 My name is Alexandre Agui and I welcome you to the second MECOLIFE workshop, which is a collaboration between Belgian, Belgian and Taiwanese philosophers about exploring questions about causation mechanisms in biology. And for the second workshop we decided the Belgium team not to present in order to have more selected papers and invitations. And I have to say that you have to check regularly the program on the website. I did not program this website, I did not code this awful interface over the program. But because unfortunately we had a ridiculously high number of last minute cancellations. So what I have to say about the preceding of our workshop. So the people that do not have access to in your home, there's a paper in the entrance so you have access to Wi-Fi if you want. Two, by default we stream every talk on YouTube and we put it on YouTube after. If you do not want to be on YouTube, please tell us because the rule of our workshop in seminars, by default we put everything on the way. But I'm very happy not to. We will comply to European Union's rules if you do not want your complete control of your image. And seriously, for the questions we often put the question period too. If you do not want to appear on the web, you can say it and we will cut this part. That was the first. Three, we are a little bit, maybe it's our Nordic style or North American style. We like conference to go in time. So five minutes before the end of your talk, because we want to have a question period, I will put that in front of you. And if, by mistake, you get to there, you have to try to finish. Port, we have also a tradition in the suffices and it's called, no, we're using the non-bullying style. So in suffices we are non-dogmatic. You can be continental, analytical, realist, unrealist. We don't care, we care about arguments, people developing their stuff. So we can have, you can critique someone else, you can be engaged, but the ultimate goal is to help other philosophers to do better. So to have better opinions. So especially in relation between senior and junior. There's not a lot of junior presenting in this workshop, but I will be part of the... Attentive. Attentive to that kind of power relations. I don't expect problems in this workshop, frankly, but it happens. Especially when we're seeing Americans. We are much more aggressive. So, without further delay, it is my honor and privilege to present the first talk, which will be a plenary talk, so we have one hour maximum to present your before question. It is Marcel Weber who will present to us a talk called, Mechanism as Constituted by Activities. Well, first of all, I would like to thank the organizers for inviting me. It's a great pleasure and an honor to deliver this plenary talk. It's a special pleasure to see some friends from Taiwan again. So thank you very much for organizing this. So, my topic will be one which you may have heard about, Mechanisms in the Life Sciences, a huge topic. And I haven't said much about this topic in recent years. So this is an attempt to sort of jump back into the discussion and offer some actually reflections on recent developments and also a new proposal. Anyway, so here's the plan for my talk today. First, I will outline what the main questions are about mechanisms or more precisely a relationship called mechanistic constitution. Then I will talk about a traditional criterion for mechanistic constitution, which is due to Karl Kraber, which is called Mutual Manipulability or MM. And this criterion has recently been replaced by actually Karl himself with Stuart Glennon and Mark Povich. Sort of a successor criterion for Mutual Manipulability, which is widely thought not to work. So then I will offer some metaphysical reflections. So there's your ontic part about construing mechanistic constitution as an apartheid, a myriological relation between activities. And activities are conceived as metaphysically as occurrence. Then I will talk about how to intervene on activities. And then I will offer sort of sketch a new idea, which I call the domain specificity of principles used to infer mechanistic constitution. So here just to bring us up to speed on the topic of mechanistic constitution. So what you see here is an abstract representation of a mechanism. This iconic diagram is known as a Kraber diagram after Karl Kraber, who invented it. And I think since they officially replaced it by something else, I'm going to miss it, I think. It has so much shaped our thinking, people's thinking about mechanisms for good or for worse. So the black dot on the bottom is supposed to represent a phenomenon. So it's something which happens, for example, a neuron firing or a muscle contracting in a biological organism. On the bottom, you see these small circles, which are some things called entities in the literature. But I think they mean things that are engaged in some activity. For example, the thing X is engaged in Phi, Wonning, etc. Then you see arrows and dotted lines. The arrows are supposed to represent causal relationships. So the activity of the things in a mechanism can affect each other causally. But in addition to this causal relationship, there is also, and this is represented by the dotted line, there is a relation called, well, a different expression, just mechanistic constitution or constitutive relevance. This is supposed to be a non-causal relation and it is supposed to be synchronic, as opposed to causal relations which are thought to be diachronic. So this is the basic picture. And the question that has engaged people for, I guess, about two decades now, is how we should think of this dotted line here. What is this constitutive relevance or this mechanistic constitution relation? And how can it be ascertained? How can it be proven or established? It is important to distinguish different questions about mechanistic constitution. So with Kraver, Glen and Povich, I think we should respect this threefold distinction of questions. There are, namely, conceptual questions, in particular the question, what exactly does mechanistic constitution mean? So it is a question about the meaning or the content of this concept. Then there are epistemological questions, namely, what should count as evidence for the presence of mechanistic constitution relations, or also, perhaps, something stronger proof, so something like decisive evidence, or evidence that is sufficient to establish, a claim about constitutive relevance. And thirdly, there are metaphysical questions, such as, what are the truth-makers for claims of mechanistic constitution? And I will be speaking today mostly about the first two questions, but also about the third, although I will not address directly the question of truth-makers, but maybe indirectly. Anyway, so in his original proposal from 2007, Carl Kraver proposed a sort of an operational criterion for detecting constitutive relevance, and this is known as mutual