ComplexityBlog

ComplexityBlog http://complexityblog.com/blog/ en-us Nucleus CMS v3.51 © Weblog http://backend.userland.com/rss http://complexityblog.com/blog//nucleus/nucleus2.gif ComplexityBlog http://complexityblog.com/blog/ Studying Complexity vs Studying Complex Systems xml-rss2.php?itemid=91 types of inter-disciplinary research, I now will go into more detail on a related topic: the difference between studying particular systems that happen to be complex, and studying complexity itself. The main point is that complexity theory includes several commitments related to levels of organization and to there being shared principles/mechanisms underpinning the dynamics of disparate systems. Studying complexity is the overt researching of these commitments and underpinnings. However, most scientists that describe themselves as doing complexity research are not doing that. Instead they are studying particular complex systems and typically ignore the commitments and underpinnings that define complexity science.As my first example, imagine a generic agent-based model of some social system (you can think of Schelling's segregation model as an example, or your favorite if you have one). Because the agents act autonomously via a set of neighbor-contingent rules, the resulting model is some kind of complex system (see later posts for types of complex systems). And the real-world phenomena it models is also a complex system; in the Schelling case it's people moving in and out of neighborhoods. Using this ABM of a complex system we want to better understand an aggregate property of the system; in the Schelling case that's the resulting segregation patterns. When you are using a ABM like that to study an aggregate property like segregation then you are using a complex system to study a domain specific phenomena, and that is a different project than using the complex system to understand how patterned macro-level phenomena are generated by micro-level agent behaviors.

The above is a case of people studying a complex system (society) with a complex system (an agent-based model). Not all complexity research does even that. In many cases a complex system is modeled as a set of equations, a static network, or a set of descriptive statistics. All of these approaches can be useful for revealing new insights into the modeled phenomena, and as such they can all be used to better understand complexity itself as well. An example would be discovering the number of disparate systems that produce power-law distributions for some property. Of course, there are a dozen different methodological explanations for seeing a power-law distribution, and the substantive explanation will be unique to each domain in which it was discovered. The point is that if you are studying why city sizes follow a power law distribution by investigating the growth dynamics of cities then you are (possibly) studying a complex system, but you are not studying the complexity of that system. If you are instead investigating why cities and war causalities and asteroids all have power-law size distributions then you are possibly studying complexity itself. I say "possibly" because most of the purported explanations I've heard for the observed similarities across systems also do not address any of the commitments or underpinnings of complexity science either.

To help clarify my point, I came up with the following list of projects for a recent presentation on the foundations of complexity science. I asked the audience (and now you) to vote on which is a project for studying complex systems vs studying complexity.

Build a model with rule-based interacting agents and measure the aggregate properties resulting from their behavior.

Model the relationships among objects/places/people and how those relationships affect aggregate properties across that system.

Capture the nonlinear dynamical behavior of a system's properties and measure the sensitivity of detected patterns.

Demonstrate how endogenous niche creation by evolving agents switches from favoring generalists to specialists.

Determine how system behavior differs when captured as a dyadic bipartite graph versus a hypergraph.

Of course it's a trick. The last one is a methodological question: it is not about complexity or a complex system, it's about how to best capture the system of interest. For the first four it depends on whether you are applying the model to a particular situation/environment/problem or to a general class of related phenomena (possibly across disciplines). That is, it depends on whether you are applying/harnessing the complexity of the situation to help answer your specific domain-related question or you are trying to answer a theoretical question about the complexity of the studied systems. It is not the model that makes the difference -- it's the questions being asked.

Some researchers believe that the best (or only) way to study complexity (or other general, abstract, unifying principles) is through copious examples. So, these people claim, by studying specific complex systems they are studying complexity itself, albeit indirectly. And that by seeing lots of specific complex systems they can generalize the commonalities and gain insight into those general principles. I do not disagree with this claim as a path to understanding complexity in a loose way, but it is not a way to research complexity as a scientific principle. To do the latter requires asking a different set of questions; questions that are not about the specific phenomena being modeled or the problem it was designed to address. Some such questions are the following:

Are there coherent behaviors at the macro-level (that are generated from the micro-level) that can be captured in its own set of behavior rules for the revealed macro-phenomena; i.e., can you transform this macro-level into the micro-level for a model of higher-level phenomena?

Can you change the interpretation of the model (without changing the model's construction) to make it about a different system? Or, how much do you have to change?

Can you provide equivalent explanations of the observed micro and macro behaviors at both the micro and macro levels? Are the inter-level translations lossy or lossless?

What aspects of the macro-level are insensitive to what details of the micro-level?

Can a different set of micro-objects and/or micro-behaviors also generate the same observed macro-phenomena?

The key area of inquiry for these questions is the micro-macro relationships. Another common (though less informative) way of phrasing these questions is to ask about emergent phenomena: What are they, how do they arise, and when do they appear? I avoid this terminology because people talk loosely about "emergence", and what I am proposing is to establish formally described relationships among the levels of a system's phenomena. Also implicit in the partial list of questions above is a special concern regarding the relationship between micro and macro behaviors. For example, in the Schelling model the macro-phenomena of interest is the observed segregation of agent types. That can be measured in various ways, but neighborhoods in that model don't "do" anything. Contrast that to the "glider" configuration in the game of life; it clearly moves diagonally across the screen via the on/off behaviors of the micro-level. Whether we call both phenomena "emergent" is not at issue; regardless they are clearly different kinds of phenomena.

