Active Learning is a specialist branch of Machine Learning, also known as Optimal Experimental Design, Query Learning or Adaptive Incremental Learning. It is about finding out “what is going on”, efficiently and reliably, starting from a position of relative ignorance, without wasting valuable time and resources.
It is a different thing from the “Active Learning” approach found in education, where each students actively engages in and directs their own learning experience.
In the real world, Active Learning’s main use is in the optimal learning or exploring of a poorly known ‘environment’ or ‘universe’, where each exploration is costly, dangerous or difficult. The outcome of active learning is an accurate and fast understanding of that environment.
You can also think of it as an efficient interrogator, asking the best sequence of questions from a large Question Bank of unanswered questions.
At Epifini, the ‘environment’ we want to understand is the way your business functions and is structured. Our system explores your business by asking informative Questions.
Our understanding develops from algorithmic interpretation of your answered Questions, ultimately contributing to a collection of composite scores in modelling your business.
These provide a broadly scoped diagnostic profile of your business in three high level Category scores: People, Leadership and Operations.
These Categories are in turn comprised of six more specific Element scores: Culture Spread and People Development, Leadership Effectiveness and Growth Strategies, and, Process Management and Execution Effectiveness.
Each Element is further comprised of four concrete, actionable Business Factors, which identifies business areas either require urgent attention, improvement, or are satisfactory or better. The system also advises on how to make changes that address specific weaknesses in the deficient Business Factors
The Active Learning algorithm selects questions from the Epifini Question Bank which was produced by our team of domain experts – executive coaches, business consultants and strategic planners. Each Question has been marked up with semantics details to reflect what topics each question is about, as well as rules for scoring and interpreting your answers. The semantic details of each Question also reflect a knowledge representation of the domain experts’ reasons for why they might ask that specific question, given what they currently know about the business and what more they want to find out.
The Active Learning algorithm optimises a Policy for selecting the next Question to ask from the Question Bank, based on the current level of understanding of the six elements of the business and what value the answer of each unasked question would provide.
Specifically, Active Learning’s Question selection is a policy-driven combination of:
The optimality of Question selection is designed to arrive at a stable, informative and actionable model of your business – a wide-based diagnosis of business performance and health – while asking the shortest sequence of questions, thereby saving business executives precious time while at the same time providing the most useful findings and guidance.
Every respondent will experience a different sequence of Question depending of the finding of Active Learning’s exploration. For this reason, the number of necessary Questions asked by the algorithm before compiling it’s focused Report will also vary depending on strengths and weakness of the business. Businesses in need of major changes generally are asked more questions, because there is more to find out.