Re: Celebrating the Higgs - explaning and predicting

[Eric Charles]

Steve,
Interesting paper, but I'm not sure if I follow. The basic argument seems to be that we often explain things by imagining (with the help of statistics) hypothetical constructs that cannot be directly measured. As those constructs can't be measured directly, they don't help us predict things. Thus, predictive models are limited to using things that actually exist, while explanatory models are not so limited.

That seems like a really good argument for coming up with better explanations, not an argument for distinguishing and reifying two distinct modeling tasks.

This is a topic I am quite interested in. I would presume that an ideal explanatory model would be identical to an ideal predictive model, though I grant that non-ideal cases might differ. What am I missing?

Eric



On Tue, Jul 10, 2012 12:37 AM, Steve Smith <[hidden email]> wrote:

Bruce -

I second the motion (very good post)!

Mendeleev's Periodic Chart was *my* first introduction (back when) to the very concept of having a predictive model that was (almost) entirely void of explanatory ability (as it stood when constructed).  I found the notion *fascinating* and it drove me into the field of Visual (Perceptual) Analytics many years later... seeking patterns that yield useful prediction without necessarily waiting for an explanatory model.

For those vaguely interested in the philosophical underpinnings of science, it's methods and utility, I recommend Galit Schmueli's (George Washington U's) paper on Predictive vs Explanatory Models (as well as *Descriptive* models)...

arxiv.org/pdf/1101.0891

Thanks! Glad you liked it!

I have long been bemused by the strong parallels among the various
tales I was able to tell in that post.

Bruce

On Mon, Jul 9, 2012 at 4:43 PM, Pamela McCorduck <pamela@...> wrote:

Bruce, that blog post is marvelous in its simplicity and power.

Pamela



On Jul 9, 2012, at 2:39 PM, Bruce Sherwood wrote:

See my blog:

<a class="moz-txt-link-freetext" href="http://matterandinteractions.wordpress.com/2012/07/09/the-higgs-boson-and-prediction-in-science/" onclick="window.open('http://matterandinteractions.wordpress.com/2012/07/09/the-higgs-boson-and-prediction-in-science/');return false;">http://matterandinteractions.wordpress.com/2012/07/09/the-higgs-boson-and-prediction-in-science/

Bruce

On Sun, Jul 8, 2012 at 10:18 PM, Owen Densmore <owen@...> wrote:

Lets chat about the Higgs discovery, its likely-hood of being correct, and

the impact it will have going forward .. at the next Friam @ St Johns.


Could someone see if Hywel White is available .. or anyone you know who'd

like to hold forth on the topic!


  -- Owen


============================================================

FRIAM Applied Complexity Group listserv

Meets Fridays 9a-11:30 at cafe at St. John's College

lectures, archives, unsubscribe, maps at <a class="moz-txt-link-freetext" href="http://www.friam.org" onclick="window.open('http://www.friam.org');return false;">http://www.friam.org


============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at <a class="moz-txt-link-freetext" href="http://www.friam.org" onclick="window.open('http://www.friam.org');return false;">http://www.friam.org



"Im Deutschen lügt man, wenn man höflich ist."

"In German, if one is polite, one lies."

Goethe, "Faust"


============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at <a class="moz-txt-link-freetext" href="http://www.friam.org" onclick="window.open('http://www.friam.org');return false;">http://www.friam.org

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at <a class="moz-txt-link-freetext" href="http://www.friam.org" onclick="window.open('http://www.friam.org');return false;">http://www.friam.org

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

Eric Charles

Professional Student and
Assistant Professor of Psychology
Penn State University
Altoona, PA 16601

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

[Steve Smith]

Eric -

Interesting paper, but I'm not sure if I follow. The basic argument seems to be that we often explain things by imagining (with the help of statistics) hypothetical constructs that cannot be directly measured. As those constructs can't be measured directly, they don't help us predict things. Thus, predictive models are limited to using things that actually exist, while explanatory models are not so limited.

I think I have been considering it from the opposite perspective...  that predictive models aren't *necessarily* explanatory... I think that *explanatory* models are at least minimally *predictive* (how else can they be validated)?  They may not be sufficient to predict (m)any of the things we are interested in, but I think (though I'm not sure exactly why) the are in fact predictive (by definition?).

That seems like a really good argument for coming up with better explanations, not an argument for distinguishing and reifying two distinct modeling tasks.

I think Science is always seeking models that have more explanatory power.  The business of Science is *understanding* (my contention) with *prediction* being merely a useful mechanism for hypothesis testing and even generation.   Most of us get excited when Science *predicts* something, but I am not sure it is an end in itself (except for engineering, technology development, business purposes).

