Confidential doctoral project : No

Project title (thesis topic) : Visual-attention-based assessment of simulation fidelity in virtual reality

Description : Virtual reality (VR) simulations are becoming widespread in education and professional training, particularly for skills that are costly, hazardous, or otherwise difficult to teach. Their pedagogical effectiveness depends on their fidelity, that is, the degree to which they reproduce real objects and environments, both physically and functionally. A fidelity defect — a poorly modelled object, inconsistent scene lighting — can involuntarily capture the learner's attention and divert cognitive resources away from the learning task. The realism of a simulation is currently assessed mostly through subjective judgements, which allow neither the localisation nor the quantification of these disruptions.

The project proposes to measure these disruptions objectively from eye movements.

1. The first part of the thesis develops methods for analysing eye-tracking data specific to 3D environments and 3D models. These methods are validated through a first experimental study on indoor scenes containing controlled anomalies.
2. The second part, carried out with a postdoctoral researcher responsible for the EEG work, links these gaze features to the neural markers of distraction and results in a machine learning model predicting a distractibility index from eye-tracking data alone, made interpretable through explainability methods.
3. The third part designs and evaluates a learning analytics dashboard that reports this index back to teachers and learners.

This PhD will be carried out in the IEIAH team at the Laval site (Mayenne, France), in the CERIUMM building.

Keywords : data analysis - eye tracking - learning analytics - Virtual reality

Project issues

• Scientific challenge: moving from a subjective assessment of realism to an objective behavioural measure that can be localised.
• Methodological challenge: eye tracking in 3D environments still lacks an established framework for analysis and comparison. The metrics produced here extend beyond the project and will serve the whole community working on gaze in VR.
• Applied challenge: a predictive model based on gaze alone dispenses with EEG and becomes deployable in real training settings (eye tracking now being built into headsets).
• Educational and societal stakes: giving designers and teachers an actionable diagnosis of their virtual learning scenarios; supporting the replacement of costly or hazardous training (laboratory animals, industrial equipment, high-risk situations).
• For the doctoral candidate: a rare position at the interface of computer graphics, AI, vision sciences and technology-enhanced learning, with the possibility of publishing in four communities and career prospects in both academia and industry.

To apply for this position, please review the full job description by clicking on  this link.

Applications must be submitted by November 1, 2026, at the latest, by clicking on  this link.