I appreciate revisiting the problem and attempting to break it down further; understanding the real problem is the first step to solving it.
Attentional Drift Diffusion Model (aDDM) shows how decisions are influenced by fluctuations in visual attention - https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5573732/. So, using aDDM along with a visual input sensor i.e. camera can be a key to build a solution to solve this problem.
Luckily subtle cameras on the face in the form of spectacles are within the reach of consumers now - https://www.spectacles.com/, with computation becoming cheaper day by day and with rapid advancements in machine learning; this problem has a real chance of getting solved sooner than we think.