image1 image2 image3


Training our internal models

If you listen to songs on Spotify, or watch shows on Netflix, or read articles on Medium, or buy things on Amazon, you will see recommendations of more songs to listen to, more shows to watch, more articles to read, more things to buy, based on your previous interactions so far.

This is the field of data science and machine learning where algorithms and models are deployed to learn your preferences based on the actions that you take. Pretty much everything that you do (and don't do) is used to model your preferences and to come up with recommendations in the future. The items you click on, the items you spend more time on, the items you ignore all go in as inputs to the models along with things like your age, gender, time of day, how long you've been a subscriber and so on and so on.

While some of these features contribute heavily in computing your preferences, some others have meagre contribution. The strength of the models lies in getting these combinations right and is measured by whether you take the intended action when you are presented with these recommendations.

And Data Scientists obsess over getting the model to perform better and better for a desired outcome.

All of our subconscious behaviour (and much of our conscious behaviour) is the result of such models in our own neural systems as well. This is how our gut instinct is built up.

When you are about to get into a contract with a person you just met and you have that gut feeling telling you this person could be a shady partner to work with and you had better reconsider your decision to do business with them without being able to pinpoint a reason as to why exactly you feel that way, you are behaving much like a machine learning model that decides to recommend you watch 'House Of Cards' after you watched 'Orange Is The New Black' without being able to pinpoint a reason as to why exactly it picked this recommendation.

Every stimulus that we encounter during our days, every experience that we have, every person that we talk to, all act as inputs to our internal models that determine what our gut reactions to certain situations will be. Which is why babies need to be constantly watched over as they don't have any data to form strong gut instincts (except for those that are baked into their genes).

And yet, unlike data scientists, we don't obsess over these internal models to optimize them for the outcomes that we want. We focus on our shortcomings and failures and make it harder for ourselves to try again much more often than we focus on our strengths and successes to make it easier for ourselves to try again and try new things.

We ought to take a leaf out of the books of data scientists and start watching the inputs to our internal models and get the combinations right that lead to the most optimal outcomes for us.

Share this: