Re: [Usability] user levels, etc.

[Adam Elman <aelman users sourceforge net>]

On Monday, November 12, 2001, at 12:06 PM, Havoc Pennington wrote:
> An idea we've been tossing around at Red Hat and was also discussed
> some at ALS: rather than user levels, let's just remove overly
> advanced/confusing prefs from the UI entirely, and have a special
> power user control panel that exposes the most common ones, and for
> really wacky settings people can use a generic GConf editor app or
> gconftool. Windows takes a similar approach, I'm told you can get a
> power-tweaker application that lets you set the weird stuff.

I'd also put in a vote for removing user levels entirely for several 
reasons:

1) Users do not fit nicely into categories of "Novice," "Intermediate," 
and "Advanced."  Many people might be extremely comfortable with a 
couple of applications, but might have little understanding of the 
underlying system.  Others might have deep understandings of how 
low-level functionality works, but be totally unfamiliar with the GUI.  
(Alan Cooper has some other good examples of this in his book "The 
Inmates Are Running The Asylum," in which he calls this 3-level division 
the "Euphemism Pyramid.")

2) Given that users don't fit nicely into these categories, it's 
impossible to fit preferences into these categories anyway, especially 
with 3 levels -- with 2, it's a bit easier.  I know there has been much 
confusion in the past about what preferences should be in 
"Intermediate" -- that's because the division is arbitrary.

3) Moreover, since the division of preferences is unclear and hence 
arbitrary, users have no idea how to get to that one preference that 
they really need to set, so users end up having to switch user levels 
and explore the control panels again and again to find the preference, 
negating the whole purpose of having the user levels in the first place.

I don't think that a "power user control panel" or "power-tweaker 
application" are the right approach either, though.  The right answer, 
unfortunately, is hard design work: Figure out what users are actually 
doing/going to do with your application, figure out the set of 
preferences which most users will likely want to customize to their own 
needs, and make those available.  Then figure out the set of preferences 
which _some_ users will need, and make those available -- not through an 
"Advanced Settings" or a disclosure triangle, ideally, but in a way that 
clearly denotes what preferences are going to be found.

For instance (and this is hypothetical, so if you disagree with this, 
that's fine :): most people will want to be able to customize their GTK 
theme.  Fewer will want to assign custom colors or custom fonts beyond 
what the theme specifies.  So you have a theme selector, and on the 
theme selector dialog is a button that says "Customize Colors and 
Fonts..." which pulls up a secondary dialog to do that.

And if almost nobody will ever use a preference, then why have it at 
all?  Customization is really important (and certainly a hallmark of 
GNOME and free software in general), but an overwhelming amount of 
flexibility is actually one of the core usability issues in GNOME (and, 
in fact, in Windows too IMHO).

That's my 2 cents,
Adam