GOMS Analysis of Watch
Christian mogensen@cs.stanford.edu
The Watch
The watch has three buttons. The Mode button advances the little triangle
indicator from Time to Timer to Alarm to Dualt
to StopW. Advancing from StopW moves the indicator back to
Time and places you in Change Time mode, which is indicated
by having the seconds digits and the second hand display in the topleft
corner flash.
Pressing the Select button will select the next group of digits.
Pressing the Set button will advance the current selection.
The buttons are stiff and give little feedback when pressed. The
user needs to press hard in order to get a response.
GOMS framework
- Goal: Advance minutes by two
- Goal: Change time on watch
- Goal: Enter change time mode
- Goal: Advance mode if digits not flashing
- Press and Release Mode button.
- Check digits are flashing.
- Goal: Advance minutes by two
- Goal: Select minutes.
- Press and Release Select button.
- Goal: Advance minuites twice
- Press and Release Set button.
- Press and Release Set button.
- Goal: Lock in changes
- Goal: exit change mode
- Press and Release Mode button
- Check digits no longer flashing
GOMS predictions
The number of actions in this sequence are 10 to select the CHANGE
mode from TIME mode, plus 5 to set the minutes by two.
This gives us a rough figure of 15*0.25s cycle time = 4 seconds
if actions are automatic operations.
Since these are new operations, it will take significantly longer for
each operation to be performed as it has to be thought over; the
time to respond will be high: it should be close to 200 msec or so.
Motor response will also be higher since the actions will tend to be
unfamiliar (Will digital versus analog watch wearers fare differently?).
Estimate close to 100 ms.
This gives us 15 operations at 0.40s cycle time, or close to 6 seconds
to change the time when the operation is unfamiliar.
Results
These results are shared with Nat Johnson
These data are hand timed - so resolution is to about 0.1 s.
n Trial A Trial B Trial C
0 16.4 4.2 5.2
1 6.1 4.0 4.0
2 5.7 9.2 * 4.2
3 4.7 11.5 ** 3.8
4 4.5 5.2 20+ ***
5 4.2 4.2 3.6
* One error: pushed mode once too many times
** Two errors: mode error plus advanced seconds instead
*** One error: advanced hours by mistake, started over
Analysis
-
The data for subject A follows the Power Law of Learning so closely it's eerie.
The subject had a similar but simpler digital watch. This does not
appear to been an advantage.
- Subject B was vaguely familiar with a similar type of watch.
The subject appears to be very familiar with watch setting, which
is strange, since he uses an analog pocketwatch. The high error
rate (3 errors/6 trials) leads me to discount these measurements.
- Subject C was very familiar with the watch, but even here there is
operationalizing of actions: the times got progressively better,
but by very small increments,indicating that n (the trial number)
is large.
Conclusion
The experiments confirm the power law of learning. The GOMS predictions
turned out to be a little on the high side. People learn simple
operations very quickly.
GOMS modelling does not degrade gracefully when in error.
Error times can estimated for error recognition and recovery plan formulation.
I estimate three cycles - close to 0.4-0.8 s:
- error condition recognition (perception and cognition)
- error recovery planning (cognition cycles)
- error recovery task scheduled (cognition cycle)
The exception can then be handled by something like this:
- Goal: Enter change time mode
- Goal: Advance mode if digits not flashing
- Exception: Advanced mode when digits flashing.
- Redo Previous Goal Advance Mode if digits not flashing
GOMS needs subroutines - the strict hierarchy of tasks does not map
onto real world tasks, where component methods may be repeated in different
contexts.
Christian mogensen@cs.stanford.edu
http://www-pcd.stanford.edu/mogens/377/goms-watch.html