I think Bruce's description of Mendeleev's Periodic Chart is my easiest example...   the "model" in this case was an almost numerological geometric arrangement of the known elements based on correlations among their properties without any *explanation" of the underlying reasons or causes for those properties.  It was very effective for predicting elements yet to be discovered (recognized, identified?) as well as properties of known elements not yet explored, and thereby helped to structure the *search* for new elements. 

This is a topic I am quite interested in. I would presume that an ideal explanatory model would be identical to an ideal predictive model, though I grant that non-ideal cases might differ. What am I missing?

I think you are correct that an "ideal" model has strong predictive and explanatory properties.  

For Engineers perhaps, predictive models are sufficient, they may not be (very?) interested in explaining *why* a particular material has the properties it does, merely *what* those properties are and how reliable the properties might be under a variety of conditions. 

I'm sure there are others here with good perspective on this question... Bruce? 


- Steve


============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

[Bruce Sherwood]

"For Engineers perhaps, predictive models are sufficient, they may not
be (very?) interested in explaining *why* a particular material has
the properties it does, merely *what* those properties are and how
reliable the properties might be under a variety of conditions."

I don't think this currently true. A big chunk of what used to be
labeled "physics" is now in academic engineering departments with the
name "material science". This consists of exploiting models that
explain observed properties of materials, with the goal of looking for
opportunities to change parameters to get improved behavior. In the
early 1990s I heard a talk by an engineering professor at the science
museum in Toronto, where he explained how such research had led to
concrete many times stronger than it had been, and that the iconic
tall tower in Toronto could not have been built not many years before
it was built, as it relied on much stronger concrete.

In some cases someone sees how, starting from fundamental physics
principles, one can predict that such and such should happen or be. In
other cases an observed phenomenon gets explained in terms of
fundamental physics principles (post-diction), which then suggests how
changes in the situation might yield an improved behavior. Pre-diction
and post-diction both require a deep understanding of how to go from
underlying fundamental principles to the behavior, but pre-diction in
addition requires the imagination to run the argument forward, not
already knowing the answer. That's why I claim that post-diction
("explanation") is more common than pre-diction.

There's a fruitful interplay between pre-diction and post-diction. An
example I've mentioned some time ago, from our intro physics textbook:
When searching for an explanation for spark formation in air (we see
the spark and ask how it occurs, which is post-diction or explanation)
there are a couple of tentative explanations that one can rule out.
Another explanation seems to explain the phenomenon, and the validity
of this post-diction is greatly strengthened by noting that it (and
not the other explanations) pre-dicts that it takes twice the critical
electric field to trigger a spark if the air density is doubled, a
pre-diction that is consistent with observations.

Bruce

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

[scaganoff]

It seems that many scientific fields go through a phase of observation (derisively called "stamp collecting") followed by a phase of classification. If you're lucky then patterns can be picked out of the classification scheme to "predict" where to look for new entities or new interesting phenomena.

The Periodic Table is one of the cited examples. Another example (though perhaps not as good) is the Hertzprung-Russell diagram used in astronomy where stars are plotted onto a graph with luminosity and colour as the two axes. They form a characteristic pattern which had to be explained by any theory of stellar evolution.

I also recall many years ago picking up a book on atomic spectra published in 1901 - some 12 years before the Bohr theory of the atom - which illustrated hundreds of different emission spectra and talked about the relationships between spectral line frequencies in terms of waves and resonances. It reflected a very interesting point in the science where patterns were emerging and calling out for an explanation.

So it seems that a "classification" model can be used to make "predictions" - to see if the pattern extends to unobserved areas - and that this can be independent of an underlying explanatory theory. I think Gell-Mann's QCD models probably fit this idea. The image of the "ten-pin owling skittles" pattern and the mystery of what lies at the tip is very evocative.

Regards,
Saul

On 11 July 2012 06:56, Bruce Sherwood <[hidden email]> wrote:

"For Engineers perhaps, predictive models are sufficient, they may not
be (very?) interested in explaining *why* a particular material has
the properties it does, merely *what* those properties are and how
reliable the properties might be under a variety of conditions."

I don't think this currently true. A big chunk of what used to be
labeled "physics" is now in academic engineering departments with the
name "material science". This consists of exploiting models that
explain observed properties of materials, with the goal of looking for
opportunities to change parameters to get improved behavior. In the
early 1990s I heard a talk by an engineering professor at the science
museum in Toronto, where he explained how such research had led to
concrete many times stronger than it had been, and that the iconic
tall tower in Toronto could not have been built not many years before
it was built, as it relied on much stronger concrete.

In some cases someone sees how, starting from fundamental physics
principles, one can predict that such and such should happen or be. In
other cases an observed phenomenon gets explained in terms of
fundamental physics principles (post-diction), which then suggests how
changes in the situation might yield an improved behavior. Pre-diction
and post-diction both require a deep understanding of how to go from
underlying fundamental principles to the behavior, but pre-diction in
addition requires the imagination to run the argument forward, not
already knowing the answer. That's why I claim that post-diction
("explanation") is more common than pre-diction.

