Prejudice = Under-Fitting
an ML approach to a common problem of us, prejudice + what can we do?
hi, this is sina and i’m gonna write about a problem that almost all people have faced or will face sometime in their lives. so the problem is completely familiar. but my approach in this blog is not the same as the approaches you may have seen.
an introduction to the machine-learning-based approach
my approach is a machine-learning-based view of human learning. in this step, i want to explain a machine-learning concept that can be applied to some aspects of human life. this is under-fitting.
under-fitting is a common machine learning problem when you’re trying to create a model. it occurs when the model doesn’t have satisfying data to be well-generalized or it’s not strong enough to work well in the face of assessing data, that is called the test set.
to be more technical, the training set’s examples aren’t enough for the model to actually learn all of the examples that will come in the test set and your model will be “highly biased”. so its accuracy will be so bad and the answer, in many cases, is to replace the model with a more stronger model or make your dataset more rich (to learn more, see this).
isn’t this event familiar to you in your real life? i have been faced with it many many times and i will explain it in the following lines.
how do our brain and also ML models work?
think that we make a new dataset when we are learning a new thing. so at first, we have just one point and it’s obviously impossible to train a model (for judgment) with that one point. this is why we go further, to the second, third, and more points. we learn more and more and more, and we reach to a time when we can truly creat our model in our mind, completely based on the points we have already learned.
in fact, we build this model constantly and unconsciously during the time that we are adding the points. so as our dataset becomes richer, our model’s performance becomes better and better. this is how our brains and also ML models work!
when we become fanatics (biased)?
so we usually, in almost any concept, reach a point where we THINK that our model is good enough for our needs and that we don’t need to learn new things. it’s true. we don’t need new points RIGHT now, because our needs are completely met.
but the common mistake that most people will make is that they think today’s dataset and today’s model will be enough for future events. this is very very common and prejudice (especially in older adults) is exactly what came from here and why i want to say fanatic people are under-fitted!
what should we do?
so what should we do not to become a fanatic person? the answer is simple. always seek new data to let your model become better and better.
this is easy to say, but hard to act. because our brain isn’t always ok with new data, especially when the new data isn’t in the same way as the last ones. so you’d better proactively seek new data, especially the shape of data that aren’t the same as your previous thoughts!
i’m doing so since a few months ago and it truly helped me.
is there another problem that i should care about?
it may be obvious that constantly feeding your brain with new data may also bring another problem that i don’t want to delve into. something like
other people will say that your decisions don’t have “permanency”
you’ll have harder decision-making processes
and many other problems.
but here you can trust your intelligence! it’s a bit fancy, but completely true.
the most intelligent people i know have the skill to consciously ignore the data that they don’t want to use for their decision-making process. in fact, they train a model at the moment with the data they want to be their dataset and this is my definition of genius!
this is a little bit different with the term “skillful neglect” (which is a common term in the world of medicine) and i don’t know what term should i use for it. this is why i said “prejudice = under-fitting“.
it was always funny to apply mathematical approaches to real life. it’s a good way to understand how it works and i’ve already written a blog about mathematical metaphors in real life and it will be published soon. so stay tuned! :)
PS: honestly, this blog wasn’t mature enough to be published. but i published it to receive your feedback and opinions on prejudice and principle-based thinking. please send your thoughts to sina80mor@gmail.com and let’s chat!
sina
feb 17, 2024