Discussion: View Thread

Feedforward about Feedback

  • 1.  Feedforward about Feedback

    Posted 06-05-2005 11:47
    Fred Nichols offered ---
    [...]
    > And, for those who are interested, the "Feedback about Feedback" article
    > [...] it's still available on my articles web site. [..]>
    > http://home.att.net/~OPSINC/feedback.pdf

    Jack Ring responds ---
    Fred makes a valuable point. I exercised a similar idea a couple of years
    ago on a list populated by systems engineers. It escalated to a Zoomerang
    survey which results I interpreted as the majority being very imprecise in
    their thinking. Others gave me 'feedback' that the survey was not well
    designed. Sigh.

    Feedback is misused in many technical disciplines and in most
    "non-technical" disciplines. Misuse occurs in two ways. 1) assigning the
    label, feedback, to situations in which no feedback is happening and 2)
    using the label, feedback, to educe specific concepts in the minds of others
    without verifying whether the concept educed fits the intention.

    To get clear about feedback it is useful to introduce a term for stuff that
    isn't feedback. Because time is of the essence in feedback a good
    contrasting term is feedforward.

    The roots of misuse include Locus of information extraction vs. application,
    Elapsed time, and Effect of the information when applied.

    in semiconductor fabrication for example, measurements of the wafer are
    taken at Step 5 of the process and used to configure Step 6. This is
    feedforward (also called conditioning) because the point of application of
    the information derived at Step 5 is downstream from Step 5. Similarly when
    we determine what students do and don't know then apply that to curriculum
    design we are committing feedforward.

    Conversely had the output of Step 5 been measured and used at the input of
    Step 5 then that might qualify as Feedback.

    Feedback in the cybernetic sense must occur while it can have an effect, as
    in the amplifier gain setting on Fred's gun direction motor or the automatic
    volume control on your radio. Not only is the feedback signal derived
    downstream and applied upstream but also the time delay must be small enough
    to influence the current results. For example, if the time delay in the
    'feedback' loop is one millisecond then the automatic volume control will
    affect all sounds lower than about 1000 cycles per second but all higher
    frequency ones will have already happened by the time the AVC signal gets
    back to the input of the amplifier. For the higher frequencies the
    'feedback' signal even though applied upstream, is feedforward because it
    affects the next episode, not this one. Similarly, Fred's gun controller
    probably could zero in on a fixed land target or even another, slowly moving
    ship but not on a supersonic airplane passing overhead at low altitude.

    Notice that in most diagrams using 'feedback' the time domain is not even
    mentioned. Thus the problem starts with an imprecise definition of
    'feedback' in the first place.

    So both direction and latency of information flow are pertinent. Third
    comes effect. For example, the Beer Game in the System Dynamics field shows
    that corrective signals with too much time delay will have an effect
    opposite of their effect with small time delay.

    Sorry for the long post. Suffice it to say that most interpersonal
    interactions are feedforward rather than feedback --- Including why your
    spouse has never suggested a second honeymoon.

    cheers,