Machine learning and stereotypes

Now, just like with human beings, ML doesn't always provide 100% accurate data but, you gotta admit, it does a pretty good job. Why is that? For one, machines are able to process much more data in far less time than humans. Secondly, the data fed to machines are usually less biased than those fed to humans. What does this mean? Let's look at the stereotypes we have about people from certain regions or with a certain appearance. A large portion of what we know is hearsay - the result of what we've been told by other people. Even when we create stereotypes based on personal experience, our experiences are rarely enough to make accurate guesses as that amount of data would hardly be enough to train a machine learning model to 50% accuracy. Machines, on the other hand, are able to make their decisions based on a LARGE collection of ACTUAL events. This is what makes their stereotypes more accurate.

But take heart, we humans are not entirely useless. Idiosyncratic to the human species, is our ability to analyse complex human actions. This is what we call "intuition". We can override our already existing biases by observing individual behaviours and picking up on hints that a machine would easily overlook.

Finding the balance in the use of stereotypes

In his book, ‘Zero to one’, Peter Thiel says. "Humans have the advantage of making sense out of complicated situations. Machines have the advantage of making sense out of large amounts of data”. In light of this observation, I’ll point out two possible steps to take when forming opinions:

• Mimic computers - Try to take in as much actual data as you can before coming to a conclusion. This will require some real research on our parts and not just blind information consumption.

• Use the human advantage - We have the beautiful ability to give people the benefit of the doubt and to observe distinct qualities that make an individual stand out from a group. Even when supposedly correct data places probability at 90%, our intuition can help us make up for the 10%.

Our intuition is of course, also biased, but it's a better solution than extrapolating the traits of the 90% to the whole group. If you have any ideas relating to this topic, I'd be glad to hear about it. Thanks for reading!

One term that's gotten quite mainstream in the last decade is ‘Stereotypes’. According to the APA dictionary of psychology, "This is a set of cognitive generalisations about the qualities and characteristics of the members of a group or social category". You probably know this already, but this word is rarely ever used to convey something in a positive light. The popular opinion these days is that all stereotypes should be disbanded and one should draw conclusions on an individual basis. I too have, at some point shared this opinion and like everything in life; that view has its pros and cons. I believe a more effectual solution lies in a more nuanced view of things. We can start by asking the following question:

Why do stereotypes exist?

In our bid to completely cast off the negative effect of stereotypes, we seldom consider the fact that they are, not just a constant but, an essential part of our daily lives. They are a part of our natural instincts to keep us from danger; helping us group people in ways that will be beneficial to the general population and to us as individuals. Let's take our food stereotype for example; we generally associate a kind of smell with food that can be eaten and another with food that is very likely harmful to the body. But even this fundamental classification has exceptions. There are many meals that have a repulsive smell and a delightful taste.

Another example is the voting age limit. In Nigeria, that limit is 18. This was based on the stereotypical assumption that the majority of people below 18 are not fit to make life-altering decisions. However, we all know fully well there are 16 year-olds who would do better with a voting card than some 60-year olds. Now imagine if, based on this assumption, it was determined that every individual, regardless of their age, would be personally assessed on voting ability before a voter’s card is issued. The point I'm trying to make is that stereotypes may not always be correct, but are not without use.

Machine learning

Another term that's picked up popularity in the last few years is ‘Machine learning’. If you haven't heard about this, then you might very likely be living under a rock. But just incase you haven't, "this is a branch of AI which is based on the idea that systems can learn from data, identify patterns and make decisions with minimal human intervention", according to sas.com. From this definition, you might have already begun to make the connections between machine learning (ML) and stereotypes. At first glance, ML and stereotypes seem to have nothing in common; until you realise that they both involve recognising patterns in a group and coming to conclusions based on those observations. I would put it this way - ML requires the formation of stereotypes. Ever wondered why your Spotify and Netflix suggestions are so great? It's because a machine made stereotypical assumptions about you based on similar data it had previously been fed.

Blog

Music