The 90-9-1 principle: 10% of HN users decide on what the other 90% see

20% of HN posts are unambiguous and don't need any clarifying comments, but 80% of the time comments are helpful: see Pareto distribution https://en.wikipedia.org/wiki/Pareto_distribution. It's fascinating too if not new.

That 90% could spend more time upvoting articles in New

I wander over there periodically to see if there's interesting stuff. I upvote it when I either think its interesting or think it will generate interesting discussion

Isn’t this the definition of an editorial board?

The 90-9-1 or 1% principle is well established in virtually any communications network, it's a very rough rule of thumb, and arises as do numerous other phenomena out of power law and Zipf function relations. Jacob Nielsen discussed the concept in 2006, and I recommend it for more background. There's also a good Wikipedia article. Andrew Odlyzko uncharacteristically seems not to have addressed this directly in his own (otherwise excellent) article archive, though I'm sure it's mentioned in passing at least once.

https://www.nngroup.com/articles/participation-inequality/

https://en.wikipedia.org/wiki/1%25_rule_(Internet_culture)

http://www.dtc.umn.edu/~odlyzko/doc/networks.html

The "1%" may be far smaller than the rule would lead one to believe. In the case of Google+, it was roughly 0.16% of all profiles who were active within the previous month. On the JoindiasporaCom pod of the Diaspora* network, of some 316,658 registered profiles, 546, or 0.17%. The fact that both networks show similar numbers ... may also suggest a long-term tendency of social media sites. I participated actively in both.

https://blogs.perficient.com/2015/04/14/real-numbers-for-the...

https://web.archive.org/web/20220112195336/https://joindiasp...

The implications of the 90-9-1 rule are several, and the Tweet thread here only touches on a few elements.

- Yes, a small number of participants can unduly influence what gets shown, though nudging that is a very imprecise practice at best. The "1%" itself however will reflect multiple divergent interests. The Hivemind is vast and contains multitudes. It contradicts itself.

- In reverse-engineering possible usage metrics, the 90-9-1 rule offers at least some basis for making estimates and drawing inferences, though these may be off by an order or two of magnitude.

- With larger networks, the "1%" tends to become an ever-smaller fraction.

- With older networks, participation tends strongly toward attrition.

- Activity metrics such as MUA are far more relevant than total registrations.

- The impressions drawn from any voluntary-participation exchange are all but certain to be entirely divorced from those determined by systematic random sampling. At best you might be able to state how the most active participants of a group will respond. If that corresponds to other power alignments, it may be useful. Quite often it is not. A tiny minority can make a deafening noise.

(The above point was a late add to this comment. I'd meant to include it initially.)

- Total users tells you very little about prevalence or impressions --- how often a given item is seen on a site.

Counts of things --- profiles, posts, comments, and the like --- are easy to come by. Counts of actions, and especially of who saw what and for how long are much harder to come by, but far more significant measures. Recall that each individual person's time is non-fungible. We all have 24 hours, 1,440 minutes, and 86,400 seconds, in a day, and those are allocated to a set number of tasks. There's a pretty fascinating field of research in time-use studies, how many messages people are presented with, how many words are spoken, books or articles read, etc.,, within a day, week, month, year, decade, or lifetime. And the totals are often far smaller than you might think. Our attention and personal data access and processing bandwidth are very finite.

I've just downloaded ten years worth of posts, some 2,700 items, from a social media site that's going offline. The compressed text amounts to just over 18 MB. That's ten years of content, in just under 13 floppy disk's worth of storage (1.4 MB is a truly puny quantity today). Put another way, a 1 TB drive could hold either roughly 60,000 people's data for a comparable period of time, or my own musings for well over a half-million years. (I'll see how much of that I can deliver.)

Ten years' writing at a comparable pace from all the world's population would occupy 130 PiB of storage, which would cost roughly $2.5 million. Well within the budget of a modest commercial entity, let alone a state actor.

The real services of social media aren't content but in providing audience, discovery, visibility, and interaction. That's the principle benefit I find from HN, better on some days than others.

And finally, "ranking" within the most-active queue has very little to say about quality or value. In yesterday's discussion of Donald Knuth's work, drfuchs made a number of highly insightful comments. That's David R. Fuchs, best I can make out. (https://dl.acm.org/doi/10.1145/2363.2367)

Their HN presence is modest. That has virtually no bearing on their history and contributions to the field, however. Many easy-to-acquire metrics provide little if any useful guide to overall quality, fitness, or value.

Quantity is not quality.

(Though it has a quality all its own, has has been noted.)

Edit: see late add above.

Because people rarely can contribute in equal amounts, unless it's something simple, like walking.