Algorithm isn’t the answer always

In 2009, Pilot Sully landed a full fledged plane in New York’s Hudson river when its engines malfunctioned. Although he saved a lot of lives but he was still criticized for taking such an extraordinary decision. Simulations showed that he could divert the plane to another nearby air strip. But as they show in the eponymous movie, that simulation forgets to take human struggle into account. When the plane malfunctioned pilots kept on trying for at least a minute to get it back on track. They could have panicked, they could’ve gotten cold feet, but instead of giving up, they tried for a minute or so to get the plane working. Simulations later showed that if you added that minute, it was not possible to reach the airstrip. So human mind’s strategy at that time was actually correct and simulation wasn’t.

I am studying Data Science and I work for a company making software for Retail Data Science. Data Mining, predictive modeling, training, etc. are the terms I see and read all day. There are factors at play in the world today which are forcing everyone to find the method to the madness from the huge chunks of data we have at our hands. But relying on data and algorithms is a two edged sword. In fact, whatever I am studying and working with professionally has certain limitations.

We know that Covid Vaccine is now being given to front line workers. In one of the cases, an algorithm was used to find out the order in which the vaccine should be given. The algorithm, which is obviously highly dependent on what we feed as initial parameters, chose some names of the Medical staff. However, it did a mistake which led to actual front line workers to protest. It didn’t chose names of staff who were on ICU duty and who had the real imminent danger of infection.

We should be grateful to the work done by scientists but we forget that an algorithm is as good as what it takes as input, how it is supposed to run models, and how much tested it is. Over relying on it blindly isn’t the answer and it might lead to more problems than it is supposed to solve.

I guess, data scientists and softwares have some tradeoffs when it comes to accuracy and the quality of data fed. And not all works with 0s and 1s. Human emotions aren’t tangible but still play a huge role in almost everything.

P.S.: This.