jump to navigation

Gomory on Research, Industry, and National Competitiveness July 30, 2010

Posted by Will Thomas in 20th-Century-Science Historiography.
Tags: , , ,
trackback

Click for the Ralph Gomory profile at the IBM archives

One of my activities on my recent blogging hiatus was an oral history interview with Ralph Gomory.  The interview was originally instigated as part of the AIP History Center’s History of Physics in Industry project, on which I’ve helped out here and there.  Our discussions with researchers at IBM all pointed to Gomory as a crucial figure in that company’s history.  Personally, I had a strong interest in the interview, because Gomory’s background is in mathematics, and he is a notable figure in the operations research (OR) community, primarily on account of his foundational work on integer programming.  (For those keeping track, I wrote my dissertation, and am currently polishing up a book manuscript, on the history of certain sciences of policy analysis, including OR.)  This post is mainly based on the background research I did ahead of the interview.

Gomory was director of research at IBM from 1970 to 1986.  IBM Research had been established in its present form in the late 1950s by Emanuel Piore.  Piore had spent much of his postwar career at the Office of Naval Research, culminating in a stint as Chief Scientist.  Careful readers of Zuoyue Wang’s recent book on the President’s Science Advisory Committee (to be discussed on this blog presently) will know that Piore became a ubiquitous figure on various high-level government panels (i.e., though not well-known to historians, he was a big deal).

The idea behind establishing IBM Research was the general sense, widespread in the 1950s and ’60s, that technologically-oriented companies would be well-served by conducting their own basic research.  Piore’s goal was to establish an environment — housed in a modern building designed by Eero Saarinen — where researchers could freely explore their own ideas.  Gomory had originally been brought in to be part of the new mathematics department (along, incidentally, with fractal geometry pioneer Benoît Mandelbrot).

Now, going back to my previous post’s interest in basic research and the “linear model” in history: once one had established the importance of the link between research and technological development, one was faced with a series of subsidiary questions, to which one would have devoted more or less thought. At what level should this research be funded, overall?  What sorts of organizations should house research activities?  In what ways, and to what degree, should research activities be connected to, or liberated from, organizational (or simply others’) goals?  Of all possible specialties, what sorts of specialists should particular organizations hire?  According to what criteria should organizational managers initiate, discontinue, prioritize, and fund competing research projects?  Answers to these questions necessarily depended on more specific notions of the importance and character of research activities and their connection to technological work.

Gomory’s long term as director of research was, in many ways, centered around an attempt to better integrate IBM Research’s to-that-point freestanding activities into corporate strategy.  He became convinced that the best opportunities for research contributing to IBM’s products were generally limited to very certain points in the product development cycle (the succession of generations of products).  Developers and manufacturing engineers were often better positioned to offer judgment on what research results would prove the most useful to them, even as their ability to do so was predicated on researchers effectively communicating potential implications of their work to the engineers.  IBM Research, meanwhile, would continue to pursue open-ended academic-type work — including Nobel Prize-winning work — but the research division also began to concentrate more on problems suggested by difficulties and challenges foreseen in the product development and manufacturing processes.

In the 1980s and ’90s, at IBM, and then as president of the Sloan Foundation, Gomory began to publish short articles about the management of research, initially concentrating on the problem of the Japanese challenge to American competitiveness.  IBM, of course, experienced first-hand the threat from surging Japanese electronics firms.  Gomory’s articles responded to what he believed were misguided appeals to American underinvestment in research and science education as explanations for the challenge.  For example, some commentators, particularly academic researchers, were likely to point to large Japanese research investment and state projects, notably those at MITI (now METI), as a key source of the challenge.  (Incidentally, the international polemical/political arc leading from DSIR to MITI would be well worth tracing.)  Gomory preferred to point to particular Japanese methods of integrating design, manufacturing, and marketing, and their contraction of the product development cycle, to explain their successes.

Accordingly, while Gomory supported funding American basic science to maintain competitiveness in all fields, he argued that it was unlikely to make a substantial contribution to pressing problems of national economic competitiveness.  He attributed the idea that it could to what he referred to as the “ladder of science” model (essentially the linear model).  He asserted that whatever advantages might accrue from success with that model were fleeting as new industries based on novel technologies were quickly replicated in other nations.  Most economic advantage and long-term success was grounded in large industries’ ability to put low-cost, high-quality products through the development cycle more rapidly than their competitors.  Academic researchers were even less likely than industrial researchers to know what research results could be fruitfully applied in the cycle.