There's a fruitful interplay between pre-diction and post-diction. An
example I've mentioned some time ago, from our intro physics textbook:
When searching for an explanation for spark formation in air (we see
the spark and ask how it occurs, which is post-diction or explanation)
there are a couple of tentative explanations that one can rule out.
Another explanation seems to explain the phenomenon, and the validity
of this post-diction is greatly strengthened by noting that it (and
not the other explanations) pre-dicts that it takes twice the critical
electric field to trigger a spark if the air density is doubled, a
pre-diction that is consistent with observations.

Bruce

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

--
Saul Caganoff
Enterprise IT Architect
Mobile: [phone]
LinkedIn: http://www.linkedin.com/in/scaganoff

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

[Bruce Sherwood]

Good points, Saul.

If I remember correctly, before the Bohr model, people looking at the
hydrogen emission spectrum had already discovered an empirical formula
for the frequencies of the emission lines:

f = constant*(1/n1^2 - 1/n2^2)

Bohr's model yielded the same formula, with Planck's constant times f
being the energy of an emitted "photon" when the atom's "quantum
number" changed from n2 to n1, and the model also provided an
evaluation of the constant in terms of known quantities such as the
mass of the electron.

I should mention that the ten-pin diagram is a graphical
representation of a particular structure in group theory.

Bruce

On Tue, Jul 10, 2012 at 11:05 PM, Saul Caganoff <[hidden email]> wrote:

> It seems that many scientific fields go through a phase of observation
> (derisively called "stamp collecting") followed by a phase of
> classification. If you're lucky then patterns can be picked out of the
> classification scheme to "predict" where to look for new entities or new
> interesting phenomena.
>
> The Periodic Table is one of the cited examples. Another example (though
> perhaps not as good) is the Hertzprung-Russell diagram used in astronomy
> where stars are plotted onto a graph with luminosity and colour as the two
> axes. They form a characteristic pattern which had to be explained by any
> theory of stellar evolution.
>
> I also recall many years ago picking up a book on atomic spectra published
> in 1901 - some 12 years before the Bohr theory of the atom - which
> illustrated hundreds of different emission spectra and talked about the
> relationships between spectral line frequencies in terms of waves and
> resonances. It reflected a very interesting point in the science where
> patterns were emerging and calling out for an explanation.
>
> So it seems that a "classification" model can be used to make "predictions"
> - to see if the pattern extends to unobserved areas - and that this can be
> independent of an underlying explanatory theory. I think Gell-Mann's QCD
> models probably fit this idea. The image of the "ten-pin owling skittles"
> pattern and the mystery of what lies at the tip is very evocative.
>
> Regards,
> Saul
>
> On 11 July 2012 06:56, Bruce Sherwood <[hidden email]> wrote:
>>
>> "For Engineers perhaps, predictive models are sufficient, they may not
>> be (very?) interested in explaining *why* a particular material has
>> the properties it does, merely *what* those properties are and how
>> reliable the properties might be under a variety of conditions."
>>
>> I don't think this currently true. A big chunk of what used to be
>> labeled "physics" is now in academic engineering departments with the
>> name "material science". This consists of exploiting models that
>> explain observed properties of materials, with the goal of looking for
>> opportunities to change parameters to get improved behavior. In the
>> early 1990s I heard a talk by an engineering professor at the science
>> museum in Toronto, where he explained how such research had led to
>> concrete many times stronger than it had been, and that the iconic
>> tall tower in Toronto could not have been built not many years before
>> it was built, as it relied on much stronger concrete.
>>
>> In some cases someone sees how, starting from fundamental physics
>> principles, one can predict that such and such should happen or be. In
>> other cases an observed phenomenon gets explained in terms of
>> fundamental physics principles (post-diction), which then suggests how
>> changes in the situation might yield an improved behavior. Pre-diction
>> and post-diction both require a deep understanding of how to go from
>> underlying fundamental principles to the behavior, but pre-diction in
>> addition requires the imagination to run the argument forward, not
>> already knowing the answer. That's why I claim that post-diction
>> ("explanation") is more common than pre-diction.
>>
>> There's a fruitful interplay between pre-diction and post-diction. An
>> example I've mentioned some time ago, from our intro physics textbook:
>> When searching for an explanation for spark formation in air (we see
>> the spark and ask how it occurs, which is post-diction or explanation)
>> there are a couple of tentative explanations that one can rule out.
>> Another explanation seems to explain the phenomenon, and the validity
>> of this post-diction is greatly strengthened by noting that it (and
>> not the other explanations) pre-dicts that it takes twice the critical
>> electric field to trigger a spark if the air density is doubled, a
>> pre-diction that is consistent with observations.
>>
>> Bruce
>>
>> ============================================================
>> FRIAM Applied Complexity Group listserv
>> Meets Fridays 9a-11:30 at cafe at St. John's College
>> lectures, archives, unsubscribe, maps at http://www.friam.org
>
>
>
>
> --
> Saul Caganoff
> Enterprise IT Architect
> Mobile: [phone]
> LinkedIn: http://www.linkedin.com/in/scaganoff
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

[Nick Thompson]

Eric,

 

I have not read the original article, but still your comments caught my attention.

