<html><head /> <body> <META http-equiv="Content-Type" content="text/html; charset=UTF-16"><title>Demystifying Simplicity</title><meta name="keywords" content="philosophy,simplicity,ancestral,biology,characteristics,cladistic,classification,derived,difference,elliott,global,identifying,inference,Local,observation,organisms,phylogenetic,related,similarity,sober,species,"><table style="font-family:Verdana; font-size:larger; " align="center" border="0" width="50%"><tbody><tr> <td style="background-color:silver; border-color:white; border-left-style:none; border-style:none; " width="730"><span style="font-family:Verdana; font-size:larger; ">Demystifying Simplicity</span></td> </tr> <tr> </tr> </tbody></table><br><table style="font-family:Verdana; font-size:medium; " align="center" bgcolor="white" border="0" width="50%"><tbody><tr> <td height="131" width="669"><p align="left"><span style="font-family:Verdana; font-size:x-small;"><strong>Editors Introduction</strong></span><span style="font-family:Verdana; "></span><span style="font-family:Verdana; font-size:x-small; "> How can you tell the difference between robins, sparrows and crocodiles? Elliott Sober uses simple models to discuss the philosophical aspects of such seemingly obvious questions. Phylogenetic inference is the attempt to identify species' genealogies using the observed features of the species themselves. Sober uses the concept of parsimony in phylogenetic inference to explore the implications of, for example, classifying a daffodil with a crocodile.</span><br> <br> <br> <span style="font-family:Verdana; font-size:x-small; ">The example I want to talk about is specific to evolutionary biology. It is the use of "parsimony simplicity" as a criterion for choosing phylogenetic hypotheses.</span><br> <br> <span style="font-size:x-small;"><strong>Robins, sparrows and crocodiles</strong></span><br> <span style="font-family:Verdana; font-size:x-small; ">So you're a biologist, and you would like to know what the true phylogenetic relationship is that connects robins, sparrows and crocodiles.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">Of course, you can't get in a time machine and go back and observe the evolutionary events. All you can do is observe the present-day similarities and differences that characterize these three groups of organisms. Thus, on the basis of these observed similarities and differences, you can attempt to make an inference concerning which species are closely related and which are more distantly related to each other.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">Figures 1 and 2 show two phylogenetic hypotheses. Figure 1 says that robins and sparrows are more closely related to each other than either is to crocodiles. Figure 2 says that sparrows and crocodiles are more closely related to each other than either is to robins.</span></p> <p align="center"><br> <span style="font-family:Verdana; font-size:x-small; "><IMG src="1892_figure1_new.jpg" id="5323" type="3" align="center" width="323" height="265" url="1892_figure1_new.jpg"></span><br> <br> <span style="font-family:Verdana; font-size:x-small; "><IMG src="1892_figure2.jpg" id="5327" type="3" align="center" width="323" height="261" url="1892_figure2.jpg"></span></p> <p align="left"><br> <span style="font-family:Verdana; font-size:x-small; ">The top of the tree represents the present, and as you go down the tree you're going into the past. When two branches come together, what you're seeing is a hypothetical common ancestor. The hypothesis in Figure1 says that robins and sparrows share a common ancestor who was not an ancestor of crocodiles. This means that robins and sparrows are "more closely related"; they share a recent common ancestor. These are two of the hypotheses you might consider about the relationships.</span><br> <br> <span style="font-size:x-small;"><strong>No wings to wings</strong></span><br> <span style="font-family:Verdana; font-size:x-small; ">How could you use the observed similarities and differences of organisms to decide which of these is a more plausible hypothesis? This problem could be addressed by considering a single characteristic.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">A biologist would consider 50, 100, 200 characteristics that describe the similarities and differences, but I'm going to focus on just a single characteristic that applies in this example, in order to give you an idea of how phylogenetic parsimony (cladistic parsimony) is used to solve the problem.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">We observe that robins and sparrows have wings and that crocodiles do not, a similarity that unites two of them apart from the third. How could you use that observation to discriminate between these two hypotheses? Well, here's how cladistic parsimony (the concept of parsimony invented by the systematic philosophy called cladism) answers that question:</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">Suppose that at the bottom of the tree you had an ancestor common to all three groups, one that lacks wings. Now ask yourself: How many changes in character state have to occur in the branches (which connect the root of the tree to the tips) to get the observations that you make at the tips of the tree?</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">Well, on the tree shown in Figure 1 only a single change needs to occur. On that branch, a single change from no wings to wings with no other changes elsewhere in the tree will suffice to generate the observations at the tips. However, on the tree shown in Figure 2, two changes need to be made. You have to have a separate evolutionary event of going from no wings to wings on the lineage leading to robins and then, secondly, a similar kind of transition from no wings to wings in the last lineage leading up to sparrows.