manipulability. So Kraver argued that there are two types of experiments that are used in research, looking for mechanisms, in particular in neuroscience, but also elsewhere. There are experiments where scientists intervene bottom up, which means they intervene on the constituent, on the constituents, on the components of a mechanism, and then they detect some change in the phenomenon. So for example, they intervene in a neuron, an individual neuron, and then they observe a change that happens in the whole organism or in the subsystem under study. The second type of experiment is a top-down experiment, where according to Carl, scientists intervene on the phenomenon itself, and then they observe or detect a change at the level of the components of a mechanism. So for example, in a neuroscientific experiments, subjects may be asked to memorize a word or to think about something or to visualize something, which is sort of intervening on the whole phenomenon of cognitive processes, and then something is observed in the brain. Some neural circuits light up in the MRI or whatever method is used. And Kraver proposed that if a component, which is implicated in a mechanism, satisfies both of these requirements, which means it can be manipulated bottom up, but it can also be manipulated top-down. This is sufficient for establishing constitutive relevance. This idea has received much criticism in the literature. So many people have questioned the whole idea of a top-down manipulation, and I think perhaps the most rigorous sort of objection, or if you want proof of the incoherence of this notion, is due to Michael Baumgartner, Alexander Gebharder and Lorenzo Cassini. You can find this in these two papers. So here is their way of sort of cutting out the problem. So on the top you see Psi, which is supposed to represent a phenomenon to be explained by the identification of a mechanism. And you see that there is an intervention variable, which targets specifically Psi. Then there are further intervention variables, I Phi 3, I Phi 2, I Phi 1, which directly target constituents or components of the mechanisms, which are some entity, some object, something which is doing something. So Phi y, for example, could be a neuron firing, an individual neuron firing or something like that. And then you could directly intervene on this neuron with an electrode, for example. And here you also see, again, the arrows mean causal relations. So the interventions are basically causes. And then you see the dotted lines, which are constitutive relations. And these are just the possible interventions that are at stake here. Now Baumgartner and Gabe Harter have proposed the following reductio argument, purporting to show the incoherence of the idea that you could intervene on the whole of a phenomenon, and then with respect to some of its component, such as in order to verify a relation of constitutive relevance. So they suppose that I Psi were an ideal intervention variable on Psi with respect to Phi I. So here this is I Psi. It's supposed to be, so this is a supposition, an ideal intervention variable with respect to those components Phi I to 3. Then according to interventionist theory, this implies that I Psi is a cause of both Psi and the components Phi. This leaves us, according to Baumgartner and Gabe Harter, with three possibilities. Either I Psi causes Psi, which causes Phi I, or I Psi causes Phi I, which causes Psi, or I Psi is a common cause of Psi and Phi I. So there are these three options and none of these three options could count as an ideal experiment on the activity Psi with respect to its component activities. Baumgartner and Gabe Harter argue that the first option is impossible because Phi I are supposed to be spatial temporal parts of Psi I, which means if they're spatial temporal parts, they cannot stand in a causal relation. The second possibility works, but it's not a top-down intervention, it's a bottom-up intervention. Now the third possibility contradicts the supposition because it's supposed to be an ideal intervention, which means it's supposed to be surgical with respect to the components, which means it should manipulate only that component and no off-track components, which means that components which are not on a causal path between the two. But this is not satisfied here. If it is a common cause of Psi and Phi, we have a contradiction to the supposition. This intervention is not surgical or ideal, it is fat-handed. So it manipulates two components at the same time. Therefore Baumgartner and Gabe Harter conclude that the notion of a top-down intervention or ideal top-down intervention is incoherent. Now enter this recent paper by Kravor Glenneman-Povic. Now I read them as accepting the incoherence argument, although I'm not entirely sure if they're really accepted, or if they just defend the mutual manipulability criterion in a somewhat modified way. But anyway, they officially announce that the MM criterion is to be replaced by a different but related, according to them, criterion, which they called matched interlevel experiment, or MIE. So here's the new picture. Now mechanisms are represented like this. You have a series of activities which sequentially act and in total form a process which takes the system from input conditions psi into output conditions psi out. Now for simplicity, this is depicted in a sequential manner, but they indicate that mechanisms could also be non-sequential, so they could branch or join or whatever. So this is the new way of representing mechanisms and the new way of experimenting, of establishing constitutive relevance relations. So now we're talking about deciding whether or not some of these activities represented as a circle is actually part of the whole activity of phi. So this has to be established, again, by three types actually of experiments. Bottom-up inhibitory experiments, where you intervene on a component activity such to inhibit it and observe an inhibition of the whole process. Or there are bottom-up excitatory experiments where you activate one component activity and observe the process, increasing the activity of the whole process. And then the third is supposed to replace the old top-down intervention. Now how they describe it is you intervene on the input conditions and then observe a change in the components of the phenomenon, the mechanistic phenomenon, and in particular you will be looking for changes that are of the same kind as and occur with quantitatively overlapping ranges with the activity psi I detected in top-down experiments. So you need to observe the right kind of dependence between an intervention on the input and the kind of change you observe. So it somehow has to be proportionate, quantitatively proportionate to the output. So now I must say that I'm practically totally on board so this is constructive work with this proposal. So what I'm going to suggest, what I'm going to present are perhaps