“You understand nothing!”, “what experience you have in decision making?” , “it doesn’t work this way or that way”, “you need to be extremely knowledgeable to make decisions”, etc. The list is endless, literally! These are the responses an employee can easily get when he has a suggestion for management. Agree with it or not, Since long, management decision-making, beyond any doubt, has slowly been made a cult!
Nobody is born super-wise, born super-intelligent, or a born superman. In fact, nobody can ever become super-wise, super-intelligent or a superman.
In the world that changes with the lightening speed, the rate of obsolescence of experience, technology or even a knowledge is rapidly increasing.
Bankers who have 10+ years of experience are becoming obsolete now with the advent of newer products like, payment banking, e-commerce partnerships, credit card schemes and offerings, customer intelligence tools, ever changing legal regulations and provisions. All management decisions are increasingly becoming heavily relied on data. This is just not for banking. It is a universal phenomenon.
In machine learning, there are various types of diversity. Diversity in training data enhances the discriminatory power ( predictive power) of the model. The various combinations of parameter values in a labelled training data, help the model to point in certain direction which is the predicted probability of outcome. Or simply a prediction. Cross validation, Resampling are some of the techniques.
Diversity in model parameters focuses on specific subsets of input parameters that could help in a better way to point to the prediction. Random forests is the best example. Every tree is composed of diverse subset of inputs. Then through voting, the best prediction is made.
Diversity in models also allows us to model the problem better. There are models arranged in a stack. Every model passes certain information to the other model thereby achieving enhanced predictive power. Or we can also think of the ensemble model which takes the individual predictions made by the models that it contains.
Diversity in inference helps us in local interpretability. It might reveal certain distribution of outcomes pertaining to the valid ranges of inputs. We can then choose inferences that are particular to a certain subspace. We can think of psephology as one of the application domains.
It only corroborates the fact that the diverse opinions are to be welcomed! In fact they are invaluable in the overall decision making process of the organization. Only a single mind or a few powerful minds, however experienced they might be, are totally insufficient to make valuable decisions. History is witness to various serious management failures.
Every voice is a key stakeholder. Just including a bunch of “selected” representatives to “comply” with “inclusion” guidelines is like running a fool’s errand!
With the advent of ML techniques, it is possible to include every voice and make the decisions that work best for the organization. We can always ruminate about how to democratize AI. Till that time, we can always use AI to democratize management decision-making!
Reinvent The Self! 