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">Biologists would look at the data and say that Figure 1 is more parsimonious and simpler than Figure 2. They would use a criterion of parsimony to decide which of these phylogenetic hypotheses--about the genealogy of the three groups--is better supported by the single observation that we're considering, namely, that some of these organisms have wings and others do not.</span><br> <br> <span style="font-size:x-small;"><strong>The significance of the concept of parsimony</strong></span><br> <span style="font-family:Verdana; font-size:x-small; ">Now, you might think that my talk about simplicity, about parsimony, is really not what's fundamentally going on. Perhaps all that biologists are doing is grouping things by similarity. Why bring in the word "parsimony"? Why count the number of changes that have to take place in a tree? Maybe the simple idea is just that you put things together that look the same. Well, that's in fact not how the concept of parsimony works, as I hope to show you by considering a second example.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">OK, we have a different group of organisms now. We're considering robins, crocodiles and daffodils. And, being keen observers of nature, we see that robins have wings, but crocodiles and daffodils do not have wings. That's a similarity that unites crocodiles and daffodils apart from robins. What does it tell us? In particular, what would a criterion of phylogenetic parsimony tell us about that observed similarity?</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">Let's assume, as before, that the species at the root of the tree (the common ancestor of all three of the groups represented at the tips) lacks wings. From the hypothesis in Figure 3, I hope you'll see that it's possible to generate the observations at the tips by a single change from no wings to wings. What about the hypothesis shown in Figure 4? How many changes have to take place in the tree's interior to get the observations at the tips? Well, once again, a single change suffices.</span></p> <p align="center"><br> <span style="font-family:Verdana; font-size:x-small; "><IMG src="1892_figure3.jpg" id="5331" type="3" align="center" width="323" height="239" name="" url="1892_figure3.jpg"></span><br> <br> <span style="font-family:Verdana; font-size:x-small; "><IMG src="1892_figure4.jpg" id="5330" type="3" align="center" width="323" height="263" name="" url="1892_figure4.jpg"></span></p> <p>&nbsp;</p> <p><span style="font-size:x-small;"><strong>Derived similarities</strong></span><br> <span style="font-family:Verdana; font-size:x-small; ">If you look at the structural difference between the two examples, you'll notice that wings are a derived characteristic in Figures 1 and 2. The characteristic that the ancestor at the bottom of the tree is assumed to have is no wings, and what unites sparrows and robins is the fact that they have wings. So that's a <I>derived</I> similarity. However, in the second example, the similarity that unites crocodiles and daffodils, namely, that neither has wings, is an <I>ancestral</I> similarity. Crocodiles and daffodils have the trait that's exhibited by the common ancestor at the bottom of the tree.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">A way to summarize the difference between these two examples could be to say that parsimony, as it's used in phylogenetic inference, is the idea that derived similarities are evidence of common ancestry, but ancestral similarity is not evidence. Therefore, parsimony rejects the idea that overall similarity is the way to infer phylogenetic relationships. Only some similarities count, namely, derived similarities.</span><br> <br> <span style="font-size:x-small;"><strong>Is simplicity simple?</strong></span><br> <span style="font-family:Verdana; font-size:x-small; ">A familiar feature to anyone who's read the philosophical literature on this is to figure out whether, and in what ways, a simplicity criterion as an epistemic criterion is a guide to how we should think the world is. For example, if you use a criterion of simplicity to judge the plausibility of theories, does it require you to believe that nature is simple? Let's look, for instance, at the beautiful purple prose from Newton, who announced that "nature affects not the pomp of superfluous causes." For scientists like Newton, the methodology was grounded in a very definite view about how nature was put together, and this idea was based in their view of how God built the world. For them, the use of simplicity as an epistemic criterion was not independent of substantive views about the way the world is.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">Therefore, if you want to let simplicity guide your judgements about which theories are more plausible and which are less, what are you committed to assuming about the way the world is? And, in fact, the issue about whether nature is simple is only an example.</span><br> <br> <span style="font-size:x-small;"><strong>The meaning of simplicity</strong></span><br> <span style="font-family:Verdana; font-size:x-small; ">The way biologists and philosophers have thought about the use of parsimony in the case of phylogenetic inference is by asking about the evolutionary process. What assumptions about the evolutionary process are made in order to use phylogenetic parsimony as a guide to figuring out genealogical relationships? For biologists, there's no notion of a grandiose hypothesis, namely, that nature is simple.