 

As argued in this article, the disassociation of predictive and explanatory power seems misguided.  I suppose a statistical function of many observational variables could have no explanatory power beyond the many variables on which it is based, but then it would only predict what it already knew.  It would “just” be an intervening variable, and not a hypothetical construct, at all.   As we have agreed, some explanations can have “facetious” content, that is not predictive, but that content is not really explanatory, either.  Darwin certainly did not believe that Nature was a breeder who chose the better adapted individuals for breeding. 

 

Further, the idea of a distinction between that which can be directly or indirectly measured also seems a bit strange.  Every measurement is based on a “measurement theory” that tells you that the reading you make on the dial is a valid measure of the thing you actually care about.  Measurement theories fail all the time.  So, what then is a “direct” measure? 

 

One possible response to this comment might be to just tell me to piss off until I have read the article you are referring to. 

 

Nick

 

PS  And isn’t  “real existence” the ultimate hypothetical construct?

 

From: [hidden email] [mailto:[hidden email]] On Behalf Of ERIC P. CHARLES
Sent: Tuesday, July 10, 2012 2:41 PM
To: Steve Smith
Cc: The Friday Morning Applied Complexity Coffee Group
Subject: Re: [FRIAM] Celebrating the Higgs - explaning and predicting

 

Steve,
Interesting paper, but I'm not sure if I follow. The basic argument seems to be that we often explain things by imagining (with the help of statistics) hypothetical constructs that cannot be directly measured. As those constructs can't be measured directly, they don't help us predict things. Thus, predictive models are limited to using things that actually exist, while explanatory models are not so limited.

That seems like a really good argument for coming up with better explanations, not an argument for distinguishing and reifying two distinct modeling tasks.

This is a topic I am quite interested in. I would presume that an ideal explanatory model would be identical to an ideal predictive model, though I grant that non-ideal cases might differ. What am I missing?

Eric



On Tue, Jul 10, 2012 12:37 AM, Steve Smith <[hidden email]> wrote:

Bruce -

I second the motion (very good post)!

Mendeleev's Periodic Chart was *my* first introduction (back when) to the very concept of having a predictive model that was (almost) entirely void of explanatory ability (as it stood when constructed).  I found the notion *fascinating* and it drove me into the field of Visual (Perceptual) Analytics many years later... seeking patterns that yield useful prediction without necessarily waiting for an explanatory model.

For those vaguely interested in the philosophical underpinnings of science, it's methods and utility, I recommend Galit Schmueli's (George Washington U's) paper on Predictive vs Explanatory Models (as well as *Descriptive* models)...

arxiv.org/pdf/1101.0891

Thanks! Glad you liked it!

I have long been bemused by the strong parallels among the various

tales I was able to tell in that post.

Bruce

On Mon, Jul 9, 2012 at 4:43 PM, Pamela McCorduck <[hidden email]> wrote:

Bruce, that blog post is marvelous in its simplicity and power.

Pamela

On Jul 9, 2012, at 2:39 PM, Bruce Sherwood wrote:

See my blog:

http://matterandinteractions.wordpress.com/2012/07/09/the-higgs-boson-and-prediction-in-science/

Bruce

On Sun, Jul 8, 2012 at 10:18 PM, Owen Densmore <[hidden email]> wrote:

Lets chat about the Higgs discovery, its likely-hood of being correct, and

the impact it will have going forward .. at the next Friam @ St Johns.

Could someone see if Hywel White is available .. or anyone you know who'd

like to hold forth on the topic!

  -- Owen

============================================================

FRIAM Applied Complexity Group listserv

Meets Fridays 9a-11:30 at cafe at St. John's College

lectures, archives, unsubscribe, maps at http://www.friam.org

============================================================

FRIAM Applied Complexity Group listserv

Meets Fridays 9a-11:30 at cafe at St. John's College

lectures, archives, unsubscribe, maps at http://www.friam.org

"Im Deutschen lügt man, wenn man höflich ist."

"In German, if one is polite, one lies."

