Monday 6 January 2014

General scheme for introducing sub-conscious feedback into wearable/persistent bio-monitoring technology

This idea assumes we have some sort of wearable tech. to capture EEG, EMG & environmental data over an extended time-period, plus a way of providing feedback: both explicit, conscious, attention-requiring feedback, as well as continuous, subtle, sub-conscious, non-attention-requiring feedback.

The basic plan is to try to establish two-way (conscious and/or sub-conscious) communication between a numerical optimization routine and the human brain, allowing us to experiment with the use of numerical optimization procedures to influence behaviour and help train the individual for increased discipline, learning and so on. I.e. to assert conscious control over unconscious aspects of thinking and behaviour.

Take EEG & EMG data & run it through some sort of dimensionality reduction algorithm, such as Hinton's Restricted Boltzmann Bachine based autoencoder network. Continuously update the reduced-dimension distribution and measure the entropy. Initially, optimize the feedback that is given to reduce the entropy in that space. I.e. the objective function for the feedback is to improve the brain-machine communications channel.

If this works, and the entropy reduces as a result of the feedback, then augment the objective function with some explicitly-designed factors to promote the aspects of behaviour that you desire. E.g. suppose we can identify that some part of the reduced-dimension feature space corresponds to stress. We could then try to optimise the feedback to reduce the amount of time spent in that space.

There are a million and one ways that we could riff off this basic idea.


No comments:

Post a Comment