hopefully some further clarifications of points that are maybe not entirely clear in the proposal and also I think we need to qualify the method and I think this will also not particularly we will have to qualify this account in a way which is not necessarily opposed to Craver's, Glennon's and Povic's proposal. I haven't discussed it with them yet but I see no reason why they should necessarily disagree with what I will offer. So let us step back on the traditional accounts of mechanisms and note that they do seem to posit because they posit things and activities both at the same time as constituents of mechanisms. It seems that there are two distinct kinds of myriological relations in a mechanism. There are part-hood relations between a mechanism or sometimes a system S and some things Xi and then there is another myriological relation between a phenomenon Xi and the component activities Phi i and one thing to notice is that the S has disappeared here. There's no longer a system. So if S once was something which was composed from other things perhaps a continuant composed from other continuance which means things like objects or entities that has disappeared. It's no longer in the picture and I think that's right. That's a good move. So I think the right view and this is very much inspired by a paper by Murray Kaiser and Beate Krikle. So I think the right way of conceiving the constitutive relevance relation is as follows. We first think of explanandum phenomena as complex processes. Complex meaning as opposed to simple. Meaning they are composed of other processes. In fact, processes or activities are not systematically distinguished between processes and activities. Now these complex processes may involve things or not. I think the things are not necessary especially since many philosophers of biology have gone into process, metaphysics, process ontology thinking that traditional things or continuance can be dispensed with in biology. Its process is all the way down so we don't really need them. I don't necessarily want to defend a process ontology but I think that an account of mechanism should be flexible enough to account for mechanisms that contain things and such mechanisms that don't. Now the second move is to construct constituent activities of these complex processes as spatial temporal parts of such processes. So the constitutive relevance relation is just a parthood relation, nothing else. And it's a parthood relation that obtains between processes and activities which are occurrence. So they are extended in time. They have temporal parts. Now just for illustration consider a Swiss watch like this one. We could say that the springs and wheels from which the watch is made are spatial parts of the watch. The watch is a continuant. You can be construed as a continuant. And the springs and wheels that make up the watch are also continuance. Then there's a relation of spatial parthood between these continuance. However, the mechanism is quite a different thing. The mechanism consists... So that's the phenomenon. So that's the watch showing the time. That's the phenomenon. And that is composed of the motions of the spring and wheels. So it's what these springs and wheels in the watch do which composes or constitutes the mechanism. So the phenomenon is extended in time. It has temporal parts. So are the motions of the springs and wheels in the watch. So the phenomenon is a process which is constituted by motions or activities. And those are occurrence. And the constitutive relevant relation is a parthood relation among occurrence. Whether we also have continuance, like in the case of this watch, this is an open question. We might not have them in all mechanisms. But in all mechanisms we will have this kind of parthood relations between phenomena and activities. So the springs and wheels constitute the watch. Their motions constitute the phenomenon. We could say that the mechanism is the set of constitutive motions and their spatial temporal arrangement. And even though sometimes things or continuance do figure in descriptions of a mechanism, the constitutive relations hold between the motions, between the activities. So this is just for illustration. Now I want to think about what it could mean to intervene on activities. If the picture I just presented is right, intervening on an activity means intervening on a temporarily extended or current. So you're not intervening on a thing. You're intervening on an occurrence. Now interventions are themselves occurrence, which is as it should be. There can be conceived of either as events or as processes themselves. And their temporal dimension, which means so their exact timing, the moment at which they occur, at the moment at which they begin, and how long they last, can make a difference with respect to the outcome. So sometimes exposing a cell to a growth factor for a longer period of time can have a larger effect or actually annihilate the initial effect. So duration, timing and duration is of the essence very often. So what about, what could it mean to intervene top-down on an activity? So it's clear what it could mean to intervene bottom-up, which means we intervene on a component and observe something in the phenomenon. But what about those top-down intervention? We could ask if the Baumgarner-Gabe-Harter-Casino-Reductile argument applies to intervention conceived as temporarily bound occurrence, acting on other occurrence. And I think they would say yes as long as so long as there are spatial temporal overlaps between the phenomenon and the components, the argument goes through because it relies on any dependence of the part on the whole not being causal. And if there's a spatial temporal overlap between two elements, according to interventionist theory, it's not a causal dependency. So the reduction goes through as long as there is this kind of spatial temporal overlap. Now Craver, Glennon and Kovic try to solve this problem. And this is my sort of analysis of what they're doing. What they're doing is they separate psychonceptually from the constitutive activities by identifying the phenomenon with input and output conditions. So they have replaced the old version of the top-down intervention which used to be as follows. In the conditions relevant to the request for explanation, there is some change to S's sighing that changes X's sighing. So that was the old top-down intervention requirement for constitutive relevance. And that gets reinterpreted or replaced by now bring about start-up conditions, sighing, measure X's sighing, also measure sighing out, so the output conditions, evaluate thereby whether X and its sighing are changed when sighing produces sighing out. So what happens is that they add a fourth possibility to Baumgartner and Gapeharter's three options. So according to Baumgartner and Gapeharter, a top-down intervention variable would either have to intervene on sigh, which then