims to operationalise and quantify interaction quality using multimodal behavioural signals such as facial expressions, gaze, gestures, speech, turn-taking patterns, response timing, and physiological or sensor-based signals. The research will combine Human-Computer Interaction, Human-Agent Interaction, multimodal machine learning, behavioural data analysis, and interactive system development. The successful candidate will design user studies, collect and analyse multimodal interaction data, develop computational models for interaction quality assessment, and explore adaptive agent behaviours based on real-time evaluation outputs. Adaptive strategies may include adjusting gaze patterns, gesture frequency, speech tone, response timing, interruption strategies, and dialogue pacing.
The expected outcome is a research framework and prototype system enabling SIAs to monitor interaction quality during live conversations and adapt their behaviours according to user responses and interaction context. This work supports the development of more natural, trustworthy, and user-centred interactive agents for real-world applications.
The project is technically demanding and interdisciplinary. Applicants should have a background in computer science, artificial intelligence, data science, human-computer interaction, cognitive science, psychology, or a closely related discipline. Essential skills include strong programming ability (preferably Python), prior training or project experience in machine learning, deep learning, data mining, or data-driven modelling, and experience processing complex data (video, audio, behavioural logs, sensor data, physiological signals, time-series, or questionnaire data). Basic knowledge of quantitative data analysis, statistical testing, model evaluation, and academic writing in English is expected. Experience in multimodal learning, affective computing, human-agent interaction, embodied conversational agents, digital humans, real-time sensing systems, Unity or Unreal Engine, MediaPipe, OpenFace, speech/audio analysis, motion capture, physiological signal processing, experimental design, or mixed-methods HCI research is highly desirable. Applicants should be prepared to conduct human-subject studies, including participant recruitment, experimental procedure design, questionnaire or interview design, behavioural data analysis, and research ethics preparation.
Applicants must have a Master's degree with Merit and a Bachelor's degree with first-class or upper second-class honors for PhD admissions. Exceptional candidates holding only a Bachelor's degree may be considered individually. Evidence of good spoken and written English is essential, with an IELTS (or equivalent) score of 6.5 or above required if the first language is not English. This position is open to all qualified candidates irrespective of nationality.
The PhD studentship is available for three years, covering tuition fees (currently RMB 99,000 per annum) and providing up to RMB 16,500 for participation at international conferences. Scholarship holders are expected to conduct the majority of their research at XJTLU in Suzhou, China, with the possibility of a short-term research visit to the University of Liverpool if required. Upon successful completion, the student will be awarded a PhD degree from the University of Liverpool (UK).
Applications are accepted year-round. Interested applicants should email [email protected] with the project title in the subject line and attach their CV, two formal reference letters, personal statement, certificates of English language qualifications, full academic transcripts, verified certificates of education qualifications, and a PDF copy of their Master Degree dissertation or equivalent writing sample. For further information, visit the XJTLU doctoral admissions and scholarship pages.
