It is stupid. The only way you can make someone an expert of a new tool in 15 seconds is by making the tool so simple that it's completely useless. There's a reason you get a few dozen hours of training before you are allowed to drive a car around. Or fly a plane. If you want to have a simplified car UI, then one exists too - it's called a taxi.
The reason of course is the competitive market pressure. The primary interest of a company is to sell the product, not for the product to be useful. And so most of current design effort is about making software easier to sell, not about making it actually useful.
It's absurd to expect people to figure out by themselves how to use their file systems, databases or - yes - even their paper calendars. All of those are tools that require some learning to use effectively. We need to make people read manuals again, and stop expecting everything to be usable out of the box by an untrained monkey.
If you've ever read a technical manual, you're probably in the 1% of the population. People don't even read two-lines popups, and you expect them to read complicated manuals?
What you really need is adaptive interfaces that grow in complexity as requirements expand, reliably detecting the newbie / power user / hacker progression. That's what works best. Hackers ridicule UIs that bury stuff under "Advanced" modes, but they're actually the best approach.
Doing three (or four, or five) times the work, though, is no fun for developers, so we go from "you're a dumb user, be happy with limited options" to "you're a clever user, surely you can read a 1000-pages bible before launching a command", because building multiple interfaces for the same routines is boring.
What I would call number 1 of user experience: The human is never at fault. Systems are there FOR people and systems should be designed to cope with the flaws of humans. If your system can't handle normal users then you are at fault and you're just pointing fingers.
> And likewise the problem with tagged data is likewise the people who can't organize tags
Extremely many attributes (which IMO is a similar but much better model than tags) can be set automatically. Date and location (and people in them by using face recognition) for photos, artist, track title for music files. It's almost comical that document names almost never match up with their filenames. File endings are another extremely silly concept.
> The classic analogy is you unleash genealogists on a village cemetery and have them produce a family tree and the output is actually useful information. Unleash the same team and ask them to digitally tag tombstones, and you just get a pile of useless, error filled data, packed with duplicates and weird capitalization issues.
The attribute scheme seems to work out fine for Wikidata. Filesystems today also have adopted attributes but are fairly limited in using them. On OS X Photos and iTunes organize files through databases that connect files with attributes. On Linux when using Gnome you have Zeitgeist that automatically indexes all your files and keeps attributes on them, which is used for instance by the Gnome Music application.
> You "could" write a noob gui for a nuclear reactor, or a UI for even newborn babies to use matches, but you probably shouldn't.
The complete opposite: You actually need to put very much consideration into the user interfaces of critical systems to prevent mishandling. Chernobyl was partly caused by a bad UI.
I beg to differ. Chernobyl was caused by flagrantly violating the simple principle of "if it ain't broke, don't fix it". It wasn't broke (well, not any more than any reactor of that type is broken, by design) and they tried to fix it.[1]
[1] https://en.wikipedia.org/wiki/Chernobyl_disaster#Steam_turbi...
The classic analogy is you unleash genealogists on a village cemetery and have them produce a family tree and the output is actually useful information. Unleash the same team and ask them to digitally tag tombstones, and you just get a pile of useless, error filled data, packed with duplicates and weird capitalization issues. This is a classic "information vs data" misunderstanding.
Another problem with hierarchical vs tagged is we have somewhat obtuse yet usable set of tools to handle truly immense spammy hierarchical filesystems, but strategies for handling and scaling tagged filesystems usually are hand waved away with "it'll never happen" or "we'll sprinkle magic AI dust on it, then it'll evaporate away". A classic "make the simple easier, while making the more difficult impossible". No thanks.
Much as there's nothing really new under the sun in IT, a generation or two ago this was "sure sendmail configuration is complicated, but all it needs is a GUI because in the small subset of problem domains that are a match for GUIs and people with a problem simple enough to be handled in a GUI, it works well, therefore its perfect for everyone to do everything in"
Also data typing is just a subset of tagging. Or rephrased a type is a very poorly implemented limited single tag.
Ditto the above with a "decent shell language" again we're dealing with people who can't handle paper calendars without a book telling them how (GTD, 42 folders, etc). They're just going to mess things up quicker with automation, while people who can handle the cognitive challenge will likewise jump into org mode or crontab without any serious effort on their part... You "could" write a noob gui for a nuclear reactor, or a UI for even newborn babies to use matches, but you probably shouldn't.
As a meta observation this is a classic "future is already here just unevenly distributed" problem, I use databases and git all the time. However see the comment about the GUI for sendmail... Much like some people are unqualified to be handed an acetylene cutting torch or a beryllium-Pu neutron source, some people should not be handed relational DBs or KV stores or version control, and if they had the background to understand and safely use those tools, the existing tools aren't really all that hard to use for folks of adequate cognitive level.