causes phi i, or it would have to be bottom-up, or it would have to be a common cause. Now Craver, Glennon and Kovic add a fourth possibility, namely, which is to intervene on sighing on the input conditions of the mechanism, then observe some change in X's sighing and a change in the output conditions, sighing out. So this is a perfectly fine, according to them, causal graph. There's no violation here of any requirements of causal graph theory in particular. There is nothing that Spatia temporarily overlaps here because you intervene as it were before the mechanism unfolds. So they seem to now are left with this. They have the mechanism being that set of changes that occur between the input and output events or processes, which means basically that the phenomenon is, as it were, it doesn't contain the mechanism. And this actually has been criticized by Marie Kaiser and Beatrice Crickl in their paper Metaphysics of Constitutive Mechanistic Phenomena. They argue that this move obliterates the distinction between etiological and constitutive mechanisms, which used to be part and parcel of the new mechanistic doctrine. So you can have etiological mechanisms where you just have a series of steps which depend on each other causally. You can have constitutive mechanisms where you have this part-whole kind of relation. Now, I'm not sure whether etiological mechanisms can be retained in the new picture. I think this needs to be discussed. But in any case, it is clear that this view doesn't allow to be the mechanism to be temporarily contained in the phenomenon, which is not consistent with the view that phenomena are myriological sums of mechanistic components. Because if they are, then the mechanism would have to be spatially-temporally contained in the explanatin phenomenon. So Kaiser and Crickl think that this is the right way of depicting the relationship between a phenomenon and the mechanism. Now, I think this is correct and I think it seems to me Kraver, Glennon and Kovic are also committed to it. So we will have to conclude that on the new picture, the excitatory top-down interventions do not intervene on the phenomenon itself, but rather on the conditions obtaining before the process begins. Or perhaps at the moment where the process begins. So the top-down intervention trigger or initiate the process such that it can be observed. And typically it will be observed in this type of experiment in vivo, not necessarily, but typically. So let us consider this question of, well, what do these top-down excitatory interventions actually do? On an example, and I'm going to use also the example, the main example used in the CGP paper, which is an example from neuroscience, from the neuroscience of the nematode center of dytus elegans. And the behavior in question is a withdrawal effect shown by these nematode worms. When you lightly tap them on the head, so when you tap a worm on the head, it will do two things. It will stop wiggling its head. And nematodes like to wiggle their heads. So all the time they do this, they wiggle their heads. When you tap them on the head, they will stop the wiggling and they will withdraw. So they will make a backward movement. This is shown here. So here you see the worm wiggling its head. Then a touch is delivered to the head. The wicking stops and the worm sort of withdraws. And then after it has turned, it will resume the head wiggling. Now the function of this behavior is actually quite scary. It's eerie. So the function of this mechanism is probably a defense mechanism against a predatory fungus. Yes, there are predatory fungi. And what they do, they build ring-like structures. So the fungus builds rings which traps nematodes. So a nematode which is unfortunate enough to swim into such a ring will be trapped. Now this reflex probably allows the worm to escape from this trap. And the first thing you want to do when you want to withdraw from being trapped in a ring is you want to stop wiggling your head because otherwise you will not get out. So you want to stop wiggling your head and then withdraw. And this is what the worm does when you tap it on the head. So probably that stimulus touch on the head will be activated when the nematode is trapped in a fungal ring. And this complex set of behavioral responses then ensues. Now to the right you see a neural circuit which presumably represents the mechanism that explains this phenomenon. And it nicely actually corresponds to the different phases. So here on top you have a sensory neuron so that responds to a touch, to a mechanical signal by activating these controlled neurons, this amount neuron. And they activate this motor neuron here by secreting the neurotransmitter tyromine. So this is a tyromine-sensitive motor neuron and what it does is it relaxes the mechanisms. So the wiggling of the head will stop. And also these command neurons will activate motor neurons that cause muscle constructs, so rhythmic muscle constructions and realizations on the other side. So to move backwards the worm will contract muscles on one side of the body and relax muscles on the other side and then reverse. So this will make this wiggling motion. And this is controlled by this motor neuron. Now what kind of experiments were actually used by neuroscientists when unraveling these mechanisms? So one of the first experiments or one of the central experiments I'm not sure if it was chronologically first is exogenous tyromine application. So just adding tyromine to the growth medium, to the growth surface of the worm. This immobilizes the worm. So in the presence of tyromine they cannot move. Now genetic studies show that there are mutants in particular SER2 which is deleted for a signaling protein, a G protein that these mutants are unresponsive to tyromine. So they are not immobilized in the presence of tyromine. But tyromine sensitivity is restored if the gene is delivered as a trans gene. So the defective gene is, as it were replaced by a functional copy of the gene. Then tyromine sensitivity is restored. There are other mutants in the G protein signaling pathway which have similar effects like SER2. Neuroscientists did expression studies. So they observed where the SER2 gene is active and it was shown to be active in head muscles. Specifically in cholinergic and gavaiergic motor neurons which are responsible for the muscle effects. So other neurotransmitter inhibitors were used. So other substances with known effects also can inhibit these responses. Scientists did laser ablation studies on various neurons. So they targeted specific neurons in the worm, in the central nervous system and obliterating individual neurons by laser beams and observing what kinds of parts of the process were disappeared. Finally, optogenetic studies were conducted where neuroscientists selectively expressed light-sensitive ion channels in specific neurons. So this is a very intriguing technique where light-sensitive ion