Goethe, "Faust"

============================================================

FRIAM Applied Complexity Group listserv

Meets Fridays 9a-11:30 at cafe at St. John's College

lectures, archives, unsubscribe, maps at http://www.friam.org

============================================================

FRIAM Applied Complexity Group listserv

Meets Fridays 9a-11:30 at cafe at St. John's College

lectures, archives, unsubscribe, maps at http://www.friam.org

 

============================================================

FRIAM Applied Complexity Group listserv

Meets Fridays 9a-11:30 at cafe at St. John's College

lectures, archives, unsubscribe, maps at http://www.friam.org

Eric Charles

Professional Student and
Assistant Professor of Psychology
Penn State University
Altoona, PA 16601

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

[Dean Gerber]

Excellent series of explanations, Bruce. Do you by chance have a specific reference to the ten-pin structure and its relation to group theory?  Thanks ... Dean Gerber

---

From: Bruce Sherwood <[hidden email]>
To: The Friday Morning Applied Complexity Coffee Group <[hidden email]>
Sent: Wednesday, July 11, 2012 9:38 AM
Subject: Re: [FRIAM] Celebrating the Higgs - explaning and predicting

Good points, Saul.

If I remember correctly, before the Bohr model, people looking at the
hydrogen emission spectrum had already discovered an empirical formula
for the frequencies of the emission lines:

f = constant*(1/n1^2 - 1/n2^2)

Bohr's model yielded the same formula, with Planck's constant times f
being the energy of an emitted "photon" when the atom's "quantum
number" changed from n2 to n1, and the model also provided an
evaluation of the constant in terms of known quantities such as the
mass of the electron.

I should mention that the ten-pin diagram is a graphical
representation of a particular structure in group theory.

Bruce

On Tue, Jul 10, 2012 at 11:05 PM, Saul Caganoff <[hidden email]> wrote:

> It seems that many scientific fields go through a phase of observation
> (derisively called "stamp collecting") followed by a phase of
> classification. If you're lucky then patterns can be picked out of the
> classification scheme to "predict" where to look for new entities or new
> interesting phenomena.
>
> The Periodic Table is one of the cited examples. Another example (though
> perhaps not as good) is the Hertzprung-Russell diagram used in astronomy
> where stars are plotted onto a graph with luminosity and colour as the two
> axes. They form a characteristic pattern which had to be explained by any
> theory of stellar evolution.
>
> I also recall many years ago picking up a book on atomic spectra published
> in 1901 - some 12 years before the Bohr theory of the atom - which
> illustrated hundreds of different emission spectra and talked about the
> relationships between spectral line frequencies in terms of waves and
> resonances. It reflected a very interesting point in the science where
> patterns were emerging and calling out for an explanation.
>
> So it seems that a "classification" model can be used to make "predictions"
> - to see if the pattern extends to unobserved areas - and that this can be
> independent of an underlying explanatory theory. I think Gell-Mann's QCD
> models probably fit this idea. The image of the "ten-pin owling skittles"
> pattern and the mystery of what lies at the tip is very evocative.
>
> Regards,
> Saul
>
> On 11 July 2012 06:56, Bruce Sherwood <[hidden email]> wrote:
>>
>> "For Engineers perhaps, predictive models are sufficient, they may not
>> be (very?) interested in explaining *why* a particular material has
>> the properties it does, merely *what* those properties are and how
>> reliable the properties might be under a variety of conditions."
>>
>> I don't think this currently true. A big chunk of what used to be
>> labeled "physics" is now in academic engineering departments with the
>> name "material science". This consists of exploiting models that
>> explain observed properties of materials, with the goal of looking for
>> opportunities to change parameters to get improved behavior. In the
>> early 1990s I heard a talk by an engineering professor at the science
>> museum in Toronto, where he explained how such research had led to
>> concrete many times stronger than it had been, and that the iconic
>> tall tower in Toronto could not have been built not many years before
>> it was built, as it relied on much stronger concrete.
>>
>> In some cases someone sees how, starting from fundamental physics
>> principles, one can predict that such and such should happen or be. In
>> other cases an observed phenomenon gets explained in terms of
>> fundamental physics principles (post-diction), which then suggests how
>> changes in the situation might yield an improved behavior. Pre-diction
>> and post-diction both require a deep understanding of how to go from
>> underlying fundamental principles to the behavior, but pre-diction in
>> addition requires the imagination to run the argument forward, not
>> already knowing the answer. That's why I claim that post-diction
>> ("explanation") is more common than pre-diction.
>>
>> There's a fruitful interplay between pre-diction and post-diction. An
>> example I've mentioned some time ago, from our intro physics textbook:
>> When searching for an explanation for spark formation in air (we see
>> the spark and ask how it occurs, which is post-diction or explanation)
>> there are a couple of tentative explanations that one can rule out.
>> Another explanation seems to explain the phenomenon, and the validity
>> of this post-diction is greatly strengthened by noting that it (and
>> not the other explanations) pre-dicts that it takes twice the critical
>> electric field to trigger a spark if the air density is doubled, a
>> pre-diction that is consistent with observations.
>>
>> Bruce
>>
>> ============================================================
>> FRIAM Applied Complexity Group listserv
>> Meets Fridays 9a-11:30 at cafe at St. John's College
>> lectures, archives, unsubscribe, maps at http://www.friam.org
>
>
>
>
> --
> Saul Caganoff
> Enterprise IT Architect
> Mobile: [phone]
> LinkedIn: http://www.linkedin.com/in/scaganoff
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

