Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Mental State Sensing and the Goal of Circuit-Synapse Synergy Patrick L. Craven, Ph.D. Senior Member, Engineering Staff Advanced Technology Laboratories Cherry Hill, NJ Copyright © 2007 by Lockheed Martin Corporation. All rights reserved. Goals of Artificial Intelligence (AI) • Replicate human-like intelligence and cognition? • Media portrays this technology in different ways – Technology is Evil – 2001: A Space Odyssey (1968) – The Terminator (1984) – Technology is Good – The Terminator II (1991) – Short Circuit (1986), – “Number 5 is Alive” Copyright © 2007 by Lockheed Martin Corporation. All rights reserved. 10/31/2007 2 Human-Machine Synergy • Complement and enhance human cognition – Human cognition is different than computer processing – Brain anatomy provides us with powerful computation • Combining human and machine is advantageous – Firefox (1982) – The Matrix (1999) Copyright © 2007 by Lockheed Martin Corporation. All rights reserved. 10/31/2007 3 The Brain • Large – Neurons = 100,000,000,000! – Neural Connections = 100,000,000,000,000+ !!! • Electrochemical transmission – Axons – Dendrites – Neurotransmitters Copyright © 2007 by Lockheed Martin Corporation. All rights reserved. 10/31/2007 4 How do we measure brain activity? • • • • Electrical activity: EEG Bloodflow: fMRI, fNIR Skin response: GSR Heart Rate: EKG Copyright © 2007 by Lockheed Martin Corporation. All rights reserved. 10/31/2007 5 Collecting EEG • Measures electrical potential (µv microvolts) of cortex • 10-20 electrode placement – Even numbering: right hemisphere – Odd numbering : left hemisphere • Measured from nasion to inion (occipital bone) • Voltage between two points (ground or another sensor site) – Referential – Differential • Does not measure neuronal current (amps) Copyright © 2007 by Lockheed Martin Corporation. All rights reserved. 10/31/2007 6 Types of EEG waves • Delta: up to 3 Hz. Slow wave sleep • Theta: 4 to 7 Hz. Seen normally in young children. Seen also in drowsy adults • Alpha: 8 to 12 Hz. Observed in adults when awake and relaxed • Beta: 12 to ~30 Hz. Associated with active, busy, or anxious thinking • Gamma: 26+Hz. Not easily recordable with EEG. May indicate neuronal binding for common purpose Copyright © 2007 by Lockheed Martin Corporation. All rights reserved. 10/31/2007 7 Uses of EEG • Detecting abnormality – Head/brain Trauma – Seizure – Sleep problems • Applied EEG – Use time-domain or frequency-domain profiles to convert signal into a gauge of ‘mental states’ (drowsiness, cognitive workload) – Create closed-loop systems where mental activity influences software Copyright © 2007 by Lockheed Martin Corporation. All rights reserved. PSD Bins 1 4 7 10 13 16 19 22 25 28 31 34 37 40 Hz 10/31/2007 8 Current Technology - Applied EEG EEG Eyetracking EKG • Available cognitive state gauges – Cognitive workload – Visual engagement – Distraction – Drowsiness Copyright © 2007 by Lockheed Martin Corporation. All rights reserved. GSR 10/31/2007 9 Augmented Cognition • Goal: Maximize operator cognitive performance in dynamic, complex operational environments • Approach: Physiological-data based assessment of operator cognitive state – Detects, predicts, avoids overload to reduce operator error and maximize effectiveness User C2 System Sensors Copyright © 2007 by Lockheed Martin Corporation. All rights reserved. Cognitive State Assessor • Benefit: Mitigate negative effects of cognitive overload – Increase task speed and accuracy – Improve critical situation understanding 10/31/2007 10 Education and Training • Goal: Maximize effectiveness of training by customization based upon individual cognitive capacities and current cognitive state • Approach: Monitor trainees’ physiological data during training sessions – Identify and apply most efficacious training strategies User Training application CSS Training Monitor Sensors Copyright © 2007 by Lockheed Martin Corporation. All rights reserved. • Benefits: – Helps identify potential design problems – distinguishes workload-induced and confusioninduced errors – Provides objective, real-time, taskindependent workload measures 10/31/2007 11 Next Steps • Sensors: smaller, faster, more accurate, cheaper, easier to apply • Processing: quicker, cheaper, more portable • Algorithms: more efficient, more accurate Copyright © 2007 by Lockheed Martin Corporation. All rights reserved. 10/31/2007 12 Challenges Remaining • Technical – Gain a greater understanding of the tremendous individual variation • Ethical – Should we be doing this? • Practical – What information can it provide that we humans can’s share via actions or words? Copyright © 2007 by Lockheed Martin Corporation. All rights reserved. 10/31/2007 13