# The coming data science disaster By:: [[Brian Heath]] 2022-08-24 There is a problem lurking in the analytics world. It isn't messy or incomplete [[data]]. It isn't the need for better software or the accidental creation of computer consciousness. It isn't even business leaders not valuing it, because they generally do - or at least claim to. It's that data science as a discipline is operating in the world of illusion such that eventually, their very unscientific [[models]] will cause an unintended disaster in the real world. Many data science practitioners today operate from the position that [[understanding]] real-world systems is not necessary. With enough data and powerful computers, a model will emerge that accurately predicts an outcome. This seems reasonable enough. After all, computers are getting faster and faster while the world's systems are becoming more and more complex. So, why spend time trying to understand how the system works, when you can just let the computer do it? In other words, why take the time to create a hypothesis and test it as the scientific method would direct, when you can just jump to the end and not even need to explain the results? Well, because it's theoretically and practically wrong. It's theoretically wrong because this "toss it all in and see what happens" approach is a self-fulling prophecy not based on anything systemically tangible in the real world. If the models are built to find patterns, guess what they will find? Patterns. These patterns do not necessarily mean anything. They are likely random correlations found in the 13th dimension of data abstraction that doesn't resemble anything we observe. It's practically wrong because no one understands it. Thus, it's very unlikely that many people with skin in the game will risk it all on a model that no one understands. Those who do implement these models are doomed to eventually create an unintended disaster because all it takes is one system change for these models to go haywire. The data science field has traded thinking and understanding for apathy and speed. As painful as it is to [[critique]] the sexiest job of 2010, let's spend the time bringing it back to earth so data science powers can be used to combat and prevent [[disasters]] versus creating them. #### Related Items [[Data Science]] [[Analytics]] [[Science]] [[Scientific Method]] [[Computers]] [[Correlation vs Causation]] [[Systems Thinking]] [[Thinking]]