[Steve Smith]

Nick (aka NST) -

 

One possible response to this comment might be to just tell me to piss off until I have read the article you are referring to. 

I think you caught the implications without having read the specific article that I recommended.  We are obviously wandered into territory which you have explored before... I definitely welcome your weighing in here. 

 

As argued in this article,

Which I just read.   I was previously not familiar with JASS.
    http://jasss.soc.surrey.ac.uk/12/1/9.html

    
the disassociation of predictive and explanatory power seems misguided.
    http://jasss.soc.surrey.ac.uk/12/1/9.html

 I suppose a statistical function of many observational variables could have no explanatory power beyond the many variables on which it is based, but then it would only predict what it already knew.  It would “just” be an intervening variable, and not a hypothetical construct, at all.   As we have agreed, some explanations can have “facetious” content, that is not predictive, but that content is not really explanatory, either.  Darwin certainly did not believe that Nature was a breeder who chose the better adapted individuals for breeding.
 
Quoting NST (you) from the article:
    "Modeling is the systematic deployment of the human capacity for metaphor and is central to all scientific activity. Models    don't stand or fall on their detailed verisimilitude, but on their capacity to capture the essence what is already known about a phenomenon and to generate expectations concerning what more might be discovered if the scientist were to look where the model pointed."

This point is a very key one IMO...  it is roughly what I base my own work in the development and application of Metaphor (Complex Metaphors and Metaphor Complexes) in Information/Data Visualization and Visual Analytics.  

I *think* that what we are discussing here is the role of *explanation* in Science?  I think what you are referring to as "facetious" content above (e.g. Darwin's description of Nature as an animal/human husband, selecting individuals for selective breeding...) is merely an extreme end of the use of analogy to explain.   I presume that Darwin's choice of analogy was deliberately extreme to make it as familiar as possible to the totally uninitiated.  To those already somewhat on board with the general model, I presume they *all* dispensed with the misunderstandings implied.

We should perhaps talk more about his offline as I can already hear poor Doug's eyeballs rolling in his sockets, but I would like to elaborate for/with you what I mean by Metaphor Complexes...  as they may be directly responsive to this problem of "facetious" content.   In particular, if we admit to a whole series of layers of explanation from the most fanciful but accessible to the most complete and accurate but mundanely obscure.   My interest is to build a scaffold from the most fanciful to the most mundane, or the most accessible to the least in the interest of A) Helping an individual build a mental (and possibly mathematical) model of a phenomena for themselves in the pursuit of exploration and discovery in some phenomenological domain; and B) helping said individual blaze a trail that others can follow from an accessible if fanciful explanation to a more complete and accurate if obscure (and presumably useful one).    Perhaps what I'm suggesting is to build a stack of models that span the spectrum from explanatory to predictive, fanciful to mundane.

For this I need to retain the distinction.  I would prefer to think of my Metaphor Layers as various renderings or projections of aspects of a *single* model which has *all three* Explanatory, Descriptive, and Predictive qualities.  

- SAS

 

Nick

 

PS  And isn’t  “real existence” the ultimate hypothetical construct?

 

From: [hidden email] [[hidden email]] On Behalf Of ERIC P. CHARLES
Sent: Tuesday, July 10, 2012 2:41 PM
To: Steve Smith
Cc: The Friday Morning Applied Complexity Coffee Group
Subject: Re: [FRIAM] Celebrating the Higgs - explaning and predicting

 

Steve,
Interesting paper, but I'm not sure if I follow. The basic argument seems to be that we often explain things by imagining (with the help of statistics) hypothetical constructs that cannot be directly measured. As those constructs can't be measured directly, they don't help us predict things. Thus, predictive models are limited to using things that actually exist, while explanatory models are not so limited.

That seems like a really good argument for coming up with better explanations, not an argument for distinguishing and reifying two distinct modeling tasks.

This is a topic I am quite interested in. I would presume that an ideal explanatory model would be identical to an ideal predictive model, though I grant that non-ideal cases might differ. What am I missing?

Eric



On Tue, Jul 10, 2012 12:37 AM, Steve Smith <[hidden email]> wrote:

Bruce -

I second the motion (very good post)!

