OR Forum Paper on the State of Operations Research
Mike Trick
Research. Written by ManMohan Sodhi and Chris Tang, it is entitled
“The OR/MS Ecosystem: Strengths, Weaknesses, Opportunities, and
Threats” and analyzes the state of the field, with a particular
emphasis on the relationship between academia and practice. Check out
the paper and commentaries, and provide your own thoughts on the
issues at http://orforum.blog.informs.org (disclaimer: I am the Area
Editor of the OR Forum).
One bittersweet aspect: Mike Rothkopf provided a commentary one week
before his death where he talks about the how academia and industry
enrich each other. I wish he was here to continue the conversation.
Mike Trick
-------------------------------------------------------------------------------------------------------------
Michael Trick Tepper School of Business,
Carnegie Mellon
tr...@cmu.edu http://mat.tepper.cmu.edu/blog
-------------------------------------------------------------------------------------------------------------
A.L.
<mtr...@gmail.com> wrote:
>There is a new paper in the OR Forum area of the journal Operations
>Research. Written by ManMohan Sodhi and Chris Tang, it is entitled
>“The OR/MS Ecosystem: Strengths, Weaknesses, Opportunities, and
>Threats” and analyzes the state of the field, with a particular
>emphasis on the relationship between academia and practice. Check out
>the paper and commentaries, and provide your own thoughts on the
>issues at http://orforum.blog.informs.org (disclaimer: I am the Area
>Editor of the OR Forum).
Interesting article, The issue "academia vs. practitioners" pops up
from time to time, including the last INFORMS Practice Conference.
My personal view:
1. INFORMS is by academics for academics. As individual paying
membership fee from my own pocket, I don't see any value that I would
get in return. This is the reason why I stopped paying some time ago.
If somebody has any doubts, I suggest to visit INFORMS "national
conference" in fall. This is sandbox for graduate students desperately
building publication record needed by administration to defend their
thesis. This is the ONLY conference I know that doesn't have any
selection/rejection process
2. Academia has different goals than practitioners. The most important
issue for the academia is publishing. As practitioner, I have shredder
instead of garbage basket. I cannot publish and present papers on
conferences. If we spend million bucks to develop new algorithm, we
will not run to next INFORMS conference or write paper to give this
algorithm to our competitors for free. NDA will not help. It is prety
easy to develop derived work not covered by NDA that will be published
3. Academic grants are not cost effective due to high administrative
overhead and because these monies are used for all possible things,
not necessary connected with research,
4. Academia develops models and have problems with distinguishing
models from reality. Practical problems are considered either
unsolvable or "not scientific", and models have features that make
them practically useless. Example is VRP, when transportation cost is
consider linear. In real problems it is not linear, it is calculated
using very complex algorithms, and function that describe cost is not
ever written on paper - cost is computed by software licensed from 3rd
parties
5. Academia operates in different time scale than practitioners. When
I presented some problems to my colleagues from the academia, I got 3
years as estimate of time needed to solve the problem. In industry,
for the same problem we had 8 month to find solution, productize the
solution and deploy
6. In practice, publication is not the final proof of correctness or
quality of work. In academia, it is.
7. Academic solutions don't scale. Example are Solomon's tests for VRP
whose size is order of magnitude smaller the size of practical
problems and time needed to solve these tests is order of magnitude
longer than needed in real world
I could continue. As practitioner and former academic
A.L.
Samik R.
> On Wed, 23 Apr 2008 19:45:13 -0700 (PDT), Mike Trick
> <mtr...@gmail.com> wrote:
>
>> There is a new paper in the OR Forum area of the journal Operations
>> Research. Written by ManMohan Sodhi and Chris Tang, it is entitled
>> “The OR/MS Ecosystem: Strengths, Weaknesses, Opportunities, and
>> Threats” and analyzes the state of the field, with a particular
>> emphasis on the relationship between academia and practice. Check out
>> the paper and commentaries, and provide your own thoughts on the
>> issues at http://orforum.blog.informs.org (disclaimer: I am the Area
>> Editor of the OR Forum).
>
> Interesting article, The issue "academia vs. practitioners" pops up
> from time to time, including the last INFORMS Practice Conference.
