I am at a turning point in my research carrier. I have worked for several years in the philosophy of quantum mechanics (in particular for my dissertation -- see the description attached). That said, I have always approached the technical details of scientific theories with more general questions in mind. I have recently switched gears and decided to refocus my research in two ways. On the one hand I would like to pursue more general issues in the philosophy of science. I am currently involved in a research project on scientific understanding as I detail below. On the other hand, I have developed an interest in the philosophy of ecology. I find the philosophy of ecology interesting in its own right, and would like to contribute to foundational issues in it. Most recently, I have started a project on the notion of downward causation and the identification of ecological systems. Additionally, I think many issues in general philosophy of science, like the notion of scientific understanding, can be benefit from consideration of the science of ecology. As a first step towards developing an expertise in the philosophy of ecology, I will be teaching a seminar on that in the Spring of 2012.

• Scientific Understanding

  I am interested in the following puzzle. If there is a common view concerning scientific understanding, it is this: scientific theories provide us with understanding only if they provide us with true explanations of their target phenomena.That said, we know that many of our scientific theories and models are poor representations of the way in which the world works. The question is: does this mean that our scientific theories and models do not provide genuine understanding of their target phenomena?

  One could submit that the epistemic value of such theories and models resides in their predictive power and unificatory power. Most of us, though, would like to grant that, say, Newton's theory, has more epistemic value than that. I suggest, as a working hypothesis, that the epistemic value that such theories and models afford is that they provide understanding of their target phenomena (Note that by this I mean a genuine kind of understanding, and not a subjective and potentially misleading feeling of understanding of the type that J.D. Trout criticized, for example in his (2002)). In order to make such a claim possible, we have to decouple understanding from possessing a true explanation. There have been some attempts to do so in the recent literature (see for example Cartwright 2004, Elgin 2007, 2009, de Regt 2009, and some other contributions in de Regt et al. 2009 such as Lipton 2009), but in my view they have not produced a clear notion of understanding. My aim is to articulate a notion of understanding that (1) scientific theories and models can afford independently of whether or not they faithfully represent the way in which the world actually works, and (2) that has genuine epistemic value. Below I will describe two different papers related to this aim.

  When confronted with the problem I describe above, many philosophers have the following reaction (see Mizhari 2011 and Strevens (Unpublished)). A scientific theory that provides a poor representation of how a phenomenon actually comes about in the world, or alternatively makes poor predictions about a phenomenon, does not afford any understanding of that phenomenon. The only notion of understanding that one might associate to such theories is an understanding of how a theory and its models generates the predictions that they do. Call this type of understanding understanding within a theory. For example, one can understand how Newton's theory entails elliptical orbits for a two body gravitational system given certain initial conditions. The understanding does not depend on whether the theory is a faithful representation of how the world actually works, nor does it matter whether the predictions it makes represent any phenomena. In short, such understanding is quite independent of anything that actually happens in the world, hence understanding within. One could well understand how Newton's theory supports various predictions about the behavior of massive bodies without ever thinking about how these models may represent the actual orbits of planets in the world. As such, one may well have understanding of how a theory generates a prediction X without understanding of an actual phenomenon P that may correspond to X.

  I will happily grant that an important sense of scientific understanding is captured by this notion of understanding within a theory. That said, I am also willing to maintain that there is another important sense of scientific understanding that is worthy of attention. Simply put, I believe that in addition to allowing us to understand how they entail such and such predictions, scientific theories provide us with some understanding of the phenomena they target in the world. For example, while it is true that it is epistemically valuable to understand the behavior of a massive body orbiting around another one within Newton's theory, a great deal of the epistemic value that we grant to Newton's theory relates rather to the understanding that such a theory provides of the behavior of actual mechanical bodies in the world, say Mars' orbit around the Sun. And this is true even if, as we know, Newton's theory does not provide a faithful representation of how mechanical bodies actually interact gravitationally.