That last statement has severe import: it means that Schelling's segregation model is a complex system, but it is not useful for studying complexity because all the behavior of the system is at one level (the caveat being that if you expand the scale enough, then with some parameter settings and rule combinations it looks like the neighborhoods grow, and so that growth behavior could be codified as macro-behavior). However, there are benefits to approaching a specific problem with complexity in mind (including modeling it as a complex system) even when the resulting model is not useful for studying complexity itself. Here are some questions you should ask when modeling a complex system if you are interested in harnessing the complexity of that system:

What objects/phenomena/behaviors to choose for the micro and macro levels?

Do you need more than two levels?

Which micro-level phenomena generate which macro-level phenomena?

How do the micro-level phenomena generate the macro-level phenomena?

How do you identify and measure the macro-level phenomena?

Is the relationship between the micro and macro levels underdetermined or overdetermined?

I think many people who are self-described complexity enthusiasts are often addressing these questions (although typically in a very basic way), and they take that to be doing research on complexity. What I am claiming is that they are instead doing research within complexity, but not studying complexity itself. And my observation is that more and more models, and the questions the models are built to answer, are of this variety. Good for addressing questions about a particular complex system, and perhaps abstract enough to apply to multiple systems in a non-trivial way, but not rich enough for direct research into the system's complexity. And even when a model could support such inquiries, those inquiries are not made. This may just be part of the general trend in science of pushing away from basic research. It may also be a symptom of the conceptual mess that complexity science is still stagnating in. However it does neither the field nor the particular researchers justice to call all this research by the same name. Asking questions about the theory of complexity and asking questions using the current (messy) theory of complexity is as different as theoretical physics and engineering.

Yet, despite the rise in complex systems thinking and of complexity models, there has been little progress on understanding complexity itself. And, I claim, this leads to an insufficient ability to harness complexity and use a system's complexity to improve our understanding of the complex systems models we build. We have tantalizing hints of the promise of complexity to unify our thinking across many disciplines, but it remains merely a promise as long as we fail to engage the necessary theoretical research. And the first step to making this theoretical progress is to recognize the differences outlined above; ask different questions and build richer models.]]> Commentary xml-rss2.php?itemid=91 Sun, 31 Mar 2013 09:38:08 -0500 Types of Inter-disciplinary Research xml-rss2.php?itemid=90 not trying to correct the way people talk about their research, or impose my own nomenclature on others. I don't care which words people use for these categories. The point is that these are different kinds of projects, and so they require different skills and involve different approaches. I chose different words that I think fit for ease of reference only. One purpose to this disambiguation is to determine which of these approaches, if any, constitutes a core element of complexity science. Another (related) purpose is to defuse overblown and unsupportable claims about particular research projects and their relation to complexity (see Studying Complexity vs Studying Complex Systems coming up soon).

First, research that isn't inter-disciplinary is:

Intra-disciplinary: research addressing questions/problems within one established discipline, and using techniques intended for that discipline.

This is not to claim that the topics/questions/problems within the disciplines do not change over time, or that the methodologies do not progress. Of course they do. However, an economist using game theory to study banks or a physicist using simulations to study condensed matter are clearly within this category.

I have identified the following four distinguishable approaches to inter-disciplinary research (there are probably more, but this is a good start). Also note that a single research project could be multiple of these types.

Cross-disciplinary: using a technique or insights from one established discipline to address questions/problems within another distinct established discipline.

Multi-disciplinary: research requiring insights from multiple disciplines to address problems/questions shared by those disciplines.

Trans-disciplinary: a problem/question or insight that has useful applications in more than one established discipline.

Extra-disciplinary: research that does not fall within any established discipline or combination of established disciplines.

Applying these categories to research is clearer for some than others. For example, using game theory to study animal behavior or airplane wing design was once clearly cross-disciplinary, but after it has been done and published it seems that game theory became a technique within those fields. Using ideas from population ecology to understand inter-bank loan networks seems to count; as does using fluid dynamics models to understand traffic flow.

Multi-disciplinary research is easier to spot and recently quite common. Combining gene expression with cognitive/behavioral studies is a clear example. Physical models of protein folding is another. Developing self organizing materials also falls into this category, and could also be cross disciplinary insofar as they are sometimes bio-inspired (e.g. artificial gecko-like adhesive materials). Political economy and political philosophy seem to fit here as well. Both fields contribute to and benefit from the research. Neuro-psychology fits this too, but in a different way: both fields benefit, but the benefits are distinct rather than shared.

Trans-disciplinary research can occur any time a basic insight has implications for other fields at a higher level of organization. If we take genomics as a basic insight, it has implications for microbiology, physiology, population ecology, social evolution, etc. But it can also be across domains at the same level; e.g., learning an important feature of human decision bias (a psychological insight) can affect research in economics, political science, marketing, ethics, etc. In each of those cases the flow of insight is one-directional.

