OT? PSP development

Hi George,

On 10/21/2013 3:44 PM, George Neuner wrote:
> On Mon, 21 Oct 2013 03:06:08 -0700, Don Y <this@isnotme.com> wrote:
>
>> My preferred UI modality is earpiece and touchpad.
>
> I just want to talk to the computer and have it understand and answer
> me intelligently.  But I don't think I'll make it to the 23rd century.
> Even with training, voice recognition still makes too many mistakes,
> and AI still isn't anywhere close to where it needs to be.

For unconstrained, speaker-independant speech, I suspect the goal
is a *long* time away (not sure I would say 23rd century, though).

OTOH, for speaker dependent, fixed vocabulary, I think you can get
"decent" performance, today.  Esp if you pay careful attention to
the design of the vocabulary at each stage of a "user interaction".

>> It seems (empirically) that a handheld interface can be smaller
>> for a given set of "disabilities".
>
> Mostly correct.  Palsy is a problem for any touch interface.  I have a
> friend who can't hit a button smaller than about 2 inches square.

I would *love* to see/hear how folks with any type of tremor deal
with *pen* interfaces.

>> Interface is intended to be terse.  Imagine how the exact same
>> interface is mapped onto an aural modality.  "Lots of text"
>> read to the user would place a high cognitive load on him/her.
>> Instead, you'd opt for abbreviated representations.
>>
>> "71 degrees cooling to 68" instead of "The HVAC system is set to
>> cool mode.  The current setpoint temperature is 68 degrees but the
>> present indoor temperature is 71 degrees.  As a result, the ACbrrr
>> is actively cooling the house at this time.  The override is not
>> currently in effect."
>
> It's just as complex visually: the user must at least be able to
> distinguish the current temp from the set temp (by label, ordering or
> whatever), and it's helpful to have indication that the system is on
> and in which direction it's going.  Whether indication is given by
> colors or by symbols, etc. doesn't really matter.

Yes, but, once learned, a user can know that "the big number" is
the current temperature while the "little number" is the setpoint.
They could be arranged visually above/below to indicate heating
vs cooling, etc.

You have more ways of conveying information over the visual channel
than, for example, aural.

Use different voices for temperature vs. setpoint?  Or different
pitches, etc.?

>> [I believe "aural memory" is significantly different/shallower
>> than "visual memory".  I.e., it seems like it takes more effort
>> to remember a long string of words received aurally than the
>> same information presented in a visual form.  Though that may
>> just be a personal limitation?  I've been exploring other
>> options to convey information aurally -- "sound icons" - in
>> the hope that you can remember a particular *sound* instead
>> of specific words.  But, that "lexicon" quickly becomes yet
>> another cognitive challenge!  :-/ ]
>
> The problem is that words, in general, convey much less information
> than do pictures.

Exactly.  You want to remember an *idea* not some arbitrary set
of sounds (phonemes).

> Memory is organized and accessed in terms of
> "symbols" which are largely independent of the amount of information
> associated with them.  Short term memory can handle only 7 to 10
> symbols, regardless of the amount of information associated with the
> symbols.

It depends on the relationship of the "ideas" (symbols in your terms)
to each other.  Remembering a 10 digit number is relatively easy.
Remembering the names of ten "random" items is considerably harder:
    lion pencil blue eucalyptus overflow chromium bread irish cold buoy
When multiple words are thrown together to convey a single "idea"
(current temperature), the surplus of words cuts into the ability to
remember the *next* "idea" (which is also wrapped in multiple words).

> E.g., for most people using directory assistance, "7 1 6" would be a
> string of more or less arbitrary digits.  To me, it means "Buffalo"
> and conveys a whole lot of information - including that its area code
> is 716 - that the string of digits does not.  When I hear a phone
> number starting with "7 1 6", I immediately substitute "Buffalo" and
> discard the digits - one symbol for three.  Doesn't work with "2 1 6"
> because that string of digits means nothing to me.

Yet "7316" might represent a considerably larger cognitive load (for
you).

I meet a lot of new people each week/month.  Many of these I may see
only once -- then stumble across them, again, many months later!
Invariably, they are amazed (that I can recall their name and some
"specifics" about them) and embarassed (that they can't recall *my*
name -- though they invariably recognize my visage).

The trick, of course, is to *force* an association between them,
their name and some "other idea" that gives you (me) a variety of
handles to retrieve the data/tuple.

[I had a relation who's address was "356 <name_of_street>".  I
associated the address with # days in a year -- only different! :> ]

In a user interface (to some sort of "system"), you have to provide
terse ways of conveying data in "harmonious" sets -- the equivalent
of (programming) *idioms*.