|
Author
|
Topic: Constructing a relevant Avida experiment
|
Micah Sparacio
Member
Member # 6
|
posted 21. April 2004 08:36
RBH has recommended that IDists take advantage of the Avida environment and construct some Avida world to test various ID hypothesis.
This summer, I'll have quite a bit of free time to work on such a project and would like to start Brainstorming with others (in particular RBH) on what such an experiment might look like.
I'll let RBH have first dibs since he's the one who mentioned the idea. I'm particularly interested in doing this in phases, where at the beginning we focus on the conceptual issues: what issues such a project should be dealing with. After dealing with the conceputual issues, maybe we can start moving towards implementation and even have periodic online development meetings.
What do you think RBH? Is this viable?
[ 21. April 2004, 08:38: Message edited by: Micah Sparacio ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 21. April 2004 12:02
Micah wrote quote:
This summer, I'll have quite a bit of free time to work on such a project and would like to start Brainstorming with others (in particular RBH) on what such an experiment might look like.
It depends on whether you want to do some actual ID research or continue in the "avida work tells us nothing about biological evolution" vein that Truman and Bracht are mining.
On the assumption that it's the former, I'd start with my sugggestion in the thread about Truman's critique, with the validation and calibration of the design detection/measurement methodology that has been offered by Dembski. What ID needs in order to become "scientific" is not more philosophical musings or "evolution can't do this!" posturing, but some actual systematics using the metric that IDists claim is capable of detecting design. The most striking thing about Darwin's career for me is the years and years and years he spent in close study of the systematics of the phenomena he was to ultimately explain. He spent six years studying how to distinguish among species of barnacles, fer cryin' out loud!
In avida one has available the whole evolutionary history of all the lineages of a population evolving in a controlled environment. Under those conditions a veritable zoo of digital critters come and go - lineages emerge, flourish for a while, and then go extinct. And all that information can be stored to disk. That's a huge resource for calibrating and validating the purported design detection methodology that is so far ID's only potential contribution to a scientific understanding of biological phenomena. One knows the independent specifications in Dembski's sense - the various logic functions - and in the stored data one has detailed records of the emergence of those specified structures in a wholly controlled system. ID as Dembski conceives it in Specified Complexity is essentially a claim about the history of a phenomenon that can be inferred from an examination of the phenomenon itself without reference to its history. One can calculate the specified complexity of a structure knowing nothing whatsoever about the history of that structure.
So in an avida ID study, one could initially calculate from first principles the specified complexity metric for each of the structures (code sequences) corresponding to the various logic functions without reference to their histories. Then one could go back and examine the histories, looking for correspondences between the various calculated SC values and the actual histories, to test the proposition that SC is a metric that really does tell us something about the history of structures. Does the metric distinguish among lineages that prosper and those that go extinct? Does it distinguish between evolved critters and human-written critters? Do the data generated by the Specified Complexity metric underpin an interpretable empirical taxonomy of any sort at all?
Until the core metric of the ID design detection methodology is validated against a wide array of phenomena of known provenance, any claims one makes based on it are suspect, so in my view that's the place to start an actual ID research program along Dembskian/Behean ID lines.
RBH
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 21. April 2004 12:57
RBH,
Yeah, I'd like to contribute something positive to the discussion, rather than prove some point (that evolution can't accomplish something). Thanks for the initial ideas.
What are your thoughts (or anyone else's) about applying Behe's modified, evolutionary definition of IC?
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 21. April 2004 14:11
Micah asked quote:
What are your thoughts (or anyone else's) about applying Behe's modified, evolutionary definition of IC?
Sure. That's also do-able in an avida context. In fact, I was thinking about that as I wrote my first response to the OP. Again, that's a metric -- the number of unselected steps -- and could be quantitatively examined in the avida system. One of Truman's criticims, in fact, is related to that issue - the 'intermediates' are held to be too close together in code sequence space - and that's a question well worth examining quantitatively.
One could look at the role of neutral drift in that context (yup, it occurs in avida to at least some extent). You'll recall that in the Lenski, et al paper they provided some information about beneficial (fitness-increasing), neutral, and deleterious (fitness decreasing) mutations for the case study lineage, and showed that in at least some cases a deleterious mutation was in some sense 'stage setting' for later mutations that produced an increase in fitness. Of course, if one examines all lineages (genotype threads) for an avida run, the great majority of deleterious mutations lead to eventual extinction of the lineage. Slews of lineages sink beneath the waves for every one that persists. The case study lineage was one of the lucky ones.