Extra-disciplinary research seems to be quite rare, and this should be expected. Once some novel research thrust is initiated, it tends to attract similarly interested scientists, and quickly blossoms into another field or subfield. The nascent field of information science is a clear example, and the study of artificial intelligence was formed like this not too long ago. So this category includes research studying new phenomena that are not already covered by another domain (in the way that nanotechnology is covered by physics, chemistry, and engineering). Complexity, if there is such a distinct and independent phenomena to study, would now be in this category. Many people study particular complex systems with the belief that there is a shared phenomena, but few study the complexity of those systems directly nor attempt to make connections across disciplines.

Methodology work is also often extra-disciplinary, although it's clear that there should be an exception for methodological fields like computer science and statistics, and for particular methods like simulation, network theory, and equation models. These are, of course, often both extra-disciplinary and trans-disciplinary according to the above distinctions, but just because a person includes (say) a network model in their research shouldn't qualify it as inter-disciplinary. It must be the case that the network model is somehow used to port results from the home domain to other domains. An example of this porting is comparing power-law relationships across words in books, book sales, city sizes, star sizes, etc. Work that actually makes use of these connections (rather than just stating them) would count as trans-disciplinary on this account. Developing new measures for hypergraphs does not fall within any established discipline, and would benefit many of them. But network theory does have its own journals and conferences, so it is becoming its own methodological field in the way that statistics is. Some people study statistics itself, and others use statistics to study specific systems within established domains.

Okay, I hope that is all clear, and now back to the original question: in which sense or senses is complexity science "inherently inter-disciplinary"? The short answer is that studying complexity, and the results from that research, would fill all of those categories. Even though complexity has its own journals and conferences, it is not "owned" by any existing field, thus it is extra-disciplinary. If one gained insight about a fundamental feature ordering the behavior of complex systems, then it would have implications across many disciplines – all the fields with systems of that type – and would therefore be trans-disciplinary. Because complexity is a general principle of ordered system behavior, the problems/questions of complexity are shared across many fields – different fields for different problems – and therefore qualifies as multi-disciplinary. And, the discovery of such a general organizing principle would allow insights from one field to apply other fields, so it is also cross-disciplinary (or maybe it just facilitates cross-disciplinary research the way methodology does).

However, much of the research done under the name of complexity is intra-disciplinary rather than inter-disciplinary. Take as a common example a physicist studying social networks: the research crosses no discipline boundaries. There are no insights from physics being applied, just a method that some physicists are proficient in being applied within a social science discipline. If you disagree, then consider whether a sociologist using social networks is in any way doing physics. Thinking about city growth using analogies from biology seems to count as inter-disciplinary, as does finding commonalities in flocking birds, stock market trades, and political opinions. But flocking birds, schools of fish, and crowds of humans are all in the same field: animal social behavior. This is not to downplay that research in any way. And I do believe that studying particular complex systems may occasionally lead to insights regarding complexity itself. But on deeper analysis, studying any particular complex systems, even one outside the discipline in which a person was originally trained, will typically fall under in the category of intra-disciplinary. So complex systems research is not inherently inter-disciplinary, although it sometimes is...it just depends on what one studies and how.]]> Commentary xml-rss2.php?itemid=90 Tue, 26 Mar 2013 11:59:05 -0500 Science Is More Than Solving Puzzles xml-rss2.php?itemid=89
When putting together a picture puzzle one works with and against the constraints of the pieces. It is rare that a piece could go in multiple locations, so you have to find its correct spot. Those same constraints can help you; for example, the corner pieces are the most constrained, you know where they must go first. Then edges must connect these corners and you typically know their required orientations. A salient feature of the image might make a particular cluster easier to fit together than other parts. Et cetera. Different types of puzzles require different methods to solve because they have different constraints. One puzzle is considered more difficult than another because it requires more sophisticated methods. A puzzle with more pieces requires more tenacity, but not specialized abilities. Regardless of one's skill level, with enough time and effort every puzzle can be solved by anybody who is capable of moving within the constraints. Hours and hours later you end up with an image that looks worse than a $5 poster of the same thing, but you did it yourself and you feel proud and accomplished.

Now I will explain how most scientific research can be seen as solving puzzles. Research science starts with either a specific problem or a particular phenomenon that needs explaining: Why do banks collapses cascade? Can one make a compound with inert atoms? Were Homo habilis capable of verbal communication? What role do bus routes have in disease spread? Et. cetera. This is akin to determining what picture to use in the puzzle. The discipline covering that problem or phenomena has established methods for solving problems: e.g. collect and analyze data, design an experiment, or run simulations. Different techniques are used for different questions even within the discipline, but there is an established way to do things.

