#Seven_deadly_sins of #datascience #AI #ML are the same as seven deadly sins of life - #pride, #greed, #wrath, #envy, #lust, #gluttony and #sloth. You ask how?
Let me explain and do avoid them if you can.
(1) Pride - if your data science model is selfish and thinks that it is superior to everything else (using any measure that justifies the end), it will lead to conclusions and actions which are biased and inferior. Many a time this is related to the data scientist's ego and self.
(2) Greed - if you want to build a data science model, which is better than the current one, even at the cost of losing explainability, at a higher cost, without any truly meaningful gain, greed has taken over. This can lead to data science which could be deemed as manipulative, full of trickery and using mathematical authority without a real need to do it.
(3) Wrath - if your data science project results in wrath of people who are affected by your model, due to incorrect individual level prediction, you are causing one of the seven deadly sins among the end users. Your model, instead of relaying on usual metrics like MSE, accuracy, F-1 score etc., should address individual level issues that can arise from your model. Do calibrate your predictive power with individual level predictions.
(4) Envy - if your data science model stays envious of all the success that other models have had, it will run into problem, due to envy. To compare your model with other published models, do ensure that like for like comparison is done, where the data, the business problem, and available computing power are aligned. Otherwise, good models will die, just because of envy.
(5) Lust - many models are data hungry and in many cases disorderly hungry. In order to satisfy that lust of data, unnatural use of large datasets, without considering the impact on business question at hand, without considering the effect of over-fitting and need of heavy computing, is done. This leads to bad data science for sure.
(6) Gluttony - is very similar to greed and lust and is seen when there are data science models, which in their desire to improve the metrics in question slightly, become a "gluttonous model", resulting in unnecessarily complex model. Avoid gluttony.
(7) Sloth - if your model results in actions which cannot be actioned, results in outcome which creates more confusion then clarity, creates situations which make decision making difficult, your are creating an atmosphere filled with sloth. As said in the original seven deadly sin definition of sloth, "a data science model which results in absence of interest or habitual disinclination to exertion, to do anything" needs to be avoided at any cost.
