Alp on Nostr: The Absurdity of Social Media Algorithms Using LinkedIn as an Example Recently, I ...
The Absurdity of Social Media Algorithms
Using LinkedIn as an Example
Recently, I read a post there about 2 guys who have been observing and analyzing the LinkedIn algorithm for five years and are now releasing an information brochure (PDF) that explains the algorithm in detail.
Among other things, it stated that one should also add photos to their posts because they are given more reach (by the algorithm). This means they are shown more frequently and to more users. The post text and images are analyzed by AI.
In the past, it was found that texts and images with a positive tone resonated better with people, so the AI is now specifically looking for indicators regarding the mood that a post evokes.
Example:
Photo with dark, brown, gray colors, low saturation -> negative mood
Photo with cheerful natural colors or people with smiling faces -> positive mood
So, if you attach a photo to your post that evokes a positive mood, it will be shown much more often. This is actually true; I’ve been running my own little tests.
So let’s recap:
1. LinkedIn folks discovered that positive images performed better (in the past, when there was no AI-powered algorithm).
2. They programmed their AI to recognize the features that evoke positive (or negative) moods (facial expressions, colors, subjects, etc.). If the result is positive, it shows the post to more people. The view numbers increase.
3. As the views go up, the experiential knowledge accumulates: "Positive photos are perceived by more people and generate more engagement."
And so on and so forth. Through the AI and the decision to give a post more or less reach, the network has entered a feedback loop. It would have been much better to leave this entirely to the users, like it's the case here on Nostr.
At least now I know that if I write something negative, like a rant, I need to attach a photo of a mischievously grinning Joker clown that consists of beautiful colors. 😈
Using LinkedIn as an Example
Recently, I read a post there about 2 guys who have been observing and analyzing the LinkedIn algorithm for five years and are now releasing an information brochure (PDF) that explains the algorithm in detail.
Among other things, it stated that one should also add photos to their posts because they are given more reach (by the algorithm). This means they are shown more frequently and to more users. The post text and images are analyzed by AI.
In the past, it was found that texts and images with a positive tone resonated better with people, so the AI is now specifically looking for indicators regarding the mood that a post evokes.
Example:
Photo with dark, brown, gray colors, low saturation -> negative mood
Photo with cheerful natural colors or people with smiling faces -> positive mood
So, if you attach a photo to your post that evokes a positive mood, it will be shown much more often. This is actually true; I’ve been running my own little tests.
So let’s recap:
1. LinkedIn folks discovered that positive images performed better (in the past, when there was no AI-powered algorithm).
2. They programmed their AI to recognize the features that evoke positive (or negative) moods (facial expressions, colors, subjects, etc.). If the result is positive, it shows the post to more people. The view numbers increase.
3. As the views go up, the experiential knowledge accumulates: "Positive photos are perceived by more people and generate more engagement."
And so on and so forth. Through the AI and the decision to give a post more or less reach, the network has entered a feedback loop. It would have been much better to leave this entirely to the users, like it's the case here on Nostr.
At least now I know that if I write something negative, like a rant, I need to attach a photo of a mischievously grinning Joker clown that consists of beautiful colors. 😈