One lesson we could draw from the arc at IBM Research from the ’50s to the ’80s is a progression in ideas from Piore to Gomory.  This would map well onto existing narratives detailing a widespread questioning of the wisdom of unalloyed support for research in the 1960s and ’70s, which has been linked to a decline of the perceived validity of the linear model more generally.  The classic example is the increased Congressional questioning of military support for university research, punctuated by the Defense Department’s mid-1960s “Project Hindsight”, a study that failed to find a substantial link between advances in military technology and investment in research.

Framing this story in terms of the rise and fall of the linear model makes sense, because it renders a rationale for the support for research as a path to technological and economic prowess.  However, my own preference (and I think this mainly accords with David Bruggeman’s suggestion for thinking of the linear model as “incomplete”) is to think of a sort of undefined virtue as having been attributed to research, with little further reflection being given to problems such as who should be responsible for supporting research, and what institutional frameworks best mediate between university research, industrial research, and technology development communities.

This could all reduce down to “I say po-tay-to, you say po-tah-to”, but my feeling here, also expressed in my previous post, is that doing away with the historical idea of a linear model frees us up to look at, and evaluate the relative significance of the history of other rhetoric, other ideas, and other practices.  For example, while it seems likely that high-level managers and policymakers were convinced to support research perhaps out of some vague notion that it would yield occasional windfalls, this support would likely have been disconnected from their management or policymaking regarding technology development activities, which do not seem to have been substantially chained to any linear model, even though these activities were ostensibly a part of it.

From this perspective, the focus in the IBM Research narrative can be detached from Gomory’s reforms and criticsm; instead those reforms and criticism become an invitation to look at the significance of the history of company management, technology engineering, and marketing, and its relationship to scientific research for what it was, rather than for what it failed to be.  Fortunately for us, the house history of computing leaves us with a good head-start in the case of IBM, and studies such as Christophe Lécuyer’s of Silicon Valley, or Joan Bromberg on the joint-history of quantum optics and the laser industry, give us a good look at what a more integrated history would look like (although one should note both these cases focus on novel industries).

Indeed, we have a good start on a long-term historiography of these areas, as early modern technology-knowledge confluences in areas such as naval architecture, waterworks, practical medicine, and chemical dyes have found historians’ interest, and there is also good work in later periods on topics such as metrology, telegraphy, and forestry.  The difficulty, as ever, is broader survey and synthesis.  To develop what I would view to be a satisfying historiography, it is not enough to say, “but so-and-so did their case study of X, which amply demonstrates the historical connection between science and technology; how can you say there is not a sufficient historiography on the matter?”

The point is, we need to find better ways to talk about developments and trends en masse, to get out of the “view from the archive folder”, to deal not with just the actions of a single committee, for instance, but to describe how the work of thousands of committees coordinated the scientific and technological world.  I am convinced that if this happens, the historiographical importance of things like some “linear model” will start to seem very odd in retrospect, that, somehow, we became distracted by a few snippets of rhetoric that, while prominent and even influential in some respects, can only be properly evaluated amid a much larger, and more complex context.  From this view to focus historiography on a few items of rhetoric would be to make the same mistake of incompleteness as those who deployed that rhetoric in the first place.

Advertisements

Comments»

1. David Bruggeman - August 2, 2010

Gomory’s comments about the ladder model, combined with some remarks on the NSF’s Science of Science Policy listserv, suggests to me that economists probably have a fair amount to do with the entrenchment of the linear model in thinking and in policy following World War II. History of economic theory (particularly with how it treated technology) covering the postwar period might be useful here.

2. Will Thomas - August 2, 2010

My suspicion is that there’s a division between the economics of science and technology, on one end, and sort of naive academic opinionating along the lines of “science and technology are very important what with spin-offs and stuff, more funding please” on the other. Gomory’s articles were mainly designed to combat naive opinionating — he also had a couple of barbs about calls for business schools to start teaching “exotic new corporate culture”. His most immediate aim seems to have been to alert industrial managers to under-discussed issues, especially the need to shorten the product development cycle.

