Neurodynamic Organization presentations
AI & Uncertainty
Machines Learn Human Uncertainty
How Real are Simulations?
How Real are Simulations?
Nearly every human decision, whether conscious or not, has its origins in uncertainty (Tononi, Boly, Massimini & Koch, 2016). Uncertainty in turn has its origins in what is meaningful for the system, in terms of either short or long-term survival. We have presented evidence for a candidate measure of uncertainty that applies equally well to individuals, teams, or even groups of teams that is quantitative, dynamic, has high resolution, and is understandable to external observers as well as those experiencing uncertainty. These are all characteristics that would be needed to teach a machine to recognize uncertainty, and to become a transparent and useful partner for humans. The classification of different dynamic trajectories of uncertainty is a beginning step towards these goals.
A paradigm shift is underway for understanding how teams are assembled, trained and supported. It is being driven by the generation of dynamic data streams using biometric sensors with seconds’ resolutions. It is expected that analyses of these data will shape the creation of new forms of simulations, performance measures and practices that are based on rapid neurodynamic and other physiological models.
For the most part these dynamic understandings will be derived from simulation studies. However, it is currently unknown at the neural / cognitive level how well simulation training elicits the types of dynamic thinking that is actually used by operating room physicians and staff during real-life neurosurgery, i.e. the ecological validity of simulation environments is unknown for neurodynamic measures. We begin to fill this gap by making quantitative comparisons of the neurodynamic organizations of individuals and teams in simulation and live operating room surgical situations.
Can machine learning be used to forecast the future uncertainty of military teams?
High stakes teams instinctively know when progress is being delayed and that uncertainty and hesitation are early indicators of potential disruption. The ability to rapidly measure the uncertainty of a team would have implications for future educational and training development as this intelligence could help target reflective discussions about past actions, support in-progress corrections, and feasibly generate forecasts about future uncertainty. Here we briefly describe an approach combining neurodynamics and machine learning to provide measures with these capabilities.
Quantitative models of information were derived from EEG brainwaves that provided detailed neurodynamic histories of US Navy Submarine Piloting and Navigation (SPAN) team members. Persistent periods (25-30s) of neurodynamic information were seen as discrete peaks during the taking of Rounds. These peaks were identified as periods of uncertainty by neural networks previously trained to recognize the frequency, magnitude and duration of different patterns of uncertainty in healthcare and student teams. Transition matrices of neural network SOM-states closely predicted the future uncertainty of the team during the three minutes prior to a grounding event.
These exploratory studies suggest that the dynamics of uncertainty may have common characteristics across teams and tasks and that forecasts of their short-term evolution can be estimated.