2024/Berlin/timelines

Timelines was a session at IndieWebCamp Berlin 2024.

Participants

capjamesg
Sara Jakša
Kagami
Jakob
... add names

Notes

Welcome to the session!
Idea: See a timeline where you have an equal chance of seeing people, rather than seeing the most vocal people.
- Want to see people that post once a week equally to thost that post a lot.
- Considering timezones: make sure that see content posted at differe
https://fediscover.kagami.rs/

https://github.com/saschanaz/fediscover (sorry for using github)

- Kagami: Created a reader that lets you see posts posted since a certain number of days ago; see posts from the last week, today, a few hours ago, etc. If you choose "since 1 days ago", you see posts from the last day.
- Only shows one post from a single person, not more than one.
- Intention is to give an equal chance for any account to be displayed on the screen.
- How to choose each post from each user.
- Algorithm:

1. Pick random ten accounts 2. Read recent 80 posts from each account. 3. Pick one random post from each account from what was just read. 4. Result: Random ten posts for each button click. 5. If you click read more, it loads different accounts picked by random. 6. Applies deduplication. This algorithm solves less frequent posters having equal opportunity to be seen.

Idea: how would you order data to prefer posts from friends / in a category, etc.
Idea: look at pace rate someone is posting at. If someone who posts really rarely suddenly posts a lot, there is a brief period where you see a lot of their posts. If they keep up with posting more, the presence rate could decline.
Some people make lists to curate different content; some people on Mastodon make lists depending on the frequency with which different people posts. But this is something that has to be manually updated.
Not having lots of cognitive load is important.
Question: what do we want out of our feeds? There are trade-offs from all algorithms.
No matter what algorithm you use, there are always things you are not going to see.
Writing your own algorithm gives you the ability to explain how things work.

Please note that all contributions to this pad and other IndieWebCamp documents are considered to be released under the public domain according to CC0.