On possible connections to the history of economic thought, I know Phil Mirowski has written some about it, but don’t know how far the historiography goes beyond that. Per Edgerton, it is in these economic/policy studies discussions that the “linear model” seems to be invoked as some sort of theoretical ur-state of ignorance, but it is not clear that serious modelers ever subscribed to it. As near as I can tell, economists were invoking select bits of rhetoric to give the theoretical ur-state a historical reality. I know it’s been good sport to criticize the lack of realism in economic models of research, but I’d like to learn more about what specific ills it would be possible to hold them responsible for: particular policies, or just giving comfort to naive opinionators (which seems to be the usual beef with economists).

On a related side note, just today, on one of my occasional browses through the Physics Today website, there’s a new article about STAR metrics (I don’t think there’s a paywall, but I can’t tell for sure when I’m at work) and the pressure to apply metrics in the face of inquiries relating to funds received from the American Recovery and Reinvestment Act.

Finally, as far as historians are concerned (so not necessarily directed at you, David, but more as addendum), the entrenchment or persistence of certain kinds of theorizing and rhetoric should not necessarily make the history of those theories or rhetoric the central focus of investigation when many alternative lines of thought have existed. Edgerton was emphatic on this point in his “Linear Model” piece:

“We could, if we wished, label thesse obviously self-interested claims [academics’ exaggeration of the importance of their work for technological and economic development] the ‘linear model,’ but to call the propaganda of academics ‘the linear model’ is to flatter the claims, and to avoid stating the obvious: that these are generally claims by academic researchers for the power of academic research. To call it a ‘linear model’ also runs the risk, especially in the context of the current use of the term, of smuggling in the assumption that what academic research scientists said about innovation was the most influential discourse of innovation around. In other words, to believe that ideas about innovation were created by academic research scientists, and diffused out to engineers, to government, to industry and to the public [is] to believe in a linear model not of innovation, but of ideas about innovation.”

Later: “…it was convenient to invent labels for naive positions influentially peddled in the public sphere by scientists and engineers, and found in particular in the views of science and engineering undergraduates taught by STS and other academics. The very failure of STS to define the public discourse around science and technology doomed us to attacking public science. Instead of engaging with the arguments of critics of naive positions, the naive conceptions still had to be attacked. These naive conceptions created by academics thus achieved greater prominence than they would otherwise have had.”

This line of historiographical thought is very influential for my own thinking. By assuming the historical importance of certain persistent positions, generally as a way to explain inadequate policy outcomes, do historians reinforce the idea that these positions were of unusual historical importance? This line of thought could write the importance of alternatives ideas out of history (what Edgerton calls “anti-history”), possibly as a way of increasing the cogency of our own historical work among non-historians.

This leads into a thought I’ve started to entertain more seriously recently: is the academic alliance between history of science and STS a massive conflict of interest?

3. David Bruggeman - August 3, 2010

Thanks, Will. I try to be mindful that I’m not approaching these issues from an historical or historiographical perspective, but that doesn’t always come across.

I was trying to suggest that the economic models of Simon and his colleagues, particularly with how they treated technology, could be used to support the naive opinionating. That’s a separate question from whether Simon and colleagues were aiming to support such opinionating. I need to be more careful about distinguishing between the two questions, as they have distinct implications. FWIW, I’m more interested in the former.

I am curious about your parting thought, Will. Perhaps it’s worthy of a separate post?

4. Will Thomas - August 3, 2010

Aside from whether economic models could be used to support opinionating, and whether they are intended to be used to support opinionating, I would add a third (sociological) question: are explicit models really important in sustaining the legitimacy of certain opinions?

Say, if naive models really were dispensed with, or at least effectively quarantined somehow, would this affect the legitimacy of policy positions we view as distracting or dangerous? Or (and this again reflects Edgerton’s argument above), do we simply presume their importance in securing legitimacy, because their (historical or present) importance justifies reflection and publication on them?

This sort of gets at the possible HoS-STS conflict of interest, which I will hopefully flesh out in a post before too long. I sort of get at it here and here, but at the time I was viewing it as more of a quirk in the history of historiographical ideas, and less as an outright conflict of professional interest, which was a thought spurred by some recent reading in preparation for this latest series of posts.

Are you talking about Herb Simon, by the way? If so, he’s of interest in my work, and if you could pass along the references for the pieces you have in mind, I’d be grateful.

Thanks!