The problem or phenomena to be explained is the solution to the puzzle, the methodology of the field determines what the pieces of the puzzle are, and that methodology combined with the available components determines which heuristics are appropriate. Often there aren't enough pieces to solve the desired puzzle (although that rare stops scientists from trying). Sometimes a problem is attacked with multiple techniques, some performing better and are more interesting than others, but the progress can rarely be combined. If the problem can be solved, then it is only a matter of time and effort to find the solution. If the problem can't be solved, then either you need more pieces or you chose the wrong kind of methodology for that problem (wrong puzzle time for that image). Likewise if it's too easy to solve (because then it's not interesting).

Many scientific projects have real consequences for policy, technology, etc. and so the solution is more than just a crinkled picture. But what scientists do, and what they are rewarded for, is picking the right methodology for a problem and then following the required steps to navigate the constraints and reach the solution. Nothing about this process is creative. The constraints of the problem fully determine the best analysis methods, and one must only follow through with that method's protocols. The method may not be sufficient, and some trial and error manipulation might be required to fill in gaps. Occasionally the methodology is extended to include protocols for filling gaps of that sort. This isn't real scientific progress, this is just getting better at solving puzzles of a certain type by solving puzzles of that same type. It is methodology progress, though, because it allows others to solve more complicated puzzles more easily.

What people call "interdisciplinary" research is usually just using a solution heuristic from one kind of puzzle and applying to another kind of puzzle. This is usually obvious, but sometimes one can transform the problem in a way to make this insightful. If you can translate a picture puzzle into a kind of maze, and solve it like a maze, then that's interesting. That's also a methodology problem -- how to analyze and manipulate the information -- and not a new result about the problem domain. For example, I have a project to translate N-person games into network structures that can then be "solved" by finding the optimal network flow on the resulting structure. This will turn certain situations that are unsolvable by current game theory techniques into ones that can be solved. Pictures that would not be good for current puzzle types would become fun and interesting with the new constraints.

However the potential for new and interesting puzzles is not valuable to scientists. Methodology research is only valued (these days) if you can already demonstrate that there is a problem people have been unable to solve that can now be solved. Of course people only care about problems they can already solve (or think they can), and what counts as methodology research is merely refining the existing heuristics to make more complicated puzzles of the same types already employed. But even that is rare most science is just applying existing tools to a well defined problem, going through the motions until everything fits together the way it's supposed to.

This is contrast to what basic research is supposed to be. Basic research is comprised of high-risk, high-payoff projects that explore new possibilities and totally new questions/problems. Not just problems that can't currently be solved, but problems that haven't been considered. And to solve these new kinds of problems one needs new methods. It's akin to inventing the first maze or the first sudoku. Once the field is discovered, it opens up new avenues of research. Unfortunately, because of how deeply the puzzle-solving mentality is ingrained in current academic training, most scientists can't even recognize a kind of problem they haven't seen before, let alone see why it might be interesting and worthwhile to pursue solving it.]]> Commentary xml-rss2.php?itemid=89 Mon, 18 Feb 2013 07:59:43 -0500 Converting CSV to Netlogo Lists xml-rss2.php?itemid=88

file-open "MedianIncomeData.csv"
while [not file-at-end?] [
let thisline file-read-line

The lines will not read in as you want them; instead you will get a string like this:

"36803 , 38260 , 39702 , 42463, 42183 ..."

I don't know why there are sometimes extra spaces and sometimes not, but just get rid of all the spaces with

set thisline remove " " thisline

Netlogo delineates items in lists with spaces instead of commas, so you need to replace each comma with a space. Unfortunately there is no built in way to do this. There is a String extension that includes this feature, but I couldn't figure out how to get that working (I mean, how to install it). So, I just wrote simple Netlogo subroutine to do this for me:

to-report replace-all [string1 with-string in-string]
while [ member? string1 in-string ] [
let index position string1 in-string
set in-string replace-item index in-string with-string
]
report in-string
end

Be careful, in case you want to use this string or list replace-all function elsewhere that the thing you are replacing with cannot contain the thing being replaced. For example, you can't replace "," with ", " using this because you'll be adding back the comma it replaced and it will run forever (in that case just adding a ton of spaces after the first comma). Anyway, though it's not the perfect general solution to replacement, it's good enough for present purposes.

With that in hand, you can convert the line of CSV, imported as a weird string, into a string formatted as a Netlogo list line with

set thisline (word "[ " (replace-all "," " " thisline) " ]")

Which gives you something like

"[ 36803 38260 39702 42463 42183 ... ]"

Just have Netlogo read that string as a line of input, the list as a real Netlogo list, and attach it to the variable storing your list of lists in Netlogo (in my case MedianIncomeData).

set MedianIncomeData lput read-from-string thisline

All together it looks like this:

to import-data
set MedianIncomeData []
file-open "MedianIncomeData.csv"
while [not file-at-end?] [
let thisline file-read-line
;show thisline
set thisline remove " " thisline
set thisline (word "[ " (replace-all "," " " thisline) " ]")
;show thisline
set MedianIncomeData lput read-from-string thisline MedianIncomeData
]
file-close
end

