The HP Omnicept Cognitive Load inference is looking for patterns in biometric signals.
As of Omnicept 1.10, it may be possible to increase or decrease cognitive load scores without actually changing cognitive load by, for example, rapidly moving your eye gaze around the virtual scene, holding your breath, closing one eye for a long period of time, or other deliberate behaviors not related to the task you’re completing. This is because Omnicept’s predictions are based (in part) on biometric response patterns from real people completing cognitively demanding tasks. If a user behaves in a way that systematically changes their biometric responses, artificial changes to cognitive load predictions can occur. However, “gaming” the cognitive load reading is more likely to increase the overall uncertainty of the model’s predictions (leading to false peaks and valleys and higher standard deviations) than to artificially predict optimal cognitive load levels (a.k.a. the “goldilocks zone”).
The HP Omnicept team is continually improving the inference engine to increase its robustness against all sources of noise. This includes adding additional biometric features and tweaking our algorithms to be more sensitive to manipulation. As more features are integrated, the performance of the inferences in different contexts will continue to improve.
Comments welcome below.