top of page

Storyboard and Video

by Vira Gryaznova, Oct 04, 2021

Storyboard

Video

The audiotrack was recorded in Adobe Audition, the video was made in Adobe Premiere Pro.

Transcript

Are you a Netflix or Spotify user? Possibly yes. People do love their recommendation systems! Have you an idea of how they work?

When online, you are your personal data. Of course, they include your general profile data – your name, age, place of living, and so on. You may also know of the technical parameters of your computer or phone, as well as of an application for web surfing. Their combination is unique for any user, and it even can be enough to identify you.

But now we will watch at another side of this story: all your likes, shares, comments, searches, and other online activities that produce your so-called digital trace or digital footprint.

Your digital trace is also unique. But there are many other people with digital traces similar to yours, and it allows to perform collaborative filtering – one of the two main approaches for creating recommendation systems.

The general idea of collaborative filtering is simple. Let’s suppose that there is a popular service provider of media clips, named NotFlix. It can analyze digital traces of its users, group them into clusters and make predictions. One of such clusters contains Alice, Bob, and Charlie. All of them watched and liked clips 8, 12, 31, 77, and 92.

The system knows that Alice and Bob also watched and liked clips 24 and 45.

Then it is reasonable to suggest that Charlie would like them also.

When Charlie next time will log into the system, he will get a recommendation to watch clips 24 and 45 – and chances are good that he will be satisfied with this recommendation.

It was a very rough explanation, but you’ve hopefully grasped the main principle. Given that you are similar to me, if I like some piece of content, possibly you will like it too.

Naturally, the more information the system has about users, the more accurate clustering is, and the more satisfied with recommendations users become.   

So far, so good. But what is efficient for marketing or an entertainment platform can become a disaster in other circumstances.

What if the goal is not to give you a good recommendation for a movie to watch, but to determine the targets for a certain type of propaganda or fake news?

We will talk about it in the next video.

References
Credits

Images:
by Free Creative Stuff and Gerd Altmann from Pixabay;
by Alexander Shatov, Pankaj Patel, George Pagan III, Piotr Cichosz, Markus Spiske, Anne Nygård, Leon, Stephen Phillips - Hostreviews.co.uk, and Peter Forster on Unsplash.
Music:

“Digital Secrets” by Unicorn Heads.

bottom of page