You'll need to do more tweaking if you want to combine strings and numbers, but the basic building blocks are there. This will make it faster and easier to import CSV data into Netlogo instead of using Excel or Perl scripts to rewrite your data as a txt file that is already Netlogo friendly for input. Though currently I'm using this for ordinary time-series data, it is also clearly useful for other data coming out of other programs; such as adjacency matrices, Markov matrices, agent attributes, etc.]]> Computer Science xml-rss2.php?itemid=88 Mon, 28 Jan 2013 01:54:19 -0500 Emergent Causal Inversion xml-rss2.php?itemid=87 emergent causal inversion.The project starts with a simple agent-based model at the microlevel. The model could be something as simple as disease spread along a network, a segregation model, or a majority imitation model. Perhaps something more sophisticated is needed to generate the desired effect, but I don't think so. As the model runs, track several variables, including both aggregate variables over the agents and properties of the macrolevel itself. This is done, as per usual, over a few hundred or thousand runs of the model to build a dataset of time-series of all these variables.

The next step is use a statistical modeling technique specifically designed to uncover causal relationships in data of this kind: Bayesian network analysis. First we build the Bayesian network of the microlevel variables using the causal relationships implied by the agent rules we specified. Then we use software to uncover (using advanced AI techniques) the structure of the Bayesian network for the macrolevel variables. We also need a reduction bridge law that translates macrostates into microstates; i.e., what microlevel variables values are consistent with which macrolevel variable values.

Once all that is complete, what we are looking for is a causal relationship in the macrolevel Bayesian network that contradicts the causal mechanisms we know are operating at the microlevel. Just to be clear (because there is a lot of confusion around the term "emergence") this is not intended to reveal how the macrolevel causes behavior at the microlevel. Causation at each level is independent, therefore such a claim is just nonsense even though such claims are often made (e.g. the level of segregation in this neighborhood caused the agent to move to another neighborhood). That is not the unintuitive contradiction we are looking for.

What we want to see is that at the macro level, the value of variable A affects the likelihood of variable X and X doesn't affect A: A->X in the Bayesian network. The variable A happens when the microstate has values a, b, and c. The variable X occurs when the microstate has values x, y, and z. Yet at the microlevel, through the rules we specified, we know that it is actually the case that x, y, and z have an affect on a, b, and c rather than the other way around. Meaning that given the data we collect from the system, the apparent causal structure at the macrolevel is incompatible with the causal structure at the microlevel. And this could all be shown in a totally formal and rigorous way.

This is important because THAT property of the macrolevel, that it has distinct causal relationships from the microlevel that it can be reduced to, is perhaps the best candidate for a truly emergent behavioral property...of a certain kind. Because the macrolevel can unambiguously be reduced to the microlevel, that is not the issue here (though the most common way to define emergence). The point is that the behavior of the variables at the macrolevel cannot be reduced to behavior of the microlevel, even though all the states can be. And it's even better because the behavior can be reduced in principle, but the reduction reveals conflicting relationships.

The microlevel is where all the real causation is in the model/data, so the apparent causal relationships in the macrolevel are "incorrect" in some sense, but not in the sense that they fail to predict future values at the macrolevel. If it were done right (i.e., if it is possible to actually get a model to generate this emergent causal inversion) then basing the macrolevel causal structure on the actual microlevel structure would make predictions worse. What this means is that macrolevel objects literally behave differently than is implied by the mechanisms that generate them. And that, I think, would be a huge success for complexity theory and the hunt for emergent phenomena. ]]> General xml-rss2.php?itemid=87 Tue, 22 Jan 2013 10:15:13 -0500 Multilevel and Multiscale Selection xml-rss2.php?itemid=86 this article by Steven Pinker. Though I don't think everything there is quite right, I am not going to go into that now. I am going to focus on a compatible analytical concept that I call multiscale selection. The idea is linked to multilevel emergence (an open problem in complex systems). The drivers of evolution (the replicators) only exist at one level, because the casual forces of any system can only reasonable exist at one level (though you can pick whatever level you want to consider). Phenomena at other levels of organization are just different descriptions of phenomena occurring at the causally efficacious level. Some phenomena are easier to recognize (by humans) at certain levels, and we build a conceptual hierarchy out of these levels (although it's not really a hierarchy). As a result selection pressure seems to occur at multiple levels, but that is not a necessary or supported model/theory of evolution.

I use "multiscaled selection" to mean that the process is recognizable at many scales, and "multilevel selection" in the standard way that selection operates at many levels. Scales and levels have a similar meaning here; so cell -> organ -> body can be seen as both increasing level and increasing scale (the difference is that levels are determined by the appearance of coherent phenomena, whereas scales can be defined arbitrarily). The salient difference here is that "selection" describes a behavior, and a necessarily causally efficacious one. Selection can only happen at one level because it is causally efficacious and causes can only operate within a level (i.e., not across levels). Although certainly selection at one level may produce recognizable changes in the proliferation of features found/defined at different scales. This is what the group selection and multilevel selection advocates (typically) want to deny...that selection can only happen at one level. We can explain all the social phenomena under consideration, and a great deal more, without the multilevel section pressures; therefore Occam's razor should lead us to abandon it. Here's how to generate multilevel and multiscale selection and compare them.

