A Classification Model for Sensing Human Trust in Machines Using EEG and GSR
summary
A real-time classifier-based trust sensor uses EEG and GSR to predict human trust during a driving-obstacle simulation, trained on behavioral labels from 523 MTurk trials and physiological recordings from 45 lab participants. A general trust-sensor model uses 12 common features and QDA to reach roughly 70% accuracy on held-out validation, while a customized trust-sensor model picks ~4.3 features per person, cutting training time per person and raising mean accuracy to about 78% (some individuals reach >99%).
AI-written summary · about
chapters (10)
- 0:00Start
- 1:07Trust Calibration
- 4:36Galvanic Skin Response
- 5:31Introduction of the Data Collection
- 8:19Feature Extraction for Eeg
- 8:52Gsr Signal
- 9:33Feature Selection
- 10:29General Trust Sensor Model
- 11:17Customized Stress Sensor Model
- 11:55Model Training and Validation
related talks
21m“Trust Us”: Mobile Phone Use Patterns Can Predict Individual Trust Propensity
31mMeasurement of Trust in AI
01h 01mKeynote: Using EEG and Machine Learning to Perform Lie Detection
06mThe Empathetic Car: Exploring Emotion Inference via Driver Behaviour and Traffic Context
26mBrainDrain Using Machine Learning and Brain Waves to Detect Errors in Human
01h 05mUsing EEG and Azure Machine Learning to Perform Lie Detection
07mEvaluating the Reproducibility of Physiological Stress Detection Models
46mKeynote: How Can We Trust Machine Learning?