channels like channel rhodopsin are expressed in the membrane of neurons and then activated by light. When these channel rhodopsins can be controlled by light flashes so this is a way of specifically activating specific neurons. And all these findings were then used in order to construct this neural circuit diagram. So these various connections here which are inhibitory and excitatory relations were discovered with these experiments, these kinds of studies. Now most of these experiments I think can be understood as straightforward bottom-up interventions trying to determine if changing in activity changes the phenomenon. So you add tyrosine which inactivates the motor neuron which immobilizes the worm. You take out a neuron with a laser beam and observe that some part of the process no longer occurs. So these are bottom-up interventions. But I would like to highlight the importance of experiments trying to localize activities. So it's not only important what happens when you wiggle at some point, it's also important where it happens. I think this is crucial information when constructing a mechanism. This has I think not been sufficiently acknowledged in the literature. So I think that not only the kind and quantity of activities must match as required by the new matched inter-level experiment criterion but also their spatial temporal location and dynamics. So the effects observed by the various interventions must show the right dynamics and they must occur in the right places. So how should we think of these activation experiments? And according to Kraver, Glennon and Povich the top-down experiment is basically the touch on the worm's head. That's a top-down experiment. Because it activates the whole process, that's what it does. And I suggest that we change our interpretation of what such activation experiments do. I think they should not be viewed as interventions designed to test for constitutive relevance at all. No matter where the bottom-up or top-down. I suggest to construe these experiments as some kind of a preparative intervention. The purpose of which is to produce and exhibit the phenomenon you want to study. I mean here, the phenomenon doesn't occur when you don't tap the worm on the head. And so produce and exhibit. Here I'm missing a word in English, the German word Darsteln. It's very nice because it can mean represent but also produce. So a very nice verb which you don't have neither in English nor in French. So the idea is that the experiment, the goal of the experiment is as it were to make the system manifest the phenomenon by changing the input conditions or the initial conditions of the experiment in such a way that the phenomenon will actually occur. I think it's this type of experiment and it's what it does, it prepares the system for the experiment. So that's why I call it a preparative intervention. So what about mechanistic constitution? How can we get a sufficient proof as it were a sufficient criterion for mechanistic constitution assuming that all bottom-up interventions will not be sufficient in particular. They will also lead to things showing up which we do not want to be part of the mechanism. So my take on this is that it has been a mistake in the literature to think that there is a domain neutral sufficient criterion for mechanistic constitution. I do not think there is such a criterion, neither epistemologically nor metaphysically. It is my view that not all processes that are accessible to mechanistic explanation are decomposable into constitutively relevant activities in the same way. In various domains of the life sciences and other sciences as well perhaps, there are criteria for piecing together mechanisms. In particular there are different criteria for the connectedness of activities such that they produce a phenomenon. So for example, in the worm withdrawal example that I have just discussed, scientists use principles for constructing neural circuits. So these are very specific methods used for reconstructing, for mapping neural circuits and only these principles are sufficient for deciding on constitutive relevance. Bottom-up interventions can be helpful the alleged top-down interventions are clearly necessary but in order to really establish whether some component activity is part of a process you need methods or principles that are specific to the domain. So in biochemistry for example there are principles for mapping or constructing metabolic pathways which means there are specific methods for mapping metabolic pathways but there are also principles on the theoretical level that determine whether two activities jointly form a process. In developmental biology we could mention principles for mapping gene regulatory networks and so on. So my hypothesis, and this will have to be confirmed by further work, my hypothesis is that in various domains of science you have different principles for proving constitutive relevance. There is no domain neutral method in contrast to causal reasoning. So I think causal reasoning is the only domain neutral method here. Now each of these principles comes with its own methods for piecing together or mapping mechanisms and their activities. So for example in neuroscience you have electrophysiological recordings or optogenetics as being salient methods in biochemistry, radioactive tracer methods, enzyme inhibitors, excuse me. Now of course some of these experimental methods are used for quasi-ideal interventions to test for causal relations and here the principles of causal reasoning are pertinent but there are also, I contend, methods for ascertaining constitutive relevance which are not domain neutral, unlike causal reasoning. So here are my conclusions. I think we should answer the conceptual question, what is, or what does it mean to say that two things are related by constitutive relevance. It just means parthood as applied to processes. So it's a myriological relation, conceptually, a myriological relation between occurrence. So mechanistic activities are nothing but perhaps essential spatial temporal parts of processes. Essential, that's just a hypothesis. I have no good sort of argument for establishing. I just have a hunch that there will turn out to be essential parts of processes. Now the good way of answering the epistemological questions is as follows, bottom-up experiments can provide evidence for mechanistic constitution basically using domain neutral causal reasoning principles but as I have argued such evidence needs to be complemented with domain specific data methodological principles and constraints. For example, such principles connected to the mapping of neural circuits, metabolic pathways, gene regulatory networks, etc. Now the so-called activation or top-down experiments I think are important but we better understand them as basically just preparatory interventions that are supposed, the purpose of which is the production of the phenomenon such that