::: about
   ::: news
   ::: links
   ::: giving
   ::: contact

   ::: calendar
   ::: lunchtime
   ::: annual lecture series
   ::: conferences

   ::: visiting fellows
   ::: postdoc fellows
   ::: resident fellows
   ::: associates

   ::: visiting fellowships
   ::: postdoc fellowships
   ::: resident fellowships
   ::: associateships

being here
   ::: visiting
   ::: the last donut
   ::: photo album

::: center home >> being here >> last donut? >> underdetermination

21 March 2009

Kyle Stanford is visiting the Department of History and Philosophy of Science this term. He is the expert in the topic of underdetermination in philosophy of science. So we thought we should seize the moment and organize a workshop on the topic with Kyle at its core. So was born the "Kylefest," as it instantly came to be know in all our internal documentation. Kyle and I formed a little organizing committee and started to solicit speakers.

The idea behind the underdetermination of theory by evidence is simple enough at a broad level. It is just the idea that, in science, there is a gap between theory and evidence. To use the most familiar example, if you plot finitely many points (the data) on a sheet of paper, you can then draw unimaginably many curves (the theories) through them. The views start to separate as soon you ask what that means. Those who like underdetermination see this as an illustration in miniature of a greater problem whenever we seek to pair data and theory. The one set of data lives comfortably with many theories. Those who are averse to underdetermination take comfort in the fact that we don't draw just any line through the points; in practice we call up things like regression algorithms to tell us just which is the right curve to draw.

The community separates along these lines. I belong to the latter group. I see underdetermination as a bump in the road of science. The passage of science may be rockier when the science is newer and more esoteric, but generally these bumps get smoothed out as we get smarter in our analyses. Why, I wonder, do my colleagues spend so much time worrying about it? On the other side is the view that this sort of underdetermination is pervasive and unavoidable. We see it most prominently in newer theories; but it affects them all. Our scientific consensus does stabilize. That creates an illusion that the evidence has determined the right theory. But it is only an illusion since our choice of theory is conditioned by other factors many of which are invisible to us. Our workshop, we hoped, would be a productive meeting of the two sides. I was looking forward to seeing both sides laid out in their many variant formulations and to advancing the discussion in the comfortable context of a small workshop.

It was fitting that Kyle should be the first speaker at the Kylefest. He began in a standard way for those who support underdetermination. He reminded us of the doctrine of scientific realism, which holds that our best theories are approximately true. It is threatened if underdetermination prevails and we cannot be assured of the truth of our best theories. A novelty of Kyle's work is the notion of "unconceived alternative." It labels the gap between evidence and theory in way that proves remarkably convenient when we need to capture elusive thoughts about underdetermination in words.

If you were a good Newtonian a hundred years ago, you may have thought that all the evidence has given us Newton's mechanics as the one, true theory. But that is only because the alternatives that ruined Newton, relativity and quantum theory, remained unconceived. In principle a similar danger threatens all our theories today and forever.

The Newtonian example does not move me. I think Newtonians had overwhelming inductive evidence for their theory. They took a small inductive gamble in accepting it and, famously, lost. But don't we mostly win these gambles? Don't we mostly win when we turn to much more prosaic notions? How could there be any confusion, say, that fossils are just mineralized once-living organisms? How could it be doubted? They are perfect pictures in stone of the beings that engendered them.

That example is the case study to which Kyle turned. His point was to illustrate that even in apparently simple cases, all sorts of alternatives may lurk in the neighborhood. In the sixteenth century, when serious discussion of fossils was engaged, it was anything but clear that fossils were mineralized, once-living beings. As Kyle ran through the way things seemed in sixteenth century, I could feel my gestalt switching. Then they did not think as we now do that organic forms can only arise in nature as the result of very specific biological process involving DNA molecules. The processes that generated these forms were more varied and may produce objects that more or less resembled perfected living forms. Why could they not also be working in the ground? And then came the combination laugh line-coup de grace, delivered with just the right level of gesturing. And if you think that these forms were produced by ordinary biological processes, "how did they get into the rocks?!" Kyle's finger poked at an imaginary rock hovering in the space between us.

We were on schedule. The arch-defender of underdetermination had just shaken my certainty in some established piece of science. I was imagining the room polarizing into the yeas and nays with an invisible line that snaked down the center of the room. And then Kyle turned everything on its head. That was then, Kyle told us. Now is now. Underdetermination no longer prevails. It has been settled by the science of taphonomy. In it we have been able to observe in the lab the very processes that generate fossils. We have done so at such a level of detail that the most minute accidents of fossilization are now intelligible to us as the result of the mineralization of the once-living.

