Notices by deutrino (deutrino@mstdn.io), page 27

@Gargron Week old idea re: follow suggestions, but maybe the "don't include me in search results" privacy button can do multiple duty in adding NOINDEX and excluding the user from all discovery features. IMO the single switch might be a better solution than a whole bunch of switches, though my supposition is that most people will have them either all turned on, or all turned off.

Cool #remote #FOSS #jobs alert!

https://www.gnome.org/foundation/careers/

#job #work #libre

(*and by "implementations" I include the human aspect, not just the software.. bleh, I think you all know what I mean)

I really hope #ForkOffTogether succeeds even though I'm not super likely to use what they create.

I just wanted to say that as somebody who vastly prefers "open speech" instances. I think there's absolutely room for more implementations in the fediverse, and I hope the #ForkOff folks will federate widely.

We are all building something totally new and amazing together, even though we have differing ideas as to what's best for our particular portion of it.

#Fediverse

@Gargron

I think this use of hashtags for onboarding removes most of the harassment concerns raised in prior hashtag feature arguments, particularly if the "don't index my toots" button in privacy settings also opts those users out of being promoted to noobs here.

Further, especially with a little algorithmic randomness, it lacks the thundering herd / concentrating eyeballs problems.

And, it promotes the use of hashtags. If we're not to have full-text search, that's good 👌🏻

@alcinnz There is certainly something to be said for human curated lists. I like to listen to certain internet radio stations because the DJs and music directors may not perfectly align with my taste, but they almost always produce a more interesting mix than an algorithm.
@Gargron

@Gargron

To be clear, I'm suggesting an algo that would surface hashtags like "linux" and "cats" and "photography" to present to the user, not "kde" and "blepvideos" and "nikon" - the key would be very broad interest hashtags.

The "algo" here could also be another human-curated list, it'd just need to be well tuned to be broad, and probably based on a study of hashtag use on the fediverse, and updated from time to time.

@Gargron

Idea: aggregate the most general, longest-lived hashtags above a certain usage cutoff, present them to the noob in a "do any of these interest you?" onboarding multiselect. Then:

* during onboarding, suggest one of the more active users in some of the selected hashtags as people for the noob to follow

* until the noob crosses some non-noob threshold, occasionally put recent toots with the selected hashtags into their timeline

@Gargron

IMO in this scheme the interest keywords should be restricted to strongly interest-based instances, not general instances with a lean towards a few interests. This feature could in fact be user-hostile otherwise. "I thought I was signing up for an anime instance but the local timeline is 90% open source wanking" would be an obvious failure mode.

Alllll of this points to human curation of the instance chooser list, even with biases & potential drama over such.

@Gargron

Idea: assume a curated instance list in the sign-up instance chooser. In the database, interest-based instances have a curated list of keywords. In instance selection at sign-up:

1. Do more fundamental filtering (eg, CoC)

2. aggregate interest keywords from this set

3. offer multiselect of keywords

4. highlight matching instances in the chooser.

If a user says "I can join a server for bookworms (or etc), NICE!!" then they're more likely to stick around.

@Gargron

I go back to the "cool people with similar interests" sticky factor. How can we make finding those people easy for noobs?

Principle: can we get the user to specify some interests during onboarding, then make it more likely that content from users with similar interests will be displayed to them at first? (cont)

@Gargron

(cont) I don't agree with the seemingly reactionary stance against nearly every discoverability feature idea that's tossed around, nor do I agree that algorithms are never useful in this context and that all discovery must be 100.0% organic. IMO that's just not realistic. Mastodon is a firehose, and people have limited time.

What about a happy medium with unobtrusive algorithmic onboarding & suggestions which tail off over time as a new user's engagement grows? (cont)