Theory or Data first?

[Martin Taylor 2018.11.20.15.34]

Rick frequently says that my approach to PCT is "theory first", and he couches it in the form of an accusation, as though it was in some way wrong. Whether it is true of my work in general, I couldn't say. I've been theorist seeking data, experimenter seeking theory, computer hardware designer, and other things during my professional life. But I do have a distinct perception of this "theory first" versus "data first" or "phenomena first" obsession of Rick's.

It seems to me that this issue should be seen as a large-scale example of perceptual control in action. A "theory" is a perception of the way some small part of the world works, data is the sensory input on which the perception is built. Neither exists in isolation. Data is data only in light of a theory that gives it meaning, theory is theory only of it a theory about something that could be made manifest in data. It is a feedback loop.

Apart from my thesis, my very first solo published paper was very purely "theory first". It was based entirely on the mathematics of geodesics in a non-Euclidean space, using some findings from observations in my thesis of consistent patterns of error in placing dots on index cards. Using the appropriate differential geometry, I predicted three main effects, one being the existence of an optical illusion that, so far as I knew had never been described. The second thing was that the magnitude of the illusion would be linear with the radius of a disk, and the third thing was rather counterintuitive, that those subject who were most precise would be the ones showing the strongest illusion. All these were pure theory based on geometry, but they suggested that some data that had never been looked for might be worth looking for. So I did, and all three predictions were exactly as predicted, as were a couple of subsidiary predictions that depended on the patterning of the circular disk.

If you don't have theory that extends beyond what is known, you will not look for new kinds of data. If you don't have a mysterious phenomenon for which you have no theory, you won't look for other ways your Universe might actually work. To say theory or data should come first makes no sense to me. Some people do both theory and data, some people prefer to gather data, which, with luck, they will not be able to understand, while some people extend theory which, with luck, will lead other people to seek data they would otherwise not have sought.

For one example of the latter, just consider Einstein and relativity theory, based entirely on the mathematics of non-Euclidean spaces. Who would have thought to look near the edge of the eclipsed sun to see whether a star had moved in the sky? The whole idea would have seemed ridiculous, and to many people it did, since it was all based on "just mathematics". But the star did move. And now we have GPS satellites that have to be adjusted precisely according to what Einstein discovered in his mathematics or the GPS systems in our cars would soon tell us we were in a different continent.

I think is is nonsense to suggest that either theory first or data (or phenomena) first is better than the other. They are mutual complements, like an environmental variable and its perceptual complement. But neither is much use for advancing your understanding if it simply repeats what has been done before. The phenomena may suggest the need for new theory; the theory may suggest the need to seek new kinds of data.

Martin

From Fred Nickols (2018.11.19.1711 ET)

At the risk of stating the obvious, Kurt Lewin famously wrote, “Nothing is quite so practical as a good theory.â€?

···

Fred Nickols
Distance Consulting LLC
“Assistance at A Distanceâ€?
www.nickols.us

[Rick Marken 2018-11-19_19:05:51]

[Martin Taylor 2018.11.20.15.34]

Rick frequently says that my approach to PCT is “theory first”,

RM: It’s really more like “theory only”.Â

MT: Apart from my thesis, my very first solo published paper was very purely

“theory first”. It was based entirely on the mathematics of geodesics in

a non-Euclidean space, using some findings from observations in my

thesis of consistent patterns of error in placing dots on index cards.

RM: It looks like “phenomena first” to me since you used some observations as the basis of the mathematics that p[redicted the illusion.Â

Â

MT: If you don’t have theory that extends beyond what is known, you will not

look for new kinds of data.

RM: I think every good theory extends beyond what is known in the sense that it makes predictions about what will be observed that has never been observed before.Â

Â

MT: To say theory or data should come first makes no sense to me. Some people do both theory and data, some people prefer to gather data, which, with luck, they will not be able to understand, while some people extend theory which, with luck, will lead other people to seek data they would otherwise not have sought.

RM: It’s extending theory just for the sake of extending theory that I have been calling “theory first”. I should call it “theory only”. Theories get usefully extended or completely revised based on data! Extending theory just for the sake of extending theory is not just a waste of time; it’s an impediment to the the development of a scientific understanding of the nature of reality.Â

MT: For one example of the latter, just consider Einstein and relativity theory, based entirely on the mathematics of non-Euclidean spaces.