>
> My personal view:
>
> 1. INFORMS is by academics for academics. As individual paying
> membership fee from my own pocket, I don't see any value that I would
> get in return. This is the reason why I stopped paying some time ago.
> If somebody has any doubts, I suggest to visit INFORMS "national
> conference" in fall. This is sandbox for graduate students desperately
> building publication record needed by administration to defend their
> thesis. This is the ONLY conference I know that doesn't have any
> selection/rejection process
>
think, to get to meet with other like-minded practitioners at a neutral
environment (in the sidelines of the conference), and discuss about your
work, get feedback and new ideas. Sometimes, a few presentations can
also provide with a few clues. Presenting is probably not the primary
goal of practitioners at these conferences. I think that is a nice
opportunity, since OR practitioners tend to have a geeky aura in typical
companies, mostly work alone, and the boss is typically interested only
in the result obtained after solving a problem, not in the methodology
(used couple of GUBs for solving this MIP). This is quite a bit of
generalization - but not very far from truth.
A.L.
wrote:
>>
>The only value of INFORMS annual conference as a practitioner is, I
>think, to get to meet with other like-minded practitioners at a neutral
>environment (in the sidelines of the conference), and discuss about your
>work, get feedback and new ideas. Sometimes, a few presentations can
>also provide with a few clues. Presenting is probably not the primary
>goal of practitioners at these conferences. I think that is a nice
>opportunity, since OR practitioners tend to have a geeky aura in typical
>companies, mostly work alone, and the boss is typically interested only
>in the result obtained after solving a problem, not in the methodology
>(used couple of GUBs for solving this MIP). This is quite a bit of
>generalization - but not very far from truth.
Agree. Networking is the issue. But in the era of Internet,
face-to-face meetings are not longer critical
A.L.
Paul Rubin
> 1. INFORMS is by academics for academics. As individual paying
> membership fee from my own pocket, I don't see any value that I would
> get in return. This is the reason why I stopped paying some time ago.
> If somebody has any doubts, I suggest to visit INFORMS "national
> conference" in fall. This is sandbox for graduate students desperately
> building publication record needed by administration to defend their
> thesis. This is the ONLY conference I know that doesn't have any
> selection/rejection process
Interesting. I'm a pure academic and former (way former) TIMS member
who rejoined INFORMS at this past fall's conference because I saw what I
thought was more relevant stuff presented there than what I was getting
at the Decision Sciences Institute conference, my "home" for the past
25+ years. You may be right about INFORMS being populated primarily by
academics, but at least the military is represented by practitioners
(although lately I've seen some of them at DSI too, I think).
As far as the lack of a rejection process for papers, that's one reason
I quit way back -- I sat through too many sessions where the speaker had
a few index cards and half an idea what he wanted to talk about.
Unfortunately, the selection process at other conferences is becoming
more and more nominal (rejection rate approximating the fulfillment rate
for campaign promises, if you know what I mean). Happily, I didn't sit
through a single content-free session at the Seattle meeting. Now if
the presenters would just learn to embed the math fonts in their PDF and
PPT files ...
>
> 2. Academia has different goals than practitioners. The most important
> issue for the academia is publishing.
Absolutely true. Publishing something useful is nice but not essential.
:-)
> As practitioner, I have shredder
> instead of garbage basket. I cannot publish and present papers on
> conferences. If we spend million bucks to develop new algorithm, we
> will not run to next INFORMS conference or write paper to give this
> algorithm to our competitors for free.
True, but if you're ILOG or Dash, you'll run to the next INFORMS
conference and present a paper bragging on the capabilities of your product.
> NDA will not help. It is prety
> easy to develop derived work not covered by NDA that will be published
>
> 3. Academic grants are not cost effective due to high administrative
> overhead and because these monies are used for all possible things,
> not necessary connected with research,
You're saying paying salaries for a few extra university administrators
is not directly connected to the grant??? Correct again -- and this is
a sore point with many of the faculty getting the grants.
>
> 4. Academia develops models and have problems with distinguishing
> models from reality.