To keep things simple, let's assume that a level-1 group consists of two agents, and a level-2 group consists of two level-1 groups, and a level-2 groups contains two level-1 groups. Selection requires some property that affects replication. At the agent level, let's give all agents the property + or –. Level-1 groups with ++ inside have property A, groups with +– or –+ inside have property B, and –– groups have property C. Level-2 groups can only contain level-1 groups of AA, AB, AC, BB, BC, or CC (assuming order doesn't matter), and these combinations are assigned F1, F2, F3, F4, F5, F6 respectively.

Each higher level property is determined by its lower-level constituents, but these are not lossless bridge laws. Note that you can reduce any level-2 group all the way down to possible sets of agents: an F3 groups contains AC groups, which contains ++––. But given the unordered set ++––, it could also be +–+–, so you don't know if that's AC or BB, hence F3 or F4. Decreasing scale (reduction) is unambiguous, but increasing scale (emergence) is ambiguous even in a simple formal model like this.

In a good complexity model, all the behavior would be programmed at the agent level, and higher-level phenomena generated from that. So there would be rules for what +s and –s do by themselves and when they interact. Perhaps they interact with multiple other agents at the same time...or (let's say) just one at a time. At any given time, whatever pairs exist will have the property A, B, or C. But what would identify the level-2 properties if there are no pairs of pairs? If you want another level you have to add something to the model.

One thing you could do is allow agents with partners to join with other agents with partners; that allows pairs of pairs, but you might also get chains. So you limit this bonding of pairs to other non-bonded pairs. That's an ad-hoc rule to get what we want for sure, but it may actually be appropriate for some kinds of agents. Okay, that generates level-2 groups of pairs of pairs of agents and each one will have exactly one property F1-F6. You can then specify other agent rules such as preferential attachment and detachment and/or birth and death processes and produce some selection dynamics. For example, 1) assign a rule that all agents will connect to a + if one is available (otherwise a –) 2) in every iteration each agent makes X copies of itself, where X = #of+s connected to it, and 3) each agent lives 3 iterations. In most cases the population will very quickly explode with + agents, A groups, and F1 groups...multiscale selection without needing to specify rules beyond the agent level (I did something like this in my dissertation for the Biased Lane Choice game, but only two scales).

Another thing you could do is identify groups of agents, and specify rules for duplication at all three levels: for example, 1) half of the agents in F6 groups die, 2) F1 groups duplicate themselves, 3) A groups join with other A groups 75% of the times, etc. Those are all phenomena that the bottom-up agent-based model described above would produce, but without requiring the higher-level selection rules. This second, higher-level type of rules are what group selection (sometimes) is claiming. Something like "when two level-2 groups are competing for +s, the one that better fosters +s and + acquisition will flourish and reproduce at the expense of the other group, and having more +s in the group is that property that makes it better for +s." This is a description at the higher level of what is happening at the lower level, and the agent rules can explain this, but if you define this rule at the higher level, then you've lost all explanatory power.

Of course, if you don't care about explanatory power, and you just want to create different (possibly competing) dynamics at multiple levels without figuring out how to get the macrolevel or mesolevel phenomena from the microlevel, then this is a way to explore the effects of multiscale selection. That model will be conceptual nonsense, and I would argue it will likely also be useless for any purposes because it fails to preserve conservation laws in causal efficacy. But there may be some specific questions and narrowly focused problems that such a model can address.

My point is that the thing that most people are actually interested in is the multiscale selection generated by microbehaviors, and there is a major difference between identifying phenomena at higher level (because they are interesting, measurable, and comparable to data), and defining rules for those higher level phenomena as causally efficacious. Downward causation is worse than wrong, its confused and unnecessary. So is causation at multiple levels/scales in the same model. If you want to explain the phenomena you generate in a model via the mechanisms that may have actually generated those phenomena in the world, then selection (and causation in general) can only be at a single level...choose wisely.]]> Methodology xml-rss2.php?itemid=86 Tue, 22 Jan 2013 03:31:20 -0500 Call for Participation: Swarmfest 2012 xml-rss2.php?itemid=85
This year we will continue to examine the range of systems being modeled with ABM, with the possibility of providing some guidance as to the suitability of the various types of agent-based models for particular types of modeling goals. We also encourage the sharing of our experiences with attempting to gain acceptance of ABM within our own research communities, and discuss strategies where cross-domain examples/analogies can aid in that process. We will also try to identify future avenues for ABM research, including the "next" generation of ABM tools, platforms and application.

Go to the Swarm Development Group homepage for upcoming news on the location facilities, program and logistics and to get added to the mailing list.

I won't be able to go this year because of a schedule conflict with my teaching at ICPSR in Michigan, but I have gone each year for the past few years and it has always been an interesting collection of research and a supporting group of individuals for advancing modeling and showing appropriate applications.]]> Commentary xml-rss2.php?itemid=85 Tue, 22 May 2012 01:51:57 -0500 Models for Philosophy xml-rss2.php?itemid=84
Yet I want to call this process of increasingly clarifying concepts and making distinctions conceptual technology. And conceptual technology is the "product" that philosophy has to offer other fields, to improve them both academically and practically.