RM: Einstein was not just “extending a theory” to show how smart he was; or because he liked to do theory rather than observation. Einstein knew that there was data inconsistent with the existing Newtonian model of the universe and he developed his theory with an eye to solving this problem. If you were “extending” PCT for the same reason Einstein extended Newton I would have no problem with what you do. But you can’t possibly do that because PCT hasn’t got anything close to the level of observational confirmation that Newtonian physics had when Einstein did his work.Â

Â

MT: Who would have thought to look near the edge of the eclipsed sun to see whether a star had moved in the sky?

RM: No one whould have thought that until Einstein came up with his theory that explained everything Newton’s had plus the problem observations (eg., Mickelson-Morley) and predicted the gravitational effects on light.

MT: I think is is nonsense to suggest that either theory first or data (or phenomena) first is better than the other.

RM I agree. Both are equally important. What I object to is extending theory for no reason other than that you want to do it. Theory should be extended only when the existing theory fails to explain some observation – or fails to explain it elegantly (parsimoniously).Â

Â

MT: They are mutual complements,

RM:Â Righto. So let’s see some mutuality in your “theory extending”. Don’t go extending a theory unless the existing theory can’t explain some data. And when you make the extension that explains the previously unexplained data show that the extension now explains that data by showing that the model fits it.Â

BestÂ

Rick

···

like an environmental variable and its perceptual complement. But

neither is much use for advancing your understanding if it simply

repeats what has been done before. The phenomena may suggest the need

for new theory; the theory may suggest the need to seek new kinds of data.

Martin


Richard S. MarkenÂ

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
                --Antoine de Saint-Exupery

[Martin Taylor 2018.11.20.10.24]

[Rick Marken 2018-11-19_19:05:51]

  I won't bother commenting individually on your interleaved

comments. Instead, I will just say that I disagree with almost all
of them, especially your characterization of what Einstein did.

  As for my own efforts, I do not believe I extend PCT other than

by allowing findings in other sciences to be included. What I do
believe I do is to show how widely PCT can be applied to cases in
which more than one single control loop are involved. That’s not
extending the theory, but demonstrating its power. That’s the
reason why the working title of my book is the deliberate pun
“Powers of Perceptual Control”.

Martin

···
        [Martin

Taylor 2018.11.20.15.34]

        Rick frequently says that my approach to PCT is "theory

first",

RM: It’s really more like “theory only”.Â

        MT: Apart

from my thesis, my very first solo published paper was very
purely

        "theory first". It was based entirely on the mathematics of

geodesics in

        a non-Euclidean space, using some findings from observations

in my

        thesis of consistent patterns of error in placing dots on

index cards.

        RM: It looks like "phenomena first" to me since you used

some observations as the basis of the mathematics that
p[redicted the illusion.Â

Â

        MT: If you

don’t have theory that extends beyond what is known, you
will not

        look for new kinds of data.
        RM: I think every good theory extends beyond what is

known in the sense that it makes predictions about what will
be observed that has never been observed before.Â

Â

        MT: To say

theory or data should come first makes no sense to me. Some
people do both theory and data, some people prefer to gather
data, which, with luck, they will not be able to understand,
while some people extend theory which, with luck, will lead
other people to seek data they would otherwise not have
sought.

        RM:Â  It's extending theory just for the sake of extending

theory that I have been calling “theory first”. I should
call it “theory only”. Theories get usefully extended or
completely revised based on data! Extending theory just for
the sake of extending theory is not just a waste of time;
it’s an impediment to the the development of a scientific
understanding of the nature of reality.Â

        MT: For

one example of the latter, just consider Einstein and
relativity theory, based entirely on the mathematics of
non-Euclidean spaces.

        RM: Einstein was not just "extending a theory" to show

how smart he was; or because he liked to do theory rather
than observation. Einstein knew that there was data
inconsistent with the existing Newtonian model of the
universe and he developed his theory with an eye to solving
this problem. If you were “extending” PCT for the same
reason Einstein extended Newton I would have no problem with
what you do. But you can’t possibly do that because PCT
hasn’t got anything close to the level of observational
confirmation that Newtonian physics had when Einstein did
his work.Â

Â

        MT: Who 

would have thought to look near the edge of the eclipsed sun
to see whether a star had moved in the sky?