5. David Bruggeman - August 5, 2010

Yes, I was referring to Herb Simon. A couple of items that might be of particular use to you (assuming you haven’t already seen them) are:

“Effects of Technological Change in a Linear Model”, 1951, in Koopmans, editor, Activity Analysis of Production and Allocation.

“The Linear Decision Theory for Production and Employment Scheduling”, with F. Modigliani and C.C. Holt, 1955, Management Science.

Robert Solow’s work is also important here. His growth model that suggested 4/5 of economic growth came from technical progress could have fed into notions that more science => more tech => more growth. This is the usually cited article as the genesis of the model.

Solow, R.M., 1957. Technical change and the aggregate production function. Review of Economics and Statistics 39, 312–320.

6. Will Thomas - August 5, 2010

Excellent, thank you! Actually, this is right up my alley, and I’m sure I have run across it, but it wouldn’t have jumped out at me as far as this context is concerned, because I would have been looking at the history of linear programming (which, in the ’40s and ’50s, was brought into economics as an extension of Wassily Leontief’s input-output economic models). There’s starting to be some good work on the relationship between operations research, management science, linear programming, and economics. Judy Klein, notably, is doing some work on the Carnegie Tech crowd, like Modigliani. I will look into this very soon.

7. Will Thomas - August 6, 2010

I had a chance to glance over the Simon piece, and it has nothing to do with the “linear model” being discussed here (though, weirdly, it does have a bit to do with the “integer programming” that Gomory pioneered, and that is discussed at the beginning of this post). For those interested, the Simon piece is available online via the Cowles Foundation — as Paul Krugman would say, it’s wonkish.

So, best to start with Leontief, a Russian economist who ended up at Harvard. An input-output model (posited by Leontief in the ’30s) essentially imagines the economy as a series of flows. The output of one industry (say steel), feeds linearly into other industries (automotive, construction, etc.), ultimately feeding consumer demand, but, labor, too, is an input, so it’s all interconnected (the connections are linear, not the economy). An economy can have a higher or lower output depending on how the parts of this system are arranged.

“Linear programming” problems (we’re in the late ’40s and ’50s now) are designed to find optimal solutions to problems of just this structure: how do you arrange or allocate resources in such a way that you optimize an output (or minimize an input)? Koopmans’ massively influential “Activity Analysis” conference, which took place in ’49, was the first time that linear programming ideas were applied back onto an economy of inputs and outputs.

If you’re an economist, one useful assumption is that, through profit seeking, an economy will tend to organize itself in such a way as to achieve something like a most efficient allocation of its resources — this is “equilibrium”. However, economies change. In the 1930s, economists had begun to put serious thought into questions such as what happens in growing economies, i.e. “dynamic” economics. Joseph Schumpeter (who, by the way, was at Harvard in this same period, beginning in the ’30s) was a big-time advocate for understanding how technological change creates economic growth.

Among those who know about Schumpeter at all, his name is pretty much synonymous with the two-word phrase “creative destruction”. (If one reads his Capitalism, Socialism and Democracy there’s a good chance you’ll glance over his extensive history of economic ideas, particularly Marxist ideas, to get to the juicy bits about creative destruction.) OK, so it’s the Depression, and automation is widely blamed for de-skilling labor and for unemployment. Schumpeter’s argument is that technology destroys jobs, yes, but it frees skills and resources to be allocated in more productive ways.

So, what Simon is doing circa 1950 is building a model of this process. Basic science has nothing to do with it. We’re mostly talking about piecemeal improvements in industrial processes. But, the idea is these technological improvements free up resources to be reallocated in more efficient, more productive ways. Linear programming was, for obvious reasons, a remarkable opportunity to anticipate how economic inputs and outputs would redistribute themselves, leading to economic growth.

I would also stress that the discussion was exceedingly exploratory, theoretical, and addressing particular sets of academic concerns.

Incidentally, if one wants to understand the contours of current economic debates, brushing up on this history is at least a little helpful. Although there was nothing innately free-market about input-output models (indeed, linear programming,as I understand it, was seen as a boon to communist economic planning), the ideas that monetary policy and tax cuts are preferable to fiscal policy, and that fiscal stimulus will prevent the necessary structural rearrangement of America’s economic resources, are connected to these early discussions about resource allocation, and the economic drags created by “artificially” fixing some elements of these resources (as through federal spending on specified projects) rather than simply making more liquid resources available to the system as a whole.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s