Email or username:

Password:

Forgot your password?
Top-level
Geert Aarts

@rauschma Thanks for all your contributions. What would be really useful is to have a very basic algorithm that places the posts with the most boost by the ones *you* follow at the top, and to make it one of the default options, particularly for the new users. Not having any algorithm whatsoever is probably the reason many of my colleagues became completely inactive.

13 comments
Axel Rauschmayer

@geertaarts Interesting!

I’ve also seen comments from people who like that not having an algorithm (that mostly rewards drama) makes Mastodon more peaceful than, e.g., Facebook and Twitter.

What do these people miss most about algorithms?
– Finding the most relevant posts in their (busy) timeline.
– Discovering interesting posts elsewhere (while having an empty timeline)?

Well-designed algorithms could indeed be helpful.

@geertaarts Interesting!

I’ve also seen comments from people who like that not having an algorithm (that mostly rewards drama) makes Mastodon more peaceful than, e.g., Facebook and Twitter.

What do these people miss most about algorithms?
– Finding the most relevant posts in their (busy) timeline.
– Discovering interesting posts elsewhere (while having an empty timeline)?

Geert Aarts

@rauschma Yes, I can understand the discomfort many have with most existing algorithms on the for-profit social networks that try to pull you into the drama. But indeed, I think that an algorithm that just does some simple ordering based on your own social network will be a very different experience. Phanpy has that option, but it is a Beta feature, and quite hidden.

Demian

@geertaarts @rauschma for me, I don’t like the opaqueness of corporate feed/recommendation algorithms. And I really don’t like the idea of said corporations feeding me content based on their profit/political interests. I’m open to other algorithms (besides chronological) but would like better transparency and control

Stefan Elf

@dgodon @geertaarts @rauschma

I fear that "algorithms" would have a potential to eventually kill this space. I really hope that we can do without.

Axel Rauschmayer

@inwis @dgodon @geertaarts If we ever get support for algorithms they probably will be opt-in (switched off by default).

deutrino

@rauschma @geertaarts I've been thinking for 7 years or so that the answer is multiple pluggable algos which users can select a la carte, with maybe a couple suggested defaults.

unfortunately I have not (yet) been in a position to implement any such thing, for Mastodon or any other AP software.

Malfunct (he/him)

@rauschma @geertaarts I think any algorithm should be very transparent and selectable by the user not the host. This increases complexity I suppose, but it is the price of control.

Christian Kent

@rauschma @geertaarts I’m going to hype the idea of a trust score to replace algorithms

You can set a “vouch value” for as many or as few people as you like — 0 to 100, or -100 to 100 so we can block others

Then inherit a proportion of that trust number via followers. It’s the “friends of friends” idea that worked well once before.

Plus “friends of enemies” works even more — as any former BBS sysop can tell you, you get more activity overall when you reject more bad activity

But we need something to replace the algorithm. Not everybody can do full timeline, or inbox zero.

(You control consumption by sliding a filter from 100)

@rauschma @geertaarts I’m going to hype the idea of a trust score to replace algorithms

You can set a “vouch value” for as many or as few people as you like — 0 to 100, or -100 to 100 so we can block others

Then inherit a proportion of that trust number via followers. It’s the “friends of friends” idea that worked well once before.

delProfundo

@whophd @rauschma @geertaarts that trust tree gets spicey pretty quick. We had a friends of friends model in a startup I founded 15 years ago. At 4 steps away ur basically mapping the entire user base.

Jay Stephens

@whophd @rauschma @geertaarts
So, for FoFs (assuming -100 to +100) the effective score would be your friend's rating of that FoF, multiplied by the % you gave the linking friend?

Matt Mascarenhas

@whophd @rauschma @geertaarts Mmm, well I tend to be pretty hype-resistant but I'm into this idea!

I see it in the same ballpark as a bunch of ways of scoring references / citations I have floating around for a pipe dream comms system with sophisticated referencing at its core.

Janne Moren

@whophd @rauschma @geertaarts
People show very different faces to different people.

I'd worry that tech bros would have impeccable scores from a large number of peers, and the much fewer targets for their harassment would have no chance of breaking through.

It would be even worse if that all-over score decided their status on your feed, not just your own score, but I don't think that's what you're suggesting.

Look at algorithms for star rating systems for ideas on handling split scores.

@whophd @rauschma @geertaarts
People show very different faces to different people.

I'd worry that tech bros would have impeccable scores from a large number of peers, and the much fewer targets for their harassment would have no chance of breaking through.

It would be even worse if that all-over score decided their status on your feed, not just your own score, but I don't think that's what you're suggesting.

Go Up