Surrogate Learning

 Reading a biography is an easy way to learn from someone else's life. Granted the situations may not repeat but sometimes they do. Are there other ways of this surrogate learning?

In business and personal contexts, learning rules is easy. Most of the times they are straight forward, binary and crisp. Learning how to 'behave' in a context/situation is difficult as the number of active parameters is too high/unique. It is learning this human judgement that can help one avoid risks/errors and get better results. Before learning the behaviors, understanding the context is absolutely critical. No two contexts are exactly same and the same individual may make a different decision/choose a different action if one/more parameters change. Multiple actions/decisions in multiple contexts may help you distill/abstract the 'rules' that can help significantly.

More often than not, the contexts are not documented clearly /succintly. They are hidden in documents/numbers/conversations and actions. Reverse engineering the context from these is not easy but neither is extracting gold nuggets from mines.

Understanding the sources of context and actions is the first step. Choosing the contexts to learn from is important as not all situations have useful learnings.

Distilling active parameters, elimination of individual biases, abstract the contexts for better re-usability, explainability of learning are all the key next steps.

Context-Action-Learning models are not easy to build but its something that comes to us innately when we learn from others, from our childhood by observing/imitating/repeating. Bringing this to scale in an automated/structured framework is where the benefits of surrogate learning lie.