it can be observed, measured, experimented on. Finally, and I haven't argued for this here but I think also that the truth-maker so the metaphysical questions about truth-maker I think the truth-makers for claims of constitutive relevance also depend on the kind of process studies. I don't think there's a universal theory of the truth-makers of constitutive relevance relation. I think that we should be ontological pluralists about what individuates processes and also about what might constitute a natural decomposition of a process into parts which is what mechanisms are. Thank you very much for your attention. Wonderful questions. I'm going to do some... I've got to pronounce your name wrong. It's Lina. Casturner? Lina. E-E-N-A. Casturner. C-A-S-T-N-E-R. No, sorry. Yeah, okay. But I can say to the... Yes, ma'am. So she has this notion called mere interaction which is the idea that not every experiment is about investigating the mechanism. Some experiments involve interventions about changing something so that you can produce and exhibit. That's similar to your preparatory intervention, but I feel there... So I was just wondering if you familiar with her work and maybe there's some... You had thought about the similarity or difference between her concept and your concept. When she proposed this idea, I think if she wants to show that they are part of the scientific project that is not captured by the mechanism discussion and that's why she coined this term. But in your preparatory intervention, it's still in the overall mechanism framework and you want to highlight the top-down type of experiments that maybe would better understand this type of intervention. Yeah, I'm still trying to figure out whether you guys were saying the different things or saying the same thing in different ways. I'd be very happy to read this paper. This sounds like it could be related. You see, this happens a lot. The literature on mechanisms is so vast by now that it happens a lot that... Well, I certainly cannot read it all that. Sorry. And the second follow-up question is more on this preparatory intervention. Yes. You talk more about in vivo experiments, so I wonder if in vitro or even simulation type experiments... So you mean in silicone? Yeah, and also maybe just brain size. It's not like the animal in actual movement environment. So maybe let's just talk about simulation using computers. So in that case, I wonder whether you will treat that also as an example of what you call preparatory intervention in the simulation experiment. So for example, in my case study, the neuroscientists collected neural data and they input those data into the computer model and stimulate neural network. And they performed some intervention on the network assembly based on the data. And so that kind of intervention is... Would you call that kind of... also kind of preparatory intervention in your sense? Yeah, or maybe simulations of interventions. So I mean, I totally agree that computer simulations are used in discovering mechanisms, not only in neuroscience, also in genetics and development biology, so work on gene regulatory networks. I think Charles is interested in this. So they run simulations of these networks. And I think simulations indeed can be part of those methods that are used in a particular domain, in a particular way to establish constitutive relations in mechanisms. So that's totally part of the method. Now, with respect to experiments... Now, there has been much discussion as to whether in silico experiments are actually experiments. So that's how they're called in biology. And I don't really have a view about this. It only seems to me that when you look at what is done in such simulation studies, very often you see manipulations... Let me put this neutrally. So you see moves that are made by the scientists that could be described as here they are simulating an experiment. They're not only simulating the system, they're simulating the experiment. So for example, they have in their equations that they have a term for the calcium concentration. And then they will run the simulation with different initial values for the calcium concentration. And to me this looks like now they're simulating experiments there as it were trying to determine what would happen. If the target system were like the simulation and you added calcium to the system, what would happen? So I would say it's kind of factual indeed. But it can provide evidence, absolutely. So I have no problem with that, but I think I would call it a simulated intervention or a simulated experiment rather than a need to mention or an experiment. Caleb, yeah, that's me. No worries. Thank you very much. I have a question about the objection to mutual manipulability. And I confess I'm not updated on that on this literature, so apologies if I'm asking you to explain something that somebody else has already put in print, but it seems a bit odd to me that Craver and Glennon and Povich are compelled by this objection that if you were to say wiggle a sigh, then you wiggle an off-path causal variable at the level of five. Reason is because so Jim Woodward has responded to this bomb gardener sort of style objection. And I think compilately, right? And he says, well, look, it's just not, we just can't treat this kind of relationship as the kind of relationship that is represented on a causal graph. We use arrows here, but really, if we have this sort of synchronic dependence between the upper level and the lower level, then that path of manipulation is not a causal variable. So if we are wiggling a variable at the higher level, holding everything else fixed, and we see a corresponding wiggle at the lower level, that's not a causal relationship, right? Otherwise we're kind of begging the question against mutual manipulability. So I wonder why I get... So what is the force there? Why can't we just say, well, yes, we should not read that arrow as a causal arrow? So you mean in the argument, so the first possibility here, read the second arrow as not non-causal. Can you explain to me what this is supposed to be again? Oh, sorry, yeah. So psi represents the phenomenon. So that's the phenomenon to be explained. And phi represents the activity of the mechanistic constituents that are responsible for this phenomenon. Does that help? Oh, that's an intervention variable. So that's... So the i is an intervention variable. So it means that's the wiggling part. Yes, exactly. So Woodward says this, and also Elliott Silver and Larry Shapiro say this in another article where they talk about Jaguar and Kim's causal exclusion argument. They say, look, if you wiggle a supervening variable, then it just seems conceptually kind of weird to talk about a corresponding wiggle in the supervenient space as a causal relationship. Well, I mean, it's, I guess, taking the principles of the interventionist theory just basically literally. And I think Bamgon and like