Certainly conceptual clarity is useful in making models (e.g. understanding that the species concept is vague in an evolutionary model affects the form and use of the model). But how can a model of some phenomenon (e.g. of evolution of species or liquidity) impact our understanding of that concept (e.g. liquidity) or its properties (like vagueness). Think also about the evolution of the moral experience: how does the evolutionary explanation for our moral reactions and experiences affect our understanding of morality as a philosophical notion? In general, it is indisputable that clearer concepts can lead to better science, but the impact of more refined science on philosophy is much less clear.

The way I usually think about this is that once it is possible to build a model or perform an experiment on some phenomenon, then most of the philosophical work on that phenomenon must have already been done. Our understanding of the concept of liquidity is not complete, and though many experiments utilize liquids and the fluid properties of liquids, there are no tests that I can think of to distinguish features of the liquidity (how liquid it is). But on further thought there are things like viscosity, pressure, throughput, laminar vs turbulent flow, etc. We can test for these features of liquids without understanding what it really means to be a liquid or makes something a liquid. There is a real sense in which these feature are part of what it means to be a liquid and how liquidy something is.

But in our philosophical understanding of liquidity, the scientific facts of flow and pressure and viscosity seem unhelpful. The concept of liquidity includes these features, but our understanding of liquidity as a concept doesn't seem helped much by knowing (for example) the viscosity variation under different temperatures and pressures. Our normative concepts seem unaffected by any account of the evolutionary fitness of certain reactions to behaviors. Our philosophical questions seem to be about exactly those things that we can't collect data about. But that's jumping to the conclusion.

I am more and more convinced that if you have a philosophical question, and you can build a model to test it, then you didn't really have a philosophical question. If, on the other hand, you think that making a philosophical distinction can improve some scientific result, then of course you could make a model to test that by making different distinctions in the models. And what I really want to think more about it, can we have two theories about some phenomenon, build different models of that phenomena, and then the results of those models refines our concepts of how things work?]]> Philosophy xml-rss2.php?itemid=84 Thu, 17 May 2012 07:31:12 -0500 Levels of Organization are not Hierarchies xml-rss2.php?itemid=83 level of organization. This is an epistemic notion because it relates to what we can capture in models, not to the structure of reality. Hierarchies of scale are a mixed ontological and epistemic notion: the levels are still based on what people can discern, but there is an added ontological assumption that a higher level exhaustively includes the elements of a lower level. Now I will show why levels of organization, not hierarchies, are the domains and ranges of scientifically useful reduction and emergence relationships. One way to think about this is to separate strict hierarchies from roughly hierarchal organization. If one goes strict than phenomena at any level can be broken down into phenomena at any lower level. We can loosen this requirement by only requiring that the part-whole relationships follow an ordering even if not every level has the same descriptive power with respect to every other level. That's an improvement because it allows for cases like the following: A is higher than B which is higher than C. A cannot be reduced to B, but both A and B can be reduced to C.

Although, consider that we may conclude that this relationship among A, B and C implies that A isn't really higher than B, though they are both higher than C. If we find some level D that is between A and B such that A can be reduced to D, and D to B, then doesn't that also imply that A can be reduced to B via D? Seems so. When people/scientists consider this issue, their notion of levels seems to be limited to constitution: what parts make up the whole in a physical sense. And as long as we are talking about physical objects in the world, the part-whole relationships keeps looking strictly hierarchal.

But now consider this: bodies are objects, cells are objects; and since bodies are higher up on the hierarchy, bodies are made up of cells. But bodies are not made of just cells. There is water and nutrients and other chemicals in a body that are not parts of (or even inside) cells. And these molecules are at a completely different scale than cells – in fact cells are made up of (reducible to) these very objects. So this is a case in which bodies and cells can both be reduced to molecules, but bodies cannot be reduced (in the ontological sense) to only things at the cellular level. This is so even though a body contains all the molecules that its cells are made of...the cells are a proper subset of the body. But we can pull this same trick again and point out that there are parts necessary for the body that are not part of any molecule, but are rather in the domain of submolecular physics. We can always do this, and there is always going to be something missing.

This leads us to the issues that there is no lowest level and that no science is complete, but these are not the current topic. I'm more interested in taking this toward part-whole relationships that are not constitution. Translating phenomena from one level to another is typically a lossy compression, some of the details are left out. This is true whether one is translating macro to micro or micro to macro. But here comes the real kicker, the phenomena we are translating are not always objects.

A flock (e.g. of birds) is an object at the macrolevel, and flocks are made up of birds (which makes it look like an ontological hierarchy), but the flock object is clearly not reducible to its constituent bird-object parts alone. The behaviors of the birds are a crucial feature of the flock object-phenomenon, but behaviors are not part of ontology. Thus when we are talking about emergence and reduction of phenomena we must go beyond just ontology and therefore beyond the notion of constitutional hierarchy. (Note: I wonder if this is all that is supposed to be meant by the slogan "More than the sum of its parts.")

