A social media profile posts 30x daily on average.

You find this:

(One of a series of polls. Boosts appreciated.)

Follow

Which of these thresholds would be too high for typical daily posts from a social media account?

(One of a series, boosts appreciated.)

At what point would infrequent posting prompt you to drop or unfollow a profile, if any? Select the shortest applicable time period.

Posting only once per ...

(One of a series, boosts appreciated.)

For those curious as to the values selected here, it's based of the natural exponent, en.

In trying to find human-scale-but-logarithmic scales I've found that this produces ... vaguely satisfying results.

Powers of 2 are too condensed, powers of 10 have jumps that are too big.

@dredmorbius i've never heard of someone who only makes annual posts but i'll generally consider unfollowing an account that has had no activity for over a year

@dredmorbius
I'm more likely to unfollow someone for posting too often, than for posting irregularly.
If they rarely post, but still post interesting content when they do, then I see no reason to drop them, even if they just post once a year.

I might instead ask if they are more active elsewhere instead.

@FiXato @dredmorbius For me there's practically no limit. For me, it's not an individual limit but an overall limit of my feed. And with multi-column userstyles, I'm comfortable with the volume in my feed.

That said, I only add folks whose content I find interesting enough. So far, it turns out that this results in a total volume which I am comfortable with. (Note - I do NOT try and keep up with everything.)

@dredmorbius
This one really does depend on medium. Long monthly blog articles are great on ongoing projects I follow and about what I expect.

But 1 "tweet" or "toot" a month is very slow. OTOH, I tend not to notice infrequent posting.

@dredmorbius what should prompt me to unfollow someone who isn't actively bothering me? Following many low-activity accounts isn't a liability, overly active ones is!

@isagalaev I'm asking, not telling 😺

This is mostly an attempt to assess sentiments / feelings / practices.

@dredmorbius i see that my responses are in the minority but i find that the more posts i'm making the more fun i'm having at any point. similar that having more posts to read from other people is better (to the limit that the posts don't come faster than i can read them and it's not like there aren't natural lulls where i could catch up anyway)

@dredmorbius There is no set timeframe. In case of Fedi I just forget about themwhen they don't post, and in case of say rss, it took a while for me to finally pull the plug on some stuff.

@dredmorbius If they don't post I'm not sure I'd remember to unfollow them.

I'd followed a friends profile that got abandoned, and the only reason I did anything is that it started posting suspicious links, so I reported the account for having been probably taken over.

@dredmorbius I only do this rarely if I've never interacted with them and they appear to be a inactive account. I followed a bunch like this early on on Twitter and hadn't realized it.

@dredmorbius nothing. Just wanted to see if that worked better for you.

@londubh Not sure if it came through but what I posted was e^n --- e raised to the n.

I'm used to Markdown notation working from my side, it may not render elsewhere.

@RefurioAnachro Technically correct. It gives nine Fibonacci numbers, but deviates afterwards, if my awk code is valid.

The series are similar however.

Columns are iteration, e^(i/2), and fib(i), adjusted to start at 1 rather than 0.

gawk 'function ceil(valor)
{
return (valor == int(valor)) ? valor : int(valor)+1
}
BEGIN {
n[0]=1; n[1]=1;
for(i=0;i<20;i++) {
printf("%4i %12i %12i\n", i+1, ceil(exp(i/2)), n[1])
n[3]=n[1]+n[0]; n[0]=n[1]; n[1]=n[3]
}
}'
1 1 1
2 2 2
3 3 3
4 5 5
5 8 8
6 13 13
7 21 21
8 34 34
9 55 55
10 91 89
11 149 144
12 245 233
13 404 377
14 666 610
15 1097 987
16 1809 1597
17 2981 2584
18 4915 4181
19 8104 6765
20 13360 10946

Har har, very good! The correct formula for Fibonacci numbers is of course based on the golden ratio φ:

F_n = round(φ^n/√5)

The reason this works is that we have

lim F_n/F_(n+1) = 1/φ = 0.6180339...
n→oo

If you insert the innards of the rounding formula, you get an idea by how much you need to divide to get roughly to the next Fibonacci number.

And my false analogue is close enough:

exp(n)/exp(n+½) = 1/exp(½) = 0.6065306...

Of course, you could also look at the reciprocal and multiply instead of divide, but that makes it look a bit less of a good match. Well, it's not too close anyways, so after only 8 steps it breaks. Using ceil() gives a little more slack, rounding doesn't work quite as nicely.

So, your awk is most certainly correct, I hope you can pardon my cheekiness :)

@dredmorbius

@RefurioAnachro No apologies needed. I had a hunch you were needling a little, and it was an opportunity to actually code up an example to see.

My maths algos coding fu is ... weak and needs practice. This offered that.

