I am willing to hear differing opinions on this.
I sometimes see people on Fediverse speak as if there is something inherently wrong about the idea of content sorting and filtering algorithms.
There is a massive amount of content today and limited time. Content algorithms could provide the benefit of helping us sort content based on what we want. The most he urgent news, the most informative articles, the closest friends, etc. This might have some similarities with how Facebook and others do it, but it is not the same. Big social media algorithms have one goal: maximizing their profit. One metric for that is maximizing screen on-time and scrolling.
Personally, I’ve been developing an algorithm to help me sift through the content I get on my RSS reader, as there’s a lot of content I’m uninterested in. This algorithm would save me time, whereas those of Twitter and Facebook maximize my wasted time.
In my opinion, algorithms should be:
- opt-in: off my default, and the user is given a clear choice to change it
- transparent: the algorithm should be transparent about its goals and inner workings
Only with this, can algorithms be good.
What are your thoughts?
Not really my opinion, but there is a reasonable argument to be made about even benevolent algorithms ultimately increasing your engagement with online content and alienate you from your physical surrounding and people near you. Just because you set it up yourself does not mean that it healthy for you.
For me, it was actually the implementation of the forced timeline algorithm that was the breakthrough for me to finally leave Facebook after a long frustration. That was more than a decade ago. Now these algorithms became standard.
I would like to see multiple timelines, so to speak - for example, the favorite lists on Twitter would be available directly on the home page. That’s where I would actually find it useful in parts: My list for comics, for instance - sorting them by most popular posts from the last week? Why not, would be useful. Inoreader also has such a feature for premium users. For the home timeline on social networks, on the other hand, I really think it’s pure poison. On Facebook, it meant that I no longer saw posts about events in my private, healthy circle of friends - and instead anything that generated controversy was flushed to the top.
As long as they’re transparent and under user control then timeline algorithms might be ok. However, it would start to become problematic if instance admins could control the timelines of users, and it might become tempting for them to do so for monetization reasons.
Even under user control there would be a temptation for some people to try to SEO against the known algorithms, so that their posts appear preferentially in some people’s timelines, leading to the same set of problems that BigTech has.
i think when it comes to algorithms that save you time, simple filters do the job perfectly. like only people you follow vs. specific hashtags, or just full posts vs. replies included, or chronological vs. “good friends” (like in instagram) first. part of the reason modern algorithms are so complex is so they can confuse us and we end up spending more time on the platform. if you’re making an algorithm for ease of use, it should be the opposite of confusing. it should probably be clarified what people mean by algorithms, since that’s a very general word, but most of the time they probably mean the complex and confusing stuff modern social media uses, rather than the simple filters that most of the fediverse uses.
more complex algorithms might be useful for a site like YouTube, since it’s an entertainment platform not a socialization platform, so you just want to see anything that will be entertaining, and discover new content whenever possible.
@cyclohexane
> What is wrong is using them for maximizing corporate profitsIt is arguably not wrong either for a company to design an algorithm in a way that serves their interest. What is definitely wrong is that the algorithms are imposed on the user.
If I want to continue exchanging with my friends that are on the platform, I need to be submitted to the algorithm.Yep. Lemmy feed for example is algorithmic
Not really unless you have a really broad definition of algorithmic. It is just up and downvotes (and personalized subscriptions).
It uses an algorithm to determine score for sorting. What is your definition of algorithm then?
The inherently wrong thing about algorithms is that since they are made to maximize profits by maximizing engagement, they produce addiction. I don’t see anything wrong with an opt-in option for algorithms as long as there is a warning about the fact that it may produce addiction. The problem is that for an algorithm to work more data needs to be gathered. And since many FOSS software is privacy focused they don’t tend to gather much data.
Addiction is a goal of the algorithms as they are designed. Better algorithms increase your screen-on time and mindless scrolling.
A good algorithm could, hypothetically, limit the number of content you see, or save you time from scrolling to find something interesting.
I agree. As you said, an algorithm doesn’t necessarily need to be addictive or maximize your wasted time, that just happens to be the most common type we see today, due to the big platforms’ drive for financial growth. It is perfectly reasonable to want an algorithm that helps you by filtering the massive pool of content available on the Internet into the most relevant stuff for you, while also be respectful of your time. The idea you mentioned of an entry-limited algorithm is certainly very interesting, in my opinion.
My big problem with “algorithms” (by which I don’t mean the pedantic “well, pushing top-rated content to the top is ackshyouallee an algorithm, technically”) for controlling feeds is that algorithms are biased in subtly devastating ways. We like to think that “algorithms” are neutral because computers are neutral, but the truth is “algorithms” are designed and implemented by human beings and reflect what those human beings think is “normal” or “correct” or “important” or the like. Indeed there’s one huge, GLARING bias baked straight in from the outset: numericity. If it can’t be factored in some way into a number, it isn’t important to the “algorithm” because at its core the “algorithm” can’t work without numbers of some form or another.
Every “algorithmically”-curated system I deal with I can break with ease just by thinking a wee bit outside the box and flustering the poor so-called AI by selecting things on criterion that they’re likely not programmed for because in the biased view of the AI’s programmers the criterion wasn’t “important”.
At some point years ago Facebook started defaulting to relatives/family algorithmically. This is extremely biased and problematic. It makes a lot of sensitive assumptions, as everyone’s family structure is different. So “devastating” is a good choice of words.
