Posts on a Mastodon timeline are sorted in chronological order, new to old. I think algorithmic curation (e.g., highlighting popular posts) is acceptable as long as it's open source and testable.

#Fediview, which uses an open source algorithm, can selectively display popular posts from your Mastodon timeline based on posts' boosts and favorites. https://fediview.com/

Kudos to the developer, @adamghill

#OpenSource #Algorithm #Curation

fediview

@adamghill Thank you for your expert comments! Your project is separate from Mastodon, and thus it is opt-in. I think it's good to have options for users, some of which may differ from what the designers originally thought of.

I understand the potential consequences of algorithmic curation. But limiting (?) a useful tool seems a bit paternalistic... I'd rather give better ideas for improvement. Hmm, I need some time to think about this.

@Sho_Ishiguro thank you for this pointer.
Since the comments by @adamghill you are thanking for are not a reply in this toot/thread (as far as I can see), it would help to provide a link.

@KarlE Thank you for mentioning it. I intended to thank for several comments of him, not specific one. In this thread, so many topics, issues, and recommendations have been discussed! I wanted to explain my general view.

For background info, the FAQ page of #Fediview is well-written and succinct. It may answer questions that you might have. https://fediview.com/faq

fediview

@Sho_Ishiguro Thanks for the shoutout! Tons of credit goes to @MattHodges for the original script that I forked. 🤩
@adamghill Thank YOU! I read the FAQ page and really like your ideas 👍

@Sho_Ishiguro @adamghill

Provided we know what is presented to us and provided we have the choice.

@Sho_Ishiguro @adamghill @tfardet

when you say “testable” what kinds of possible tests are you envisioning?

@UlrikeHahn I mean, it is possible to test if an algorithm is programmed and works as the developer says. We can see the scripts and run them in a local environment (e.g., in a terminal of my laptop). We can compare these results with the results from #Fediview.

@Sho_Ishiguro isn’t the main issue what the algorithm does to the behaviour of the overall system, though?

If M. has been designed to dampen virality, it sounds like the algorithm would do the opposite of that?

@UlrikeHahn @Sho_Ishiguro Note that this is a client, not a proposed change to Mastodon itself. If lots of people favorite toots in Fediview, it will increase the virality of those toots in Fediview.

The only way this would thwart Mastodon's anti-virality design would be if people BOOSTED toots that were viral in Fediview, thus causing people following them and not using Fediview to see those toots.

@packy @Sho_Ishiguro

thanks for clarifying - yes, that is what I was assuming on both counts (nature of implementation, and likely behavioural outcome).

@packy @UlrikeHahn @Sho_Ishiguro When we come to Mastodon from Twitter, we're told that we need to boost more since likes alone won't make things seen. But with Fediview, likes alone *would* make things seen, so people using it won't need to boost and then the virus won't spread … hopefully.
@Sho_Ishiguro @adamghill Please don’t do this. Came here precisely because nobody has tried to anticipate what I will like and because there is no profiteering behind the scenes.
@xankarn @Sho_Ishiguro @adamghill why in the world would anyone want an algorithm to decide what to show them? I curate my feed how I want it based on who I follow. I want to see all of it
@westphall @xankarn @Sho_Ishiguro I have found it useful to surface posts that I missed previously. But, if that isn't useful for you, then feel free to ignore it. 😎

@xankarn @Sho_Ishiguro You have to create an API key for this to work, so you would have to explicitly opt-in to see your personalized summary.

The algorithm doesn't "anticipate" anything. It is literally counting favorites, boosts, and replies for posts that are already in your timelines. And it's opensource so it can be audited -- you can literally see everything it does.

But, at the end of the day... no one is forcing you to use it, so... you can just not use it?

@adamghill @Sho_Ishiguro Thanks for that explanation. Still have misgivings about others deciding what should be counted and the hierarchies that these decisions create.

@xankarn @Sho_Ishiguro Understood.

Personally, I don't love the feeling it's a "popularity contest" so I'd love to hear of some other approaches.

@adamghill @Sho_Ishiguro If the principal concern is “I don’t want to miss something good,” I’d say in response, popular is not necessarily good. I’m sure you’ve had the experience of getting tons of “likes” for posting something winsome, but then having something really creative/thoughtful passed over without much feedback. I prefer a platform/discourse that keeps these on level footing.
@adamghill @Sho_Ishiguro Also, we know that anger, outrage, and divisiveness attract attention, clicks, quotes etc. Things accumulate algorithmic value for reasons I think we should resist.
@adamghill @Sho_Ishiguro If algorithmic manipulation appears here, I’ll seek to opt out, but I hope it doesn’t for the sake of cultivating an alternative space for a different kind of discourse than what venture capitalism has already given us.
@Sho_Ishiguro @adamghill the problem is inauthentic networks juke #algorithms. Like absolutely. That actually dominated FB for years, dominating attention on Twitter for years, boosted crap on YouTube for years. Click bait scam and info ops networks easily juke algorithms. Kind of how our political system blew-up, icymi.
@LaureM @Sho_Ishiguro The default view uses your home timeline, so it would only surface people you have already followed. Although, I agree that using this on the federated view is more problematic.
@Sho_Ishiguro @adamghill I don’t think popularity is enough or is the best way. The client needs to be able to skip already read toots to show me unread and auto load instead of having me click. I scroll past that gap all the time.

@michaelhk @Sho_Ishiguro

>I don’t think popularity is enough or is the best way.

I tend to agree and I'd love to hear actual ideas. Also the code is opensource and PRs are welcome. 😉

@Sho_Ishiguro @adamghill

How does this work in a different way from the "# Explore" timeline? When I click on that (in the default web interface) it says that it's showing popular posts.

@RichPuchalsky @Sho_Ishiguro Honestly, I've never looked at the code for how the normal Explore feature works, but fediview is very configurable so you have control over which timeline to use, how far back in time to look, how to weight different factors, etc.
@Sho_Ishiguro @adamghill I'm fine with algorithms as long as they are opt-in only.

@Sho_Ishiguro @adamghill

And optional. You forgot optional. Open source. Testable. And opt-in.

@Sho_Ishiguro @adamghill Then you just have everyone designing their posts to go viral and fewer genuine interactions in my opinion.
@Sho_Ishiguro @adamghill I think you missed "so long as it's opt-in ..." 🙄 🤦‍♂️
@Sho_Ishiguro this is a great compromise; I was musing on this topic a few weeks ago, and was coming around to the idea that algorithmic timeline is a necessary evil to minimise abuse, but it needs to be trusted (to avoid accusations of bias). Ergo, it must be open source.
@Sho_Ishiguro
Oh! That's my posts being seen by nobody then......😔