I'm glad you enjoyed that. Actually, your numbers were what made me come up with it, so, thanks for that! I certainly had fun staring at the effect and I couldn't resist sharing the itch :)

@dredmorbius

@RefurioAnachro My remaining question is why these numbers seem intrinsically "right" in addressing information content / group sizes, both of which I have in mind.

I'd previously been looking at powers of 2 or 10 and finding them, generally, either too- fine-grained or too broad.

For network scales, particularly, there are "interesting" cases at 0, 1, 2, 3, 5 (second value at which edges exceed vertices, 8 (about the maximum group size suggested in the Mythical Man Month), and then more roughly the increasing values which seem to correlate at least roughly to pronounced tipping points / scale points at which group behaviours seem to change markedly.

Mind that that's just eyeballed / vague intuition, but it makes me wonder if there isn't some mathematical or causal relation here.

That's a good question! I have no idea! There is lots of literature on Fibonacci numbers in nature, and as far as I can recall more often than not these ideas relate to growth processes.

More generally, my intuition completely agrees that exponential growth is the correct way to see many things. I recently learned that there is such a thing as exponential interpolation:

Linear interpolation works like this:

lerp(t,a,b) = (1-t)a + tb

Exponential looks like this:

eerp(t,a,b) = a^(1-t)·b^t

I learned about this just recently from Freya Holmér here:

twitter.com/freyaholmer/status

@dredmorbius

@dredmorbius i voted 21 but now that i think about it i and a couple mutuals on tumblr each probably do a lot more than that on any given day

@carcinopithecus My questions don't make the distinction sharply, but there's typical versus occasional behaviour.

I've looked at my own content broken out by year, and have a really good idea of what my posting frequency is averaged over 12 months. I might kick out several times my long-term average if there's something specific going on, but I tend to settle into a reasonably stable rhythm, it seems.

So: long-term mean is really what we're looking at here.

Since the answers are self-reported, I'm taking into account that people might misjudge frequencies. That's one of numerous limitations to the poll.

long reply about timeline curation based on user activity 

@dredmorbius
it's hard for me to put an amount of it, though it's probably between 8 and 21? I voted for 21, but it very much depends on the kind of content they post, and how long their messages are, and how much of it is links (the more, the more likely I am to boot them), or boosts.
I'm quite likely to disable boosts for people who have a high amount of boosts, especially if they boost more than post their own stuff. If disabling boosts is no option, then a mute or unfollow is likely.

Basically I have an x amount of spoons to use/give, and if someone's posts are eating up a disproportionate amount for being too long consistently, or I have to spend too many sorting out interesting content from unappealing, then I'll have to curate them out of my timeline somehow.

While algorithms are flawed, I did like having the option of of having weighted timelines where my main timeline was a mix of all posts from some people, most posts from others, and just a couple from 'high traffic' ones, and having additional timelines for if I have more time to read additional content.

re: long reply about timeline curation based on user activity 

@FiXato What prompted the poll(s) was going through my own Diaspora* archives and seeing posting patterns from others.

In particular I'm finding memes, reshares from other social media (especially where those then link to some primary source, which itself might not actually be primary...), bare-naked links, images, etc., ... tend to leave me cold. Posts with some meat to them, and which generate a substantial conversation ... more my thing. Even humorous threads --- there are a few Brits whose political musings tend to veer sharply into the absurd, and yes, I participate and aid in that absurdity.

(US politics takes itself too seriously?)

And people posting 20, 30, 40, 50 times a day. I'd prefer that be collapsed somehow. They're certainly not getting engagement on those posts.

I just counted the number of toots in my "A" list (high-interest) here on Mastodon, and it's about 80. That's from a fairly substantial group, though quite a few are either inactive or very rarely active. The activity is ... a bit high, and I might trim or reorganise the list a bit, but it's within reason, and the quality is mostly good.

The spoons thing is a good one, and I've been collecting references for a few years on how much media and information people access on a regular basis (per hour/day/week, etc.), from words heard or spoken to amounts read. Time-based surveys seem to be the core concept, as attention is based on time, and there's a rigidly fixed quantity in any day. A minute spent on A is not spent on B, and vice versa.

re: long reply about timeline curation based on user activity 

@FiXato The abilty in G+ to structure the Home stream, or to switch rapidly between other streams also helped a lot.

On Mastodon I typically have a few lists visible at any one time.

On Diaspora*, switching between Aspects is slow, and any given stream only loads 15 posts, with a pretty substantial delay before adding more. That's a lot of friction in my view...

Lack of search also means that tools for increasing S/N and relevance are effectively nonexistent. Find good people or ... well, you lose.

@dredmorbius on twitter I used followerwonk to segment high volume posters into a separate lists so that the low volume ones didn't have so much competition. It helped a lot.

400+ a day is a bit much for most people to do manually, so those are typically not accounts follow anyway.

Sign in to participate in the conversation
Toot.Cat

On the internet, everyone knows you're a cat — and that's totally okay.