        RM: No one whould have thought that until Einstein came

up with his theory that explained everything Newton’s had
plus the problem observations (eg., Mickelson-Morley) and
predicted the gravitational effects on light.

        MT: I

think is is nonsense to suggest that either theory first or
data (or phenomena) first is better than the other.

        RM I agree. Both are equally important. What I object to

is extending theory for no reason other than that you want
to do it. Theory should be extended only when the existing
theory fails to explain some observation – or fails to
explain it elegantly (parsimoniously).Â

Â

        MT: They

are mutual complements,

        RM:Â  Righto. So let's see some mutuality in your "theory

extending". Don’t go extending a theory unless the existing
theory can’t explain some data. And when you make the
extension that explains the previously unexplained data show
that the extension now explains that data by showing that
the model fits it.Â

BestÂ

Rick

        like an

environmental variable and its perceptual complement. But

        neither is much use for advancing your understanding if it

simply

        repeats what has been done before. The phenomena may suggest

the need

        for new theory; the theory may suggest the need to seek new

kinds of data.

        Martin


Richard S. MarkenÂ

                                "Perfection

is achieved not when you have
nothing more to add, but when you
have
nothing left to take away.�
   Â
            --Antoine de
Saint-Exupery

Fred Nickols (2018.11.20.1157 ET)

Well, as long as we’re fessin’ up here, I might as well make clear why I’m here.

First off, I joined the list to advance my grasp of PCT and, thank goodness, Bill Powers was still here to help that along.

Second, I’m no theorist or mathematician or scientist; I’m a down-in-the-trenches, get-stuff-done kind of guy, so I’m interested primarily in the practical implications and applications of PCT to human behavior and performance - primarily in the workplace.

PCT gave rise to my GAP-ACT (Target Model) and I am extremely grateful for that.

PCT also reinforces my long-held views about what “feedback” is and isn’t.

PCT, with its focus on variables, has helped me see that a desired result is usefully viewed as a desired value for some variable. I know of no other place where that view prevails. Moreover, viewing results as desired values for variables leads quickly to an examination of the structure in which that variable is embedded and thence to finding ways and means of affecting it, individually and collectively.

Finally, I am becoming ever more convinced that successful collective control is at the heart of successful, high-performing teams.

So, to sum it up, I’m here for what I can get out of this list and put to good use.

···

Regards,

Fred Nickols

Managing Partner

Distance Consulting LLC

“Assistance at A Distance”

[Martin Taylor 2018.11.21.09.37]

···

[Rick Marken 2018-11-19_19:05:51]

        [Martin

Taylor 2018.11.20.15.34]

        Rick frequently says that my approach to PCT is "theory

first",

RM: It’s really more like “theory only”.

        MT: Apart

from my thesis, my very first solo published paper was very
purely

        "theory first". It was based entirely on the mathematics of

geodesics in

        a non-Euclidean space, using some findings from observations

in my

        thesis of consistent patterns of error in placing dots on

index cards.

        RM: It looks like "phenomena first" to me since you used

some observations as the basis of the mathematics that
p[redicted the illusion.

  Would you care to say which of these phenomena were "first",

before the theory?

  1. The existence of a previously undescribed visual illusion.

    1. The relation between the size of the illusion and the radius
      of the disk in the illusion-inducing display.

    2. That people whose vision is most precise are the most affected
      by the illusion.

    3. That the patterning of the disk affects the magnitude of the
      illusion.

    These were all unknown, at least to me and so far as I know
    unknown also to the scientific literature before the theory.

    So what led to the theory? Phenomena leading to data, of course,
    and what led to them? Earlier theory partly, unexplained earlier
    data partly, but the unexplained early data had led to different
    kinds of theory based on peculiarities of memory or of mind. The
    difference with this particular theory was that it was based on
    analytical geometry, and was not an extension of the psychological
    theory of the time, or at least not much.

    As I see the intellectual path to the idea that one should look
    to see whether these previously unobserved phenomena actually
    exist, it is a continuous feedback loop between theory (perceptual
    function) and data (an approximation to real reality).

        MT: If you

don’t have theory that extends beyond what is known, you
will not

        look for new kinds of data.
        RM: I think every good theory extends beyond what is

known in the sense that it makes predictions about what will
be observed that has never been observed before.

