A Just-So Story

[Avery Andrews 970415:1135(PDT)]

Resetting clocks & watches for daylight savings (actually, the end of
daylight savings in Australia) caused a little thought to take shape:

There's an odd phenomenon in digital clock-setting whereby, if you need
to set the time back, you often have to do it by going forward, and I
at any rate very frequently overshoot the desired time, e.g. I want
to reset the clock from 8 to 7 AM, and do it by pushing the `hour advance
button', but keep pushing it after 7 AM is reached, so I have to go
around the cycle again (and sometimes, again, and again ...).

If this behavior were being controlled by a sensible computer program
(or the TOTE architecture, if I remember that right), this effect
would be hard to explain, since there'd be `desired time' RT (7AM),
and an perceived time PT (whatever's shown on the clock), and a loop whereby
the button got pushed when RT != PT. Leaving no explanation for
overshoot.

But hierarchical control systems to provide a possible explanation.
On this account, we'd still have RT (the reference) and PT (the perception),
but these would be compared not by a step in a program, but by a
comparator in a continuously operating circuit with specific dynamic
characteristics, including a nontrivial amount of slowing, since PT
is a fairly high-level perception. By the hierarchical architecture,
the error signal ET from the (PT,RT) comparison goes somewhere and
contributes to a reference level for some other perception, and the
natural one to suspect here is one for rate of change of PT (dPT/dt),
minimally, if there is an ET > 0, co should dPT/dt be, but other factors
may affect it, such as the desire to be doing something more interesting
than setting a clock. The dPt/dt signal goes on trigger rhythmic
button-pressing (presumably via an oscillator).

So all we need to suppose is that the reference for dPT/dt tends to
get shoved up to a sufficiently hi level so that the whole circuit
isn't fast enough to shut down the button-pressing in time, which seems
plausible given the involvement of several levels of control doing
fairly complicated things, such as comparing numbers, and controlling
the frequency of oscillators.

A just-so story, but I think it at least makes sense.

  Avery.Andrews@anu.edu.au