other people, Bedler and Dunn, for example, have argued, well, that's what follows from the theory. You can't just say this is not causal where right after you have put forward a theory that says that when you wiggle something, then... So you would have sort of to... And I think now Woodward does this recently, right? He says that definitional and supervenient relations are explicitly excluded from the theory. I guess the people who buy this argument find this move ad hoc. Just motivated by the desire of keeping these top-down interventions. And I think this is how Bert Luriden argues. And I'm not sure, I was surprised to see that they're actually buying the argument. But I'm not sure if they do. I mean, the paper has, as it were, it has two phases. Okay, we totally get it. There are no top-down interventions. Let's sort of modify... Let's get rid of them and propose a new criterion. And then comes the MIE. That's one reading of the paper. Another reading is... No, we don't buy the argument as such. We will show you how these interventions that as they were originally conceived by Craver, how they can be made sense of. Maybe it's a bit of both. It's a bit sort of... It is a bit Janus-faced. And I must say I haven't had a chance to discuss it with them. But anyway, I guess this would be the response of the defenders of this argument to this move. Can I just have a quick follow-up? So I think that's right. I think then what the defender of Woodward might say in that situation as well. The way I can wiggle out of a wiggle, wiggle out of a being called ad hoc, is to say that the corresponding variation is synchronic. And so if mechanisms require causal relationships, if I'm holding everything else fixed, and I wiggle to the higher variable and a lower level variable corresponds, then I've had a synchronic relationship and now I've restored mutual manipulability. I mean, I pretty much have been held by that. Yeah, I think this concerns in particular a relation which we should properly describe as sighing at T and sighing at T. Now that's not a relationship between a process and activities as a whole. It's as it were a time slice. And when if you indeed have constitutive relation between these time slices, and I think this ought to be the case according to the metaphysical picture, you ought to have such relations. So synchronic constitutive relations between occurrence. Now here the metaphysics does get a bit blurry. I must say, but in any case, in this case you have a synchronic relation and the argument applies. I've been thinking, well what about if you don't consider the activity at some time point, but over a longer duration. Which means that you could have diachronic relations. So you wiggle the process at some time point T and then you observe what happens later. Then at least the synchronicity problem will be gone. But I think Baumgartner, Gapeharder and Cassini will still say, well if it doesn't matter if it's not synchronic, if it's a part, it's not admissible as a causal pathway. And I'm not sure I'm buying this yet. I think there might be something interesting going on here, but I haven't been able to get this point clear for me. So that's something I'd like to think about more and talk about more. I think it might turn out to be an interesting point. Well my problem is, well suppose you intervene at a process at an earlier time and then observe at the lower level what happens later. Now in what sense have you really intervened top down then? If you haven't changed the whole process but only a part of it, haven't you just intervened on a part and then it's not top down. So I guess this is part of my problem here. Well I think there might be some interesting discoveries there of a metaphysical conceptual kind lying in wait. This was really cool, thanks. Yeah I know you're right that I'm interested in you're by things so I'm going to ask you a question that I think well you'll understand, you'll see what I'm asking. But I do really, in general, I really like this move to making the constitution criteria domain specific the top down criteria a little bit more domain specific and sort of taking on the ontological pluralism that kind of comes along with that. I wonder what your thoughts are though on, so it strikes me that once you make that pluralistic step. If everything doesn't wind up cashed out in the same kind of domain independent causal currency then we want to start thinking presumably about something like adequacy criteria for these different ways in which to parcel out a big process into little processes if you will. And I wonder what you think about, I mean you could just be pragmatic here right and say the proof of the pudding is in the eating and you know if whatever neuroscience works well parceling out processes in manner X so that sort of gives a stamp of approval to whatever methods they're using conceptual frameworks and the methods to get at those concepts that they're using to break apart those high level processes. Do you just want to do that or do you have a do you want to have a more general theory about when these kinds of domain specific moves are sort of functioning adequately. Is there a more general criteria that you could growth our way toward about what makes for good composition. Right, that's an excellent question. I haven't thought about it. I mean some of my knee jerk reaction would be to say no. Sure. I mean would be to say that they're going to be very abstract criteria that need to be fleshed out differently for each case. So you will have maybe quality criteria for a good neural circuit which will have to do with its robustness with its predictive accuracy and its conformity with background knowledge etc. But these are all very vague criteria. Epistemic virtues basically. Exactly, basic epistemic virtues which I think are important but they're so vague. I mean how much robustness is enough when doing a sensitivity analysis or something like that. And that's going to depend I think on the discipline. So each discipline will have standards of quality standards for when a neural circuit can be published. It's publishable. But this is going to be highly disciplined and specific again. Cool, thanks. I just have to follow up on using my privilege as a prisoner because that's the same question but from the metaphysic point of view. The main questions that you will receive by metaphysicists is that if you define a part of relations you need an individual a way to integrate the host. So you need a way to integrate process. I have to say that your proposition is extremely appealing but as in the definition I would say how do I integrate process to other parts? And that could depend domain stuff but hopefully there's some central concept that should work even if there's different specificity by domain or we're not talking about the same thing at all. Yeah, I mean here you got me because I'm not a metaphysicist. Never