Mendeleev's Periodic Chart was *my* first introduction (back when) to the very concept of having a predictive model that was (almost) entirely void of explanatory ability (as it stood when constructed).  I found the notion *fascinating* and it drove me into the field of Visual (Perceptual) Analytics many years later... seeking patterns that yield useful prediction without necessarily waiting for an explanatory model.

For those vaguely interested in the philosophical underpinnings of science, it's methods and utility, I recommend Galit Schmueli's (George Washington U's) paper on Predictive vs Explanatory Models (as well as *Descriptive* models)...

arxiv.org/pdf/1101.0891

Thanks! Glad you liked it!

I have long been bemused by the strong parallels among the various

tales I was able to tell in that post.

Bruce

On Mon, Jul 9, 2012 at 4:43 PM, Pamela McCorduck [hidden email] wrote:

Bruce, that blog post is marvelous in its simplicity and power.

Pamela

On Jul 9, 2012, at 2:39 PM, Bruce Sherwood wrote:

See my blog:

http://matterandinteractions.wordpress.com/2012/07/09/the-higgs-boson-and-prediction-in-science/

Bruce

On Sun, Jul 8, 2012 at 10:18 PM, Owen Densmore [hidden email] wrote:

Lets chat about the Higgs discovery, its likely-hood of being correct, and

the impact it will have going forward .. at the next Friam @ St Johns.

Could someone see if Hywel White is available .. or anyone you know who'd

like to hold forth on the topic!

  -- Owen

============================================================

FRIAM Applied Complexity Group listserv

Meets Fridays 9a-11:30 at cafe at St. John's College

lectures, archives, unsubscribe, maps at http://www.friam.org

============================================================

FRIAM Applied Complexity Group listserv

Meets Fridays 9a-11:30 at cafe at St. John's College

lectures, archives, unsubscribe, maps at http://www.friam.org

"Im Deutschen lügt man, wenn man höflich ist."

"In German, if one is polite, one lies."

Goethe, "Faust"

============================================================

FRIAM Applied Complexity Group listserv

Meets Fridays 9a-11:30 at cafe at St. John's College

lectures, archives, unsubscribe, maps at http://www.friam.org

============================================================

FRIAM Applied Complexity Group listserv

Meets Fridays 9a-11:30 at cafe at St. John's College

lectures, archives, unsubscribe, maps at http://www.friam.org

 

============================================================

FRIAM Applied Complexity Group listserv

Meets Fridays 9a-11:30 at cafe at St. John's College

lectures, archives, unsubscribe, maps at http://www.friam.org

Eric Charles

Professional Student and
Assistant Professor of Psychology
Penn State University
Altoona, PA 16601

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

[Bruce Sherwood]

Thanks, Dean. http://en.wikipedia.org/wiki/Eightfold_Way_(physics)
gives a brief overview of what Gell-Mann (and Ne'eman) did, and
explains that the octet and decuplet are representations of the group
SU(3). The article includes some links to additional details.

Bruce

On Wed, Jul 11, 2012 at 10:14 AM, Dean Gerber <[hidden email]> wrote:
> Excellent series of explanations, Bruce. Do you by chance have a specific
> reference to the ten-pin structure and its relation to group theory?  Thanks
> ... Dean Gerber
>

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

[Steve Smith]

Saul -

Excellent points!  

I agree that the examples I am thinking of are of a "Classification" nature.  I also like the analogy of "stamp collecting", as I think it actually captures one of the traits of human intuition that makes us as capable as we are of making sense of the universe we inhabit.   Like many birds (Corvids in particular?) and some rodents (Pack Rats most notably), we are good at spying, collecting and organizing (do our animal cousins organize their finds the way we do?) shiny, colorful and otherwise notable objects.   We notice anomolous phenomena, record it, and organize it according to various organizational schemes of which I think "classification" is the most obvious and often useful.

I think that "classification" might be described as a simplistic example of analogy making.  The target domain of the analogy being the features of the phenomena (or artifacts) and the source domain being something like simple geometric relations (lists, categories, tables, etc.).

One step more elaborate perhaps is the organization into graphical models... into describing the relations between phenomena and artifacts.  A first stab at this was made in the following work...  establishing correlation networks among the *terminology* used by experts in a field of study:

In some exploratory work I did with Deana Pennington (UNM cum UTEP) for the NSF on "Creativity in Science" we started building models of the Lexicons of researchers in collaboration.  Her study group were climate modeling scientists with surprisingly diverse uses of what was nominally the same language.  It is not surprising at all that laymen in the field are easily confused or even offended by the results coming out of that work... not because it is wrong, but because the language is not normalized.   For example, Atmospheric and Oceonagraphic scientist have huge overlaps in the phenomena they study, but typically in different regimes... so what is an important distinction to one may not be to another, and the words they choose to use describe the same mechanism or phenomena can be blatantly or subtly different.  Similar for plant biologists, animal biologists, ecologists, etc.   Same targets, similar models, different terminology.  

