I have been listening to George Soros recently, and i came across his reflexivity theory which has has interesting inductive explanations for human behavior. Mathematical models fail to capture human behavior because human behavior is not similar to particle behavior. Particles do not change their behavior when uncertainty is measured using Heisenberg's uncertainty principal, meanwhile if the same principal is applied to humans then humans would react to that model and would adjust their behavior in an unpredictable way which would act as positive feedback for the model. An example would be if a stock brokerage model predicts a looming crash then the stock broker would adjust his behavior and sell his stocks, this would in turn act as a positive feedback which would reinforce the models prediction and blowup the error because the model itself would not adjust according to the human behavior. The model would only be reliable as long as the particles behavior is not changed by the output of the model. A model wouldn't have predictive powers as it would always be one step behind the human ability to change their behavior, any prediction would be rendered useless as soon as that prediction is made due to the reflexive nature of the object which is under study.
Nassim Taleb's randomness principal states that humans behavior is pure randomness, but it fails to capture "why" this randomness cannot be captured in a model.