was unlikely to become one until my retirement. It's a problem because the proposition is appealing. We would like to. I have been doing some reading on this I mean I'm of course intuitively drawn to the pluralistic approaches that say no there's no general individuation criterion for processes that depends on the kind of process. So a performance of a symphony if that's a process that doesn't have the same individuation criteria like an action potential in the brain. And that sounds intuitive if you're appealing. I mean there are proposals about the cause of coherence of processes of James DeFrisco who is he still here actually? No he was at the other level for a while but I think he's... Oh it's the other level. Sorry. No, no. It's a reasonable mistake. No, I see it's unreasonable mistake. It's unreasonable. Anyway, so he has a very interesting proposal about processes being individuated by the causal coherence. So that means that within a process there are more and stronger causal dependencies than between a process and its environment. That's one idea. And I think that's certainly worth discussing. But my sense is that this isn't always going to work. That's what I fear. So for some cases you will have a different... I have to use different individuation criteria. So pluralism about processes but also pluralism about or perhaps promiscuous realism about the decomposition of processes into natural parts. Because ultimately that's what we are talking about when we ask, well what are the real constituents of a mechanism? We want to know what are the natural parts of a process? Process... There may be many ways of carving up a process into natural parts. So some kind of... There may all be good. So some kind of promiscuous realism about processes and decomposition of processes will be my answer. But maybe this doesn't make sense metaphysically. At least there are some respectful metaphysicians who would accept that. Peter Simon, for example, I spoke to things that's fine. And he's a metaphysician. Peter first, Karen, if we have time. Yeah, thanks a lot for the same talk. Quite sympathetic to the proposal but I have some questions that's quite similar to... a bit about this very logical relation. So if you say that the phenomenon... the activities are parts of the phenomenon and I'm wondering whether the activities are wholly constituted of the phenomenon or whether there is still something else in the activities above constituted of the phenomenon? Completely thinking about the examples you gave like the watch or the behavior of the worm. In the two cases it seems that you have to add something above the wheels, the mechanisms, specifically the activities of all things happening in there to get to something quite complex like measuring time. In the case of the watch, you didn't mention time for the activities and all of a sudden there's a very complex notion of measuring and time and so on that partly is subjective. It's for us that it has this function but maybe not from the objective point of view. So there seems to be an element at its... if it's a neurological relation then just the parts do not seem to... the activities themselves do not seem to be enough to get to the whole. And the same for the worm. I mean you have these neural processes but to look at the behavior of the worm you have to first there is an added step of realizing that there is an organism first of all. I mean you might be blind to the fact that there is an organism at all and still recognize that there are all these neural processes and the cells behave in a certain way but that there is an organism that is the result of all of this seems like an extra step it is metaphysically costly to go to the behavior of a complex individual and so from there if that is true then it's not purely a whole constitution relation but that you only have parts could the pluralism lie in this thing you add which will always be the same in different disciplines in the case of measuring time it's more like an epistemological factor you add in the case of... it's a factor of recognizing that there is an intentional being with some property as well that could be different to all disciplines and that could maybe be the stuff that makes the manipulability relation also domain dependent would that make any sense in your view? so the extra stuff I'm not sure what extra stuff you have in mind I mean of course processes will have many properties so first of all science is typically interested not in individual token processes but in types so neuroscientists are studying a type or types of processes types of activities they do this of course by experimenting on token processes but the goal is to study type now of course you can have in the same type you can have processes that are manifested quite differently in different instances and that's also why I have a tendency to call the activities essential parts those parts that are essential for the process being of that type so you will have the non essential properties in any individual instance of a process and those are not part of the mechanism because they don't serve the explanatory goals they are not relevant to the explanatory goals now is that part of the pluralism possibly I would seek the pluralism at a more I want to say the terrible word fundamental level of individuation of processes the kind of things that Alexander mentioned individuation criteria and part of the criteria the decomposition principles I think that's where the deeper explanation if there is one could be found for this domain specificity of constitution short question okay so you I know you're friends of Doris and so when you talk about causation you immediately say you think there's a domain neutral I forget your word causal method or causation what can you say a little bit about that reasoning principles reasoning principles of causal reasoning so can you say a little bit more why you sign up by position to think there are domain neutral causal reasoning principles if you are friends of truth right well there are exceptions now I think domain neutral causal inference principle I mean things like mills method of difference and its modification so the idea that when you just change one factor in an experiment and leave everything the same then you can infer that whatever changes is an effect of that and so on so this kind of very general causal inference rule which is used I mean the whole causal graph theory which can be applied to wildly different things I mean to the economy of a nation as well as to an ecosystem or a human brain or an individual cell there are things going on that you can study by using causal graph theory as a tool so in that sense its domain neutral but you think causal graph theory a model is a theory or a tool it's a tool I mean the causal graph is a tool I would say used to study well natural natural world and the social world lets thank our speaker again