How to Get Video Platforms to Recommend Only What You Actually Want?

[msmissydc]

Is there a reliable way to 'guide' recommendation algorithms—especially on video platforms—so they only show content related to exactly what I want, without drifting into loosely related topics that don’t interest me? In other words, how can you influence algorithms to focus tightly on one specific subject, even if other viewers commonly see broader or adjacent content as relevant?

I have two situations where this specifically would be relevant one I don´t want to use as a prime example, but its technically the more relevant one. The other is the following.

I am doing larp specific DIY, but I am neither interested in structure building woodwork or larp Armor DIY. Looking for the first one always gets my feed flooded with the later two.

I know some platforms allow you to ban hashtags and this helps, but I think it often also bans the content I want to see.

Updates

1 y

This post has nothing to do with politics and I still get Maga Idiots spreading their fascism. WTF! Just a headsup, nonsense irrelevance and spamming are reasons for bans so dont play the victim card and cry censoring. Its a rule from the side, not my decision.

[Maybe_Maybe_not]

It's very specific, I'm afraid I don't run into this issue, for both IG and YT which happen to deliver precisely what I want. I can only tell that these algorithms feed on

• amount of time you spent "staring" at content
• where you click and tap
• content type, categorized, you're looking at

It's a difficult case because you don't master underlying categories in any database, it's done on server side. So if what you want to see is too specific you won't get it easily because everything, database side, is tied together on similarities not exactness

[msmissydc]

Yes and the last part is the problem. Some categories are on the one hand specifically what I want but the same category is also not what I want.
Imagine a ven diagram but the circle of the stuff I want is nearly completely overlapping with the stuff I don't want but the later is much bigger

[Maybe_Maybe_not]

yep, I see the point, that's an inherent flaw of all algorithms I assume, they can only be as precise as database they're using, especially complicated when users auto-tag content more or less accurately

[Anonymous]

I suppose there are some options on Youtube, Vimeo and Dailymotion.

[Opinion Owner]

Thanks for the MHO

[wynning]

Maybe_Maybe_not is spot on. There's also collaborative filtering going on as you have noticed where you get recommended stuff other people pe "like you" watch as well.

The DSA lecturers would describe this as exploitation vs exploration

Recommendation algorithms are designed to balance exploiting your known interests with exploring new content that you might find engaging. This strategy aims to maintain user engagement by combining familiar content with novel suggestions, enhancing the overall user experience.

[wynning]

That's a direct quote but essentially I doubt you can fully get rid of avoiding reccommended content you don't want because it's not made to be that way...

[Nikki1989]

YouTube Music or YouTube Premium gets it right most of the time.

[msmissydc]

I rarely have that problem with music. Generally speaking, algorithms work as intended, but sometimes they don't. If only 10% of my recommendations or less are what I want in one area then that's bad.

[AI Opinion]

To refine video platform recommendations, start by interacting consistently with content you like. Engage by liking, commenting, and watching videos fully. Use search terms that are highly specific to your interests and avoid clicking on loosely related topics. Adjust your watch history by removing irrelevant videos. Some platforms allow blocking or flagging unwanted content, which can help. Keep refining by continuously providing feedback, as algorithms learn from your interactions. 😊

[CupcakeTheDestroyer]

Yeah, you have to watch only what you want and dismiss everything else.

[Sofie4]

Use YouTube they have how to videos for everything.

[NicholasRedone]

Feedback