Part of the incidental work in the project was to try to normalize and fuse these models (as represented by the natural language used to describe these overlapping fields).   One methodology provided by Tim Goldsmith (UNM Psychology) was roughly as follows:   Interview each scientist and get a list of words that are important in their work.   Take the Union of these words  and build a correlation table of those words (rows and columns labeled by those words).   Have each scientist fill out the resulting NxN Matrix with numbers  (1-10)  roughly indicating how strongly correlated the words were.    By studying the patterns *between* these resulting matrices, a certain sense of how "distant" the various scientists were could be achieved.

This technique (I think) was developed to help understand how much learning is happening (do the same thing with an expert ... the teacher of a subject... and neophytes... the students).    As students progress in a classroom (or laboratory) setting, presumably their understanding (and the matrix of terms used) will begin to align better and better with that of the teacher.   This was first (in my presence) used to compare two methods for medical students to learn about the function of the Nephron in the Kidney.... a control group being presented with "conventional" learning materials and another group being presented with an immersive "experience" presenting the same material but in first-person context with the opportunity to *explore* the model.   The results were positive, but too many factors were involved to make any conclusive judgement... but the point was to begin to explore this as a technique for learning and learning about learning.

I personally think that "good science" is important but my own interest is in how the human intuition is used and how it can be engaged more effectively in the process of exploration, discovery, and analysis of the "real world" that we presume we live in (nod to NST's point).

- Steve

It seems that many scientific fields go through a phase of observation (derisively called "stamp collecting") followed by a phase of classification. If you're lucky then patterns can be picked out of the classification scheme to "predict" where to look for new entities or new interesting phenomena.

The Periodic Table is one of the cited examples. Another example (though perhaps not as good) is the Hertzprung-Russell diagram used in astronomy where stars are plotted onto a graph with luminosity and colour as the two axes. They form a characteristic pattern which had to be explained by any theory of stellar evolution.

I also recall many years ago picking up a book on atomic spectra published in 1901 - some 12 years before the Bohr theory of the atom - which illustrated hundreds of different emission spectra and talked about the relationships between spectral line frequencies in terms of waves and resonances. It reflected a very interesting point in the science where patterns were emerging and calling out for an explanation.

So it seems that a "classification" model can be used to make "predictions" - to see if the pattern extends to unobserved areas - and that this can be independent of an underlying explanatory theory. I think Gell-Mann's QCD models probably fit this idea. The image of the "ten-pin owling skittles" pattern and the mystery of what lies at the tip is very evocative.

Regards,
Saul

On 11 July 2012 06:56, Bruce Sherwood <[hidden email]> wrote:

"For Engineers perhaps, predictive models are sufficient, they may not
be (very?) interested in explaining *why* a particular material has
the properties it does, merely *what* those properties are and how
reliable the properties might be under a variety of conditions."

I don't think this currently true. A big chunk of what used to be
labeled "physics" is now in academic engineering departments with the
name "material science". This consists of exploiting models that
explain observed properties of materials, with the goal of looking for
opportunities to change parameters to get improved behavior. In the
early 1990s I heard a talk by an engineering professor at the science
museum in Toronto, where he explained how such research had led to
concrete many times stronger than it had been, and that the iconic
tall tower in Toronto could not have been built not many years before
it was built, as it relied on much stronger concrete.

In some cases someone sees how, starting from fundamental physics
principles, one can predict that such and such should happen or be. In
other cases an observed phenomenon gets explained in terms of
fundamental physics principles (post-diction), which then suggests how
changes in the situation might yield an improved behavior. Pre-diction
and post-diction both require a deep understanding of how to go from
underlying fundamental principles to the behavior, but pre-diction in
addition requires the imagination to run the argument forward, not
already knowing the answer. That's why I claim that post-diction
("explanation") is more common than pre-diction.

There's a fruitful interplay between pre-diction and post-diction. An
example I've mentioned some time ago, from our intro physics textbook:
When searching for an explanation for spark formation in air (we see
the spark and ask how it occurs, which is post-diction or explanation)
there are a couple of tentative explanations that one can rule out.
Another explanation seems to explain the phenomenon, and the validity
of this post-diction is greatly strengthened by noting that it (and
not the other explanations) pre-dicts that it takes twice the critical
electric field to trigger a spark if the air density is doubled, a
pre-diction that is consistent with observations.

Bruce

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

--
Saul Caganoff
Enterprise IT Architect
Mobile: [phone]
LinkedIn: http://www.linkedin.com/in/scaganoff

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org