When we develop a translation between phenomenon A and phenomenon B, it is sometimes the case that some of the objects of B are some of the parts of A. In this case a translation from A to B is called a reduction and a translation from B to A is called an emergence. I identify/define the levels of organization as the ranking of all ontologies that figure into a reduction or emergence relation. Thus the levels will not be strictly hierarchical. The levels will be refined and filled in as new models/theories demonstrate new reduction/emergence relationships. And although any particular translation is between two levels, there may actually be a continuum of levels.

(Afterthought: I want to be clear that some people may use the word 'hierarchy' to mean what I refer to as a level of organization. That is, they may not subscribe to the ontological requirement, or the demand that each layer can be losslessly described at each other, but refer to the levels as hierarchies. Fine. I don't care about which words people use. This isn't a semantic issue. The truth is that some very good scientists are not comfortable with these concepts, and have never even considered the differences. It is likely that many people's concepts of reduction and emergence include the ontological requirements of complete hierarchies, and they do not distinguish between applying them to the world or to models. Disabusing people of this sloppy usage is part of the point.)]]> Philosophy xml-rss2.php?itemid=83 Tue, 6 Mar 2012 07:34:25 -0500 Robustness and Fragility of Complex Systems xml-rss2.php?itemid=82
For the robust/vulnerable dichotomy we can take the human body as clear example of a complex system. It is extremely robust to certain insults, like the removal of a limb or a lung, which are large and disruptive, but can be dealt with. On the other hand, there are some specific regions in the brain and heart where the removal of just a cubic millimeter of tissue results in the immediate death of the individual. Given an average body mass of 70kg, that's just 1/70,000,000 of the total volume. That sounds like a fragile system to me. We can contrast that to a society of 70 million people that can survive the loss of any, and perhaps any quarter, of its population.

We can say that a system is robust with respect to some property (like being alive or perpetuating) when that system has that property over a large proportion of changes; i.e. the overall likelihood that the system will have that property is high. And we can say that a system is vulnerable to a particular change if the likelihood of loosing that property is high for such changes (and fragile if it vulnerable to at least one change). And from these definitions it is clear that a system can be both robust and fragile (with respect to some property). However, this does not mean that every system is equally robust and vulnerable…or robust and vulnerable through the same sorts of changes. Differences in robustness and vulnerability characteristics can be used to establish equivalence classes of system dynamics and thus allow us to categorize systems (or at least models of systems) by how they change over time.

Instead of offering exact criteria for systems to be complex, the style has shifted to listing some common features of complex systems, and features of systems that make them complex. Robustness is typically on these lists, but vulnerability is typically not. Part of the reason is political; practitioners of complexity science want to make it sound good. A better reason is that it's the complexity of the system that makes it robust, but systems are vulnerable where the complexity fades. For example, brains and hearts and eyes are very specific, optimized organs that are more like designed systems and less like self-organizing systems than other tissues, and that's why they suffer from single-point failures.

This is merely an unsatisfactory ad hoc rationalization to justify the connection between complexity and robustness; i.e. it's a political explanation. The truth is that at the current state of the art, robustness applies to some complex and some non complex systems, and fails to apply to others. Until some specific mechanistic connection between the complexity of systems (measured somehow) and their robustness and vulnerability characteristics is established in a general way, the claims are mere rhetoric. And furthermore, once this work is done, I'm sure we will find that robustness is neither a necessary nor sufficient feature for systems to be complex (or vice versa). It is a contingent property, and working out when systems are and are not robust will be interesting and difficult research.

Finally, the claims relating robustness to complexity often make unfair use of the lack of specificity of the two terms. My research on formalizing robustness has revealed that there are many distinct meanings that, though related, indicate very specific dynamics. A property can be maintained without being lost, regained if lost, obtained if not present, or present over a large proportion of possible futures (among others). Saying that a system is robust, without specifying the way in which it is robust, offers little to no insight into the relationship of that property and system dynamics.

Complexity is even more vague, and identifying what characteristics make systems complex is fraught with confusion. Something like: collections of at least an intermediate number of components that interact in such a way that self-organization and/or adaptation is possible in some situations. If that's the kind of thing that people mean by "complex system", then certainly it is folly to claim that such systems will be robust in general, or that robust systems will be complex in the sense of satisfying that loose definition. My guess is that people who consent (or acquiesce) to claims that complex systems are robust will retract their support when expressed in these more detailed (but equally vague) terms.

So it seems that under closer inspection complex systems are no more characterized by their robustness than their fragility, contrary to many unfounded claims made by practitioners of complex systems modeling, and expounders of complexity theory. This makes our whole field seem unscientific and foolish. For the field to mature we need specific, testable, and falsifiable claims about the relationships of system properties like interaction structure, scale, rates of change, behavior rule types, etc. and properties of system dynamics such as tipping points, robustness, path dependence, and adaptability. I've started on some of this work with my measures of some of these phenomena, but the scale of this project is science-wide. Building and using the appropriate formal tools is an important first step, but it is even more crucial to convince people of the importance of thinking and speaking clearly on these issues. ]]> Philosophy xml-rss2.php?itemid=82 Wed, 22 Feb 2012 07:14:09 -0500