Kyle wrapped up this new turn in more general terms. The older inductive support for the nature of fossils was abductive: we inferred to the best explanation. Such inferences are always at risk of defeat by unconceived alternatives. The modern analysis employs "projection," that is, modest, incremental, generalizations that are not prone the same danger. And that, Kyle informed us, is a great irony. Abduction was introduced and celebrated as remedying the defects of projective induction.

Kyle's presentation set the tone for the remainder of the conference. There was no invisible line snaking around the room. No one mounted an unreserved defense of rampant underdetermination in science; just as no one was willing to urge that scientific evidence always has heroic powers of determination. As the workshop and the discussion unfolded, critics of underdetermination, myself included, were cautious in their claims; and support for it was at best mixed.

My hope is that the participants will forgive me if I do not describe in any detail what they told us. We have set aside a section of philsci-archive.pitt.edu for them to post their papers. David Harker turned our attention back to what lies behind the underdetermination debate. He reviewed a version of scientific realism, "selective realism," designed to overcome the difficulties that face more traditional forms of the doctrine. Dana Tulodziecki's presentation centered on the John Snow's famous analysis of cholera. She used it to undermine the idea that the relationship between theory and evidence is the simple hypo-thetico deductivism commonly assumed when arguments for underdetermination are advanced. Snow, she found, was consciously and explicitly using principles of causal discovery of the type we now associate with Mill.

My own talk elaborated on Dana's themes. The idea of an assured underdetermination of theory by evidence can only persist because its proponents do not look seriously at accounts of inductive inference. They are sufficiently poorly developed to sustain that thesis or any other like it.

John Manchak then reviewed his results in spacetime theories in which, he showed, a generic failure of our observational past to fix our future. The final talk of the day was by P.D.Magnus. He picked up on the question of whether observationally equivalent theories are merely notational variants of the same theory, and so useless for establishing underdetermination. Following the trend, P. D. took a middle course, offering both sides as a thesis and antithesis, concluding that no rule could tell you assuredly which side to adopt.

We reassembled the following morning at 9 am. We'd advertised "coffee" on the program, having found in the past that Sunday morning breakfasts can be poorly attended. Many participants will have been up late the previous night, no doubt in earnest discussion of the grave matters raised in the workshop. When I arrived shortly after 9 am, there were already many hardy participants in the lounge, where, I was pleased to see, our prescient staff had arranged not just coffee and tea, but some pastries as well.

It was left to Nick Rescher, our last speaker, to draw the threads together. His talk took a broader perspective, seeking to identify the many different forms in which underdetermination manifested. As he spoke, I began to sense that he was probably more sympathetic to a general form of underdetermination than any of the previous speakers. One insightful remark seemed to capture it. We know, he reflected, that present science leaves future science underdetermined; and we know that future science will correct present science. So we cannot expect truth from our present science. It was a remarkably clear image and invited the same question from both me and Jim Bogen. What are we to make of the many text book examples of assured science, well removed from the uncertainties of our latest knowledge?

Nick has a kindly and magisterial presence and I could see him thinking that this was a simple question that could only be given a complicated answer for which we could not possibly have patience. He began to sketch a rich and subtle view, informed by Whewell's 19th century Kantianism and Peirce's pragmatism, and, I expect, if we had time to penetrate it further, we would have arrived at Leibniz's metaphysics. But we did not have time, for our closing event was still to come.

We had scheduled a closing free-for-all discussion session, as is common at such workshops. The chairs in the room have been rearranged ingeniously by Peter Gildenhuys and Balazs Gyenis, so that the participants could sit at the front and speak to one another while still keeping the rest of the room engaged. This discussion session, however, had a novel twist. Kyle had written the book on underdetermination, Exceeding Our Grasp: Science, History, and the Problem of Unconceived Alternatives. He was now also writing the article. It is "Underdetermination of Scientific Theory" for the online Stanford Encyclopedia of Philosophy, which has an enviable status as the authoritative, online resource for philosophy. Participants in the workshop were given Kyle's draft ahead of time and the plan was that we would provide editorial feedback.

In the abstract, that seemed like a good plan. What draft couldn't use some help, especially from a coven of assembled experts? I'd been busy and didn't get to read the draft until the day before the workshop. My heart sank as I read through it and found it quite uniformly excellent. ("We made a big mistake in choosing to read your article!" I had deadpanned to a startled Kyle on Friday.) Indeed Kyle had put in simple terms something that I had previously found impenetrable: just what was Quine thinking in his obscure but immensely influential remarks on underdetermination?

Shortly before the roundtable began, I was chatting with Jim Bogen, who had agreed to chair. "There might not be much to say," I'd warned him. Once again my prediction failed. As the discussion moved around the table and room, the helpful suggestions began to build and I relearned something I had always known. No piece of writing, no matter how perfect it may seem, is every truly finished.


John D. Norton

Saturday-Sunday, 21-22 March 2009

Revised 3/24/09 - Copyright 2009