  In light of the above, I propose to distinguish between two notions of scientific understanding. The first is the one described above, understanding within. But I also want to distinguish a notion of understanding so that theories and models, even when they misrepresent the way the world works, still afford some kind of understanding of the phenomena that they represent (the phenomena that are recovered as predictions of theories and models). Refer to this notion of understanding as understanding of phenomena via a theory and its models. In distinguishing this notion of understanding, I am making a definitive break from the current literature on understanding.

  It so happens that one can distinguish at least two different ways in which we can gain understanding of the phenomena via the scientific theories or models that misrepresent how the world works. The first notion of understanding via that I want to articulate focuses on the idea that understanding arises from the fact that scientific theories and their models \emph{get something right} about how the phenomena actually come about. Making sense of what a theory or its models gets right about the world is key to articulating this notion of understanding via. It will take one paper to do this.

  In the paper, my claim would then be something along the following lines: we gain understanding of a phenomenon via a theory and its models if the theory and its models isolate and faithfully represent the relevant dependency structures that are actually responsible for the phenomenon to occur in the world (I use the notion of dependency structures instead of the more common notion of causal structures as I want to leave open the question whether or not other kind of relationships, such as mereological constitution, symmetry constraints, or logical necessity, can enter the picture.). Scientific understanding of a phenomenon via a theory and its models in this case amounts to the knowledge of parts of the actual dependency structures that give rise to such a phenomenon.\footnote{The question whether understanding is a species of knowledge is debated among epistemologists (see Kvanvig 2003, Grimm 2006, Forth., Pritchard 2009 for example). That said, much of this debate relates to the issue whether understanding is subject to Gettier-style arguments or not. Such a debate is irrelevant to my project.} I believe I can use the notion of difference makers that Strevens (2008) employs in his account of causal relevance to make this idea more precise.

  The notion of scientific understanding described in the previous paragraph captures, I believe, an important part of what we mean when we say that a scientific theory affords us some understanding of its target phenomena, even if it provides a simplified and/or idealized representation of the way the world works. That said, I do not think that this notion of understanding via fully addresses the puzzle I described in the introduction. What is missed by this notion of understanding via is the understanding that theories and models afford of how possibly the phenomena arose. This would be a notion of modal understanding via which will take another paper to articulate.

  Here is an example of modal understanding via. One of the most fascinating results from the philosophy of quantum mechanics is the discovery that radically different views of the world can give rise to the very same phenomena. The many-worlds view of quantum mechanics is local and deterministic, but does not allow for systems to have definite properties. Bohm's view is deterministic and on it systems have definite properties, but it is non-local. Outside of Oxford and/or Rutgers, I doubt that anyone is convinced that either of these views is the right one about the world. That said, I think there is something epistemically valuable about having knowledge of how possibly the world might be in order to give rise to the phenomena. It is this notion that I seek to capture in a second paper on understanding.

  Whether this modal notion of understanding can be made compelling is an open question. The problem that confronts this notion on a moment's reflection is that, as described, it is extremely permissive. Do we really want to say that any theory that manages to predict or account for any phenomenon whatsoever provides us with modal understanding of that phenomenon? Intelligent design comes to mind and raises a big red flag. I think the only way this modal notion of understanding can be made compelling is if one can make sense of degrees of modal understanding via a theory or a model affords of a phenomenon. In particular, we want intelligent design to afford a very low understanding of phenomena, but Newton's theory a very high degree. Figuring out whether one can make sense of degrees of modal understanding is this part of this project.

  Now even if one can make sense of degrees of modal understanding, and even if one can recognize its epistemic value, one might still think that the epistemic value captured by this notion is ill-described as a form of understanding. I am open to this possibility. That said, even if modal understanding as I have described it is not a form of understanding, I will have identified an epistemically valuable part of scientific practice that has gone unnoticed in the literature.

