Assalamualaikum and greetings dear Bits and Pieces readers,
I always view the combo of big data and artificial intelligence (AI) with askance. Despite its many pros that are evident in our daily life, big data can entice majority of humans, the possessors of authentic intelligence, to quickly jump to conclusions without so much of a thought. It's ironic actually, considering we have all the data. We may also become entirely dependent on algorithms to make decisions for us. At least, this is my opinion.
Now from this book, I learn that this deadly combo can also cause imbalance between IQ (intelligence quotient) and EQ (emotional quotient/emotional intelligence); leading to acts of oppression. This issue was pointed out ever so simply by the author, Cathy O'Neil (Except the part where the author made a comparison with baseball to which, as a Malaysian, can't quite relate. To generalize, Malaysians don't play baseball 😁)
O'Neil started the book by introducing the concept of mathematical model, a familiar topic for me. Basically, a mathematical model tries to express any complex phenomena like the economic system, an office day-to-day operation, the weather, basically anything, in mathematical terms. Mostly with the use of statistics and assumptions (together with big data), a mathematical model can provide you with outcomes, predictions, and explanations. The assumptions are of course, based on evidence. With big data and AI, a mathematical model can be built more complex than ever.
The issue arises when the assumptions were not changed despite evidence proving otherwise (repeated use of the same model) or the outcome/prediction seemed illogical (see now what happens when you abandon reason?).
One may ask, "How come these cause oppression?" To illustrate the harm these pose, allow me to present to you two case studies from this book.
Case Study 1 -Modelling teacher's competency
A teacher had scored 6 out of 100 on a teaching evaluation test. It's a ridiculous score because he had been teaching for 26 years. We suspect that the score is ridiculous out of moral ethics (EQ) and human reasoning (IQ) i.e. we recognize the teacher's experience and contribution that he made in spreading knowledge to his students. Curiously the next year, he scored 96 out of 100 despite maintaining his teaching methods.
Upon investigation, it was learned that in the year he got 6 out of 100, the teacher was teaching high achieving students and low achieving students. You see, it's difficult for low achieving students to score higher than their average marks. As for high achieving students, they had already scored high marks so they simply couldn't get higher marks. Therefore on the evaluation test, the non-increase in marks indicate that the teacher's teaching methods is abysmal.
On the other hand, in the year that the teacher scored 96 out of 100, he was teaching middle achievers. Middle achievers can either score below or higher than their average marks. That year, most were able to obtain higher than average marks. Due to this increase in marks, the evaluation test captured this as the teacher possessing excellent teaching methods.
What does this tells us? The mathematical model was inaccurate in evaluating the teacher's ability to teach! The model was practically useless! Allow me to emphasize here, the inaccuracy of the model was detected by recognizing the value of the teacher and applying human reason. This is something that AI and Big Data are incapable of.
Case Study 2 - The United States (US) subprime mortgage crisis
Using algorithms (AI) and data on home mortgages (Big Data), a mathematical model was developed to segregate these mortgages into tranches. These tranches are then packaged into securities before selling them off to investors. The mathematical model had predicted that profit can be generated this way. The thing is, the outcome/prediction was illogical because these mortgages are held by people who are not eligible for loan in the first place. As such, the securities will induce loss sooner or later to the investors (see again why you need to apply human reason?). The mathematicians had somehow forgot about human beings from the model and only saw potential profit (EQ vs IQ). Due to this, many people defaulted on their loans and US suffered a recession.
There are many other case studies that O'Neil mentioned in her book. O'Neil concluded that in order to avoid these acts of oppression, the mathematical model of any phenomena must constantly be revised to include recent information. The author also alluded to the importance of reason (authentic intelligence) and moral values before trusting the outcome/prediction of a certain model. Only by holding on to these two traits reason and moral values), we are able to execute justice in this lifetime.
Source: O'neil, C. (2017). Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy. Great Britain: Penguin Random House UK