True in *some* cases. Many of us are quite capable of distinguishing
model assumptions from reality; we just can't be bothered, particularly
if reality makes the math difficult. I came across a book on modeling a
while back -- can't recall the title -- that had a chapter titled
"Assume a Spherical Chicken", which I think says it all.
> Practical problems are considered either
> unsolvable or "not scientific", and models have features that make
> them practically useless. Example is VRP, when transportation cost is
> consider linear. In real problems it is not linear, it is calculated
> using very complex algorithms, and function that describe cost is not
> ever written on paper - cost is computed by software licensed from 3rd
> parties
I'll buy the part that academics will shy away from real-life problems
as being too difficult/unsolvable in some cases. The flip side is also
true: some of us will overstate the actual difficulty of solving a
problem in order to justify some new heuristic. The "not scientific"
part I'm not so sure about. I do think there's an issue with generality
-- some academics may shy away from a very specific (quirky) instance of
some problem fearing that the solution process will not be general
enough to be publishable.
All that said, though, I thought I saw some rather relevant research
presented at INFORMS. As just one example, Eva Lee presented a paper on
methods for improving diagnostic scans (I forget the details) that
seemed highly applicable to me.
>
> 5. Academia operates in different time scale than practitioners. When
> I presented some problems to my colleagues from the academia, I got 3
> years as estimate of time needed to solve the problem. In industry,
> for the same problem we had 8 month to find solution, productize the
> solution and deploy
Some of this may be the definition of "solution". I suspect that fairly
standard practice in industry is to provide the best possible (not
necessarily optimal or ideal) solution in the allowed time, whereas many
(not all) academics are probably thinking in terms of an optimal
solution. (I'm using "optimal" in a somewhat generic way, not
necessarily assuming the problem is an optimization problem.)
>
> 6. In practice, publication is not the final proof of correctness or
> quality of work. In academia, it is.
>
> 7. Academic solutions don't scale. Example are Solomon's tests for VRP
> whose size is order of magnitude smaller the size of practical
> problems and time needed to solve these tests is order of magnitude
> longer than needed in real world
True, although in partial defense of my colleagues (and myself) we often
view our output as not the final word on a problem but building blocks
toward eventual solution of the problem.
I would like to point out, though, a non-virtuous cycle here. If
practitioners don't come to INFORMS (and other conferences) because they
don't see any value, we academics will have that much less contact with
reality, which will lead us to stray further into the parallel universe
of convenient assumptions, which will further deter practitioners from
playing with us ... I think this means all the academics are converging
toward being economists, but I'm not positive. Anyway, I'd like to
spend a little time around practitioners just to see how the other 90%
lives. Unfortunately, the INFORMS Practice conference is priced well
beyond my means.
/Paul
Gordon Sande
The relevant quote attributed to John Tukey (now deceased) is that
"An approximate answer to an exact problem is often more useful
than an exact solution to an approximate problem."
Tukey was fully capable of producing exact solutions to the various
approximations to a problem and used those to provide insight into
the approximations to the exact problem. He just did not flaunt the
immense collection of powerful mathematics that he kept hidden in
his back pocket.
Mike Trick
>
> Agree. Networking is the issue. But in the era of Internet,
> face-to-face meetings are not longer critical
>
> A.L.
I used to think that, but I am no longer sure. I find that one's
Internet persona is misleading, and that there still is no substitute
for face-to-face. Perhaps Web 4.0 will fix all that.
I would hate to judge everyone by how they appear on the Internet in
its current incarnation.
Mike
A.L.
<mtr...@gmail.com> wrote:
>On Apr 24, 9:47 am, A.L. <alewa...@zanoza.com> wrote:
>
>>
>> Agree. Networking is the issue. But in the era of Internet,
>> face-to-face meetings are not longer critical
>>
>> A.L.
>
>I used to think that, but I am no longer sure. I find that one's
>Internet persona is misleading, and that there still is no substitute
>for face-to-face. Perhaps Web 4.0 will fix all that.
>
>I would hate to judge everyone by how they appear on the Internet in
>its current incarnation.
>
I didn't say "is not needed" but I did say "is not critical".
Especially if we consider the cost of participating in INFORMS
conferences, what was also mentioned by others
A.L.