• Downward causation and the identification of ecological systems

  The notion of ``system'' is pervasive in the sciences. It is a matter of fact that scientists are not willing to admit that any assemblage of entities (broadly construed), even if their theories would apply to these entities, would constitute a system. For example, the water on my table, the gas in my car, and the wine in the basement are all entities to which chemistry applies, but most would say they are not chemical systems. Similar examples can be given for physics and biology too. I find it a fascinating philosophical problem to identify, as generally as possible, the criteria that must be met for an assemblage of entities to constitute a system (an interesting system?), and one that I would like to spend some time on.

  The problem, as stated, is so general that one might be legitimately skeptical as to whether anything non-trivial about it can be said that covers all of the sciences. The way to begin is to have a look at the problem within the context of a particular science, and within a particular subdomain of that science. In fact, if we look at the science of ecology, in particular community ecology (a similar problem exists for ecosystem ecology), we find that the problem occupies the attention of scientists as well as philosophers (see, among others, the debate between May 1973, Tilman 1996, 1999, Lehman and Tilman 2000, for scientists, and Cooper 2003, Mikkelson 2004, Sterelny 2006, Odenbaugh 2007 for philosophers).

  Let me sketch the project, which remains, at this stage, in its infancy. Sterelny (2006) analyses the various ways in which one can hope to distinguish robust ecological communities from mere phenomenological assemblages. Phenomenal communities are carved out in a rather artificial way, to serve certain practical purposes. Robust communities, if there are any, are "biologically meaningful" units. One way to characterize biologically meaningful units is in terms of causal integration. Now, as far as I know, the notion of causal integration of communities and ecosystems is only conceived as going in one direction in the current literature, that is, from the components to the whole community or system. Under this view, a robust ecological community, a system, is causally integrated when its components interact in such a way that some system-level property emerges (that is, a property emerges at the level of the system which is not a property of any of the components), and that property has some causal efficacy at the higher level. One does not consider the system to be robust if the system-level properties can be construed as merely epiphenomenal, i.e. the system level properties are not causally efficacious.

  In the science of ecology, much of the debate about whether there are robust communities has long centered on the diversity-stability hypothesis. The diversity-stability hypothesis is about the existence of two system-level properties, diversity and stability, and that diversity causes stability. The view is debated in the literature in part because there are many different ways to define what diversity and stability consist in. Additionally, whether stability and diversity are even correlated is debated, in particular when one takes not only the interactions between plants, but also between plants and animals, into account (see Loreau et al. 2001 for example). Finally, even if one grants that diversity or stability are correlated, it is unclear whether they are causally related, or whether alternative mechanisms for establishing the correlation exist, e.g.~migration or sampling effects (see Wardle 1998, Chesson et al. 2001).

  Given all of the difficulties with the diversity-stability hypothesis, it might be time to examine an alternative strategy for defining a robust community, which is nonetheless in line with the view that one looks for a causally efficacious system level property, to be the hallmark of a robust ecological community. The diversity-stability hypothesis is about whether one system level property causally interacts with another system level property. No one seems to consider an alternate way that system level properties can be causally efficacious: they might be causally efficacious on the system's component parts. This would be a form of downward causation.

  By the notion of downward causation, I mean the notion that a system as a whole could have some causal influence on some of its own parts ( There is a looser notion of downward causation that only requires a causal relation between a ``higher level" property (object, or event) and a ``lower-level" property (object, or event). Since I am not sure how to construe the notion of ``level" and of the ``height" of such ``levels" outside of established mereological relationships, I will restrain my notion of downward causation to the one I described.). Many philosophers and scientists (see various contributions in Andersen 2000 for example) seem to think that either there are instances of downward causation in the world, or that, at least, the notion of downward causation could be helpful for our explanatory needs in the special sciences. The viability of my suggestion, that perhaps robust communities are those in which downward causation is present in the manner described, hinges, of course, on the possibility of making sense of the slippery notion of downward causation.

  Much of the literature on downward causation suffers from not properly distinguishing the metaphysics of causation from the pragmatics of causation (Kim 1998, 2000, Craver and Bechtel 2007, Kistler 2010, Fazekas and Kertész 2011). Even if system level causation is not metaphysically independent of causation at the level of component parts, it does not mean that the notion is not pragmatically useful. Of course, a similar point can be made in the other direction. Just because downward causation is pragmatically useful, it does not mean that it is metaphysically independent of causation at a lower level.