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">There is also a sociological remark to make about the differences in attitude on this problem. Philosophers who read physicists will often notice that many physicists, from Newton down to present-day physicists, have a very deep conviction that the fundamental laws of nature are simple. If you ask biologists, however (especially biologists since Darwin), about how they feel about this question--without asking them to perhaps analyse it too much--a very common reaction is that nature is infinitely complex. And what they mean by that is that any effect variable one might care to study is influenced in extremely subtle and complicated ways by a large number of causal variables. They regard as a myth the idea that there's some simple true description of nature.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">So there's a very different attitude out there in the minds of many biologists and physicists about the nature of the phenomena that they're actually trying to investigate. Physicists often believe that we have to look for simple theories because nature is simple. Biologists, and I think many social scientists, hold the view that if simple theories are desirable, it's not because nature is simple; in fact, nature is just the opposite.</span><br> <br> <span style="font-size:x-small;"><strong>The philosophy of simplicity</strong></span><br> <span style="font-family:Verdana; font-size:x-small; ">Let's think now more generally about what philosophers have to think about when they think about the problem of simplicity, the problem of parsimony.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">There are really three parts to this problem. First of all, we have to figure how to measure simplicity. What feature of theories or hypotheses makes them simpler? What makes them more complex? How can one tell how simple a hypothesis is?</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">Secondly, we have the problem of trying to justify the use of simplicity in science. Why should the simplicity of a theory be of any relevance to our decisions about how the world works? And finally, the third part--in some ways the most difficult part of the problem--is to figure out how to trade simplicity off against other desirable features of theories.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">Often, simpler theories fit the data less well than more complex theories. Scientists seem to want theories that are simple and that also fit the data. However, the unhappy fact of the matter is that simplicity and goodness of fit are often in conflict. If, for example, you wanted to trade simplicity off against goodness of fit, you would have to figure out in some principled way how much you are prepared to sacrifice simplicity to improve the goodness of fit.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">Scientists do sometimes offer a more complex theory over a simpler one if it dramatically increases the goodness of fit, and, symmetrically, scientists will sometimes sacrifice goodness of fit if they can dramatically increase the simplicity of the theory. It would be nice to be able to say something precise about how much of a gain in one it takes to compensate for how much of a loss in the other, and that's the trade-off problem.</span><br> <br> <span style="font-size:x-small;"><strong>A global and local approach</strong></span><br> <span style="font-family:Verdana; font-size:x-small; ">If that's the problem, this three-part idea is a sort of abstract description of what the problem of simplicity means for philosophers. There are two broad approaches to trying to solve this, or a piece of this problem, which I call global and local.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">A global approach to the problem of simplicity would be that whenever a scientist makes an argument that is based on simplicity, there is a single idea present. When physicists evaluate competing theories, and give some weight to which is simpler and which is more complex, they are doing exactly the same thing that a biologist or an economist would do. The idea is that there is one concept, that it has one justification and that the rules for trading this concept off against other desirable features of hypothesis is the same. You can apply this across all scientific subject matters.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">In contrast, a local approach to the problem of simplicity does not assume that simplicity means one thing across all scientific subject matters. Perhaps parsimony, when used by evolutionary biologists to reconstruct phylogenetic relationships, is doing something very different from what a scientist in another subject is doing when the word "simplicity" crops up.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; ">So we have to be at least open to the possibility that there isn't just one problem here--there is a family of problems. Simplicity is to some extent a pun, and the word really means something substantively and methodologically different as we shift from science to science.</span><br> <br> <span style="font-family:Verdana; font-size:x-small; font-style:italic; ">This article is taken from a public lecture by Elliott Sober on June 26, 2000, at the LSE. Copyright The London School of Economics and Political Science.</span></p></td> </tr> </tbody></table> </body></html>