Meta: Algorithm evolution and AI in 2023

When it comes to performance on social networks, algorithms are often targeted as the source of our lack of visibility.

The cause: worrying figures, constantly falling on all Meta's social networks. On Facebook, theaverage engagement ratefor posts has plateaued at 0.060%. A hypothetical score, which raises questions about the value of publishing on the social network. Although Instagram has a higher engagement rate (between 0.47 and 3%), increasing your visibility on the network requires constant involvement and regular effort.

In a new issue, Nick Clegg, Meta's President of International Business, explains the importance of AI in the evolution of algorithms.

"Our AI systems predict the value of a piece of content to you, so we can show it to you sooner. For example, sharing an article is often an indication that you found it interesting. As you can imagine, no single prediction can perfectly assess the interest of a message for you. That's why we use a wide variety of combined predictions to get as close as possible to the right content, some based on behavior and others on user feedback received via surveys."

Meta algorithms: what do they eat?

According to Meta, three criteria are taken into account:

🔮 Where the post comes from: How engaged you are with a content creator.

🔮 Post time: The time at which the publication is posted is a key factor.

🔮 E ngagement probability: Meta algorithms estimate the performance of a post, which plays a major role in its dissemination.

AI: what role does it play in Meta's algorithms?

Meta explains that thanks to the power of its artificial intelligence, its algorithms will be able to analyze and interpret content holistically, across all available formats (image, text, video).

Meta explains:

"These production models offer capabilities such as visual recognition, object detection, text extraction and audio recognition. They also allow us to perform more specific tasks, such as theme/genre classification, hashtag prediction, similarity matching and clustering."

These improvements are interesting, as they will enable content to be better pushed to their potential targets. For the moment, this is only possible with the use of keywords and hashtags.

To be continued.

‍

‍