  Well, that is a paraphrase of what I said. At least we agree. But

I would not use your word “good”, unless you also add that the
predictions turn out to be correct. Any so-called “theory” that
does not extend beyond a description of what has been observed is
just a description undeserving of the name of “theory”.

        MT: To say

theory or data should come first makes no sense to me. Some
people do both theory and data, some people prefer to gather
data, which, with luck, they will not be able to understand,
while some people extend theory which, with luck, will lead
other people to seek data they would otherwise not have
sought.

Another paraphrase of what we both said.

        RM:  It's extending theory just for the sake of extending

theory that I have been calling “theory first”.

  Who in their right mind would do that? You are suggesting that

someone has a reference value for a perception of some theory that
differs from their current perception of the theory, quite
independent of any data the theory purports to be able to predict.
I suppose you might say that this is what all mathematicians do,
since their topic is the internal consistency of their theorems,
and in that sense, it is what all good theorists or experimenters
might be expected to do as well. Inconsistency within a theory
suggests that the theory needs to be extended, at least if you
have a reference perception for perceiving the world as being
unified under one set of internally consistent “Natural Laws”.

        I should call it "theory only". Theories get usefully

extended or completely revised based on data! Extending
theory just for the sake of extending theory is not just a
waste of time; it’s an impediment to the the development of
a scientific understanding of the nature of reality.

  Wrong, as myriads of examples show. Theories do get incrementally

extended because data is not completely explained by them. Even in
a standard PCT tracking study the data show anomalies that have
not been explained. More importantly, the only serious advances in
the “development of a scientific understanding of the nature of
reality” have come from a consideration of the intellectual base
of the current theory and the substitution of a quite different
base. The need for such a radical rethinking certainly may come
from a recognition either of some anomaly in the data, but it may
also come from rethinking the “why” of the theory. The very
existence of PCT comes from just such a radical reconception of
the underlying basis of all intentional behaviour.

  In my example of the illusion based on visual space curvature,

the earlier data had been sought and found because of theories of
how the brain and mind work. I ignored that way of thinking about
the problem, apart from recognizing that something in the brain
was capable of visually stretching and compressing perceived space
relative to other measures of the space. That wasn’t new. Lots of
space-distorting illusions have been known for centuries. What was
new was to see that looking at the problem as being of the
curvature of space rather than of the workings of the brain could
lead to seeking (and finding) previously unsought data.

  At the moment, and for the last several decades, the big problem

in physics is not that QCD or Relativity don’t predict data. They
both do, with amazing precision always limited by the ability to
measure the data. But they can’t (or haven’t yet) been reconciled.
Most probably, neither gives us a good idea of the nature of
reality, any more than the long-successful phlogiston theory gave
us a good idea about the nature of the flow of heat from a hot
body to a cold one. The anomaly to that one was the finding that
phlogiston was imponderable, not that it wrongly predicted
temperature changes. I have faith that somebody, sometime, will
find a different way of looking at the world that replaces QCD and
Relativity in the way that Relativity replaced Newtonian gravity
forces while not replacing the accuracy of Newton’s equations
where their sizes and/or velocities are human-scale.

  As for my "extending PCT for the sake of it" there are at least

two independent facets. Firstly, I don’t. If I extend PCT as a
theory, it is because either the theory is insufficiently
precisely defined to account for some data and needs extending in
the precision with which the theory is described, or because the
theory does not take into account data that other science have
discovered, such as the pervasive (and clearly necessary) use of
lateral inhibition by the brain, or the effects of hormonal
changes in the body.

  I don't consider it "extending PCT" when I consider how PCT

applies to understanding the effects of many people controlling
their perceptions by influencing closely related aspects of their
environment. Collective control is just an application of PCT, as
is the CROWD demo. I do consider it “extending PCT” when I
consider possible ways in which reorganization might happen. That
is an issue that worried Bill throughout the 60 or 70 years he
worked with PCT. It’s an extension PCT sorely needs, and in that
area I do try to find new ways of thinking about the problem. I
don’t think it to be “extending PCT” when I use findings from
other sciences, or casual observations of current political
problems as a reason for thinking about whether straight HPCT can
account for them or is consistent with them. As I said before, I
have faith that there is one world, and if PCT disagrees with some
findings in other sciences, something is wrong somewhere.

  And I have faith that PCT, whether based on HPCT or not (I

suspect it will be), has to be part of all life if life and
thermodynamics are to be mutually consistent.

Martin