  My own inclinations lean toward a form of ``causal republicanism" (along the lines in which it has been characterized in Price and Corry 2007). First, I want to clearly separate the question of whether causal talk is useful for our dealings of the world, from the question of whether one can make full sense of the notion of cause at the metaphysical level. Second, while I am willing to admit that the latter question is highly problematic, I am also willing to maintain that this leaves room for interesting work regarding the former. In other words, I am more interested in the pragmatic and functional account of causal talk -- what function does it play and what is good about it? -- than in the metaphysics of causation (Note that taking that most of our causal talk has pragmatic leaning does not mean that it is ``subjective" in the sense of arbitrary or idiosyncratic, quite the contrary. See for example Woodward (2007) on this point).

  In the light of the above, I would maintain that the question about downward causation is more interesting when conceived from a pragmatic point of view. A first question is, can we make sense of the notion of downward causation from such a pragmatic point of view?

  Such a question, it seems to me, should be relatively easily answered in the positive, by applying Woodward's manipulability framework for causation to the notion of downward causation. What is needed in order to find a causal relationship in that framework is roughly the following. First, one needs a system that is relatively contained and embedded into an environment that can be used as a source of exogenous influences. Second, one needs to be able to "intervene" on the system so that one can isolate putative causal relationships. Finally, one needs to find that an intervention imposing changes on the putative cause induces relatively invariant and stable changes in the putative effect. Importantly, one's intervention cannot directly cause changes in the putative effect, but only via the putative cause.

  To relate this to downward causation, one simply needs to intervene on system level properties and see if there is an invariant change in the component parts of the system. Woodward's framework is neutral on whether downward causation is metaphysically independent from lower level causation. It could very well be the case that the only way that one can intervene on system level properties is by intervening at the level of component parts. In this case, Woodward's account would support causal claims that are both downward and at the level of components. They happily coexist.

  This coexistence of downward causes and lower level causes might give the reader pause. If one intervenes on system level properties through interventions on component parts, then there will always be changes in component parts, thus rendering downward causation ubiquitous and hence trivial. For example, if one takes a system level property like, say, the average mass of a system, and then one intervenes on the components by say, removing one to alter the average mass, there will be, quite obviously, a change in the component systems. If this is all downward causation is, then it is trivial. That move is blocked because Woodward's account requires that the intervention which is the key aspect in determining whether a causal relationship is present not intervene directly on the putative effect. So, if one intervenes on system properties via the component parts, downward causation would be present only if the properties of the component parts that were not intervened on changed in an invariant way. So, downward causation in Woodward's framework is not trivial.

  It seems to me that there will be cases in the sciences, in particular in the special sciences, which will naturally qualify as cases of downward causation. The kind of experiments that are performed in fMRI (functional magnetic resonance imaging) could well provide such a case. In such experiments, one manipulates the behavior of a whole animal (say, a rat) and studies the effects of such intervention at the neuronal level. See Kistler 2010 on this point. The first part of the project is thus to make sense of downward causation within the manipulability account of causation.

  If the notion of downward causation can be made sense of in Woodward's framework, that would be an important achievement. But I am most interested in the pragmatic question, what is it good for? One can obviously ask this question in many different contexts. The first context in which I would like to address this question is with respect to defining robust ecological communities.

  Recall that the foundational question in community ecology is, what is the general difference between robust ecological communities and mere phenomenological assemblages? The question is whether downward causation is present in those cases in which community ecologists generally agree are instances of robust ecological communities and absent in instances they deem phenomenological assemblages. A good starting point is to examine system level properties, like diversity and stability, that ecologists have already identified as being ecologically important, and then to see whether downward causation is present with respect to these system level properties. In order for this project to get off the ground, I will do my best to enlist the help of ecologists.