Djoerd Hiemstra 🍉I love Mastodon and I research search engines, so when we set up a mastodon instance at the University of Twente, we planned to research federated search, who-to-follow recommendations, and trending topics. Students worked on this in one of our courses:
When engaging with the community, I quickly learnt that many tooters actively oppose search, recommendation and most other "algorithmic" tools. As a researcher in Information Retrieval I was kind of shocked: How could someone NOT like search??
Turns out, search can be easily misused to find persons to harass, searching for instance for "trans rights are human rights" and threaten all those persons is way to easy. Who-to-follow recommendations will then give the trolls more people to harass, as do trending topics. The lack of full-text search is a feature, not a bug or an omission.
Ton Zijlstra@djoerd I like search a lot but the location of where that search is possible matters. Full text search in social media disallows pseudo limited public conversations and creates damaging asymmetries. Unlike on a real public square where many conversations take place in a semipermeable setting and crashing one would immediately carry social consequences for the crasher, in social media that friction needs to be created. Search reduces friction in a way that favours the crashers.
@djoerd If a conversation wants to be findable using # is a way. Boosting is a more natural discovery tool mimicking how it would be on a proper public square. That said I do like search on anything I said , which I can.
@ton search on own posts and hash tags is fine.I like your town square analogy's social consequences analogy.
berkes 🐝 🚐 🏄 🌱@djoerd @ton leaves me wondering: if it is an artificial limitation,it would be easy to circumvent. It's not a limitation of the protocol, but of the client.
An instance could add search or a search tool. Then have their users (Ab)use that to bully, or simply inject themselves in conversations.
All that can be done, is defederate such instances when found out. And block users the misuse such tools.
Hence we should not rely on it for "security", which we currently do.
@djoerd @ton I'm pretty sure we'll see services soon that offer the following:
* See what people say about your brand (keyword)
* See what posts like to URL
* Find profiles with [advanced search conditions including emoji and flags]
* Reply to posts that mention your brand (keyword)
And we can't do much about it, other than try to hide from it, which is technically impossible.
We're relying on everyone being nice. Which is a very good attitude. Untill it isn't anymore.
@berkes @djoerd We generally always rely on people being nice, when crossing the street, when browsing the shelves in the supermarket. It's a pretty good assumption. Until it isn't. Where introducing higher costs for not being nice becomes necessary. Just browsed my earliest tweets, which were also nice and chatty, like here now, before it became a toxic I-need-to-win-this-conversation thing. I do get your point.
@berkes @djoerd Indeed. It's also why I think defederation is an effective tactic. Esp for smaller instances. I don't need to talk to everyone on the public square, just the _option_ of doing so. I do need the ability to refuse conversation with anyone, and esp to prevent them _joining_ my conversation nonetheless. Making it relatively harder to find my conversations outside my primary circle of interaction helps, granular control over blocking and defederation too.
@berkes @djoerd I just checked during a conversation yday: I've blocked 28 instances in the past 4 years (extremists, sh*tposters, bot-farms etc). About 25% of those in just the weeks since #twittermigration. Which supports your worry.
Dr David R Brake@berkes @ton You are right, this will happen and there are no technical solutions. (I am guilty of doing such analysis with tweers, I admit) Some instances thought about this, though, see:
https://mastodon.social/@socrates@scholar.social/109364972724279253
@djoerd @ton interesting discussion. What strikes me, is the same thing that worries me. We have to ask people to be nice. Rely on ethical researches.
That will get us far.
But we really ought to think about better protection against the inevitable actors that don't listen to our requests. That don't have to adhere to ethical principles. Just saying "we asked not to do this" isn't enough on the internet. B/C all it takes is a single bad actor.