Again, I'd first work on validating and calibrating the metric, and then on systematics - generating enough data to begin to look for interpretable taxonomies.
RBH
[ 21. April 2004, 14:18: Message edited by: RBH ]
IP: Logged
|
|
Evan
Member
Member # 164
|
posted 21. April 2004 15:56
Micah, could you post Behe's modified defintion of IC? Thanks.
IP: Logged
|
|
Micah Sparacio
Member
Member # 6
|
posted 21. April 2004 16:20
Evan:
From: Irreducible Complexity
"An irreducibly complex evolutionary pathway is one that contains one or more unselected steps (that is, one or more necessary-but-unselected mutations). The degree of irreducible complexity is the number of unselected steps in the pathway."
[ 21. April 2004, 16:22: Message edited by: Micah Sparacio ]
IP: Logged
|
|
Argon
Member
Member # 276
|
posted 21. April 2004 23:08
Micah Sparacio writes (to RBH):
"What are your thoughts (or anyone else's) about applying Behe's modified, evolutionary definition of IC?"
First, I would like to make the suggestion that we don't call it IC. The 'modified, evolutionary definition of IC' is definitely not the same beast as the original formulation. If the original definition defines 'irreducible complexity' then the second would more properly be called "unevolvable complexity". It is far from certain that a system would be considered IC under both definitions.
Even the "count the neutral steps model" has serious drawbacks in terms of biological application because it requires substantial knowledge of the history of a system. Here's a simple case: Acquisition of streptomycin resistance & compensating supressors in bacteria. This involves a two-step process. First, exposure to an environment with streptomycin antibiotic selects for mutations which produce tolerance to the antibiotic. However, these resistance mutations carry a growth penalty that ultimately slows reproduction. In an environment with many strains carrying streptomycin resistance, a second mutation becomes selected which compensates and restores growth rates similar to that of the 'wild-type'. As a result, streptomycin-resistant strains can compete with non-resistance strains even in the absence of the antibiotic (This is one of the reasons why antibiotic resistant strains persist long after exposure).
Now, either of the two mutations (resistance and growth inhibition suppressor) alone in an environment without antibiotic leaves strains at a selective disadvantage. Because resistance + good growth requires specific double mutations, it is unlikely that strains would acquire both in a non-selective environment. Absent historical knowledge of how the mutations were acquired, one might incorrectly assume that two simultaneous neutral point mutations would be required, when in fact, they were sequential and actually involved selected steps that almost always produce similar sets of mutations.
Avida might be useful for this question by retaining historical pathways. However, if the number of non-selected steps in the development of an 'output' is small, what conclusions should we reach? Would it tell us something about 'evolvability' or more likely, whether a particular IC definition is really that useful?
[ 21. April 2004, 23:17: Message edited by: Argon ]
IP: Logged
|
|
Evan
Member
Member # 164
|
posted 22. April 2004 07:51
Back in February I posted something which is relevant to this discussion. Here is part of that post.
Given that the design inference is based on probability, I offer this:
quote:
Here’s is an example (I am not a biologist, so real biologists would have to supply the actual details.)
Start with a known bacteria population with a known genome. Put different populations into different cultures with different amounts of some harmful substances, or with different substances. Watch for the evolution of some resistance or other means of coping with the various substances.
Repeat the experiment to see if there is any consistency - does the population reach the same type of “solution” to the problem in approximately the same time and to the same degree? - or does this vary, and by how much.
Then analyze the genome of the new population to see what genetic changes were responsible for the adaptation. Make estimates of the probability of this happening, taking into account the number of bacteria, the number of generations, and the types of possible changes known to be possible or common with this bacteria.
The result would be an estimate of the probability that the adaptation under consideration would happen as a consequences of natural processes, moving us closer to having a reliable method for detecting design at the genetic level.
IP: Logged
|
|
Evan
Member
Member # 164
|
posted 22. April 2004 09:12
The second part of the above idea would be to then look at the substeps involved. If it took N generations for the adaptation in question to happen, look at the overall situation at uniform intervals, such as N/4, N/2, and 3n/4, for example. Look at both changes which lead to the end adaptation and those that don't (both pieces of information necessary for estimating probabilities.)
I assume (and I think RBH has confirmed) that in computer simulations all the historical data would be available for analysis.
IP: Logged
|
|
Janitor@MIT
Member
Member # 125
|
posted 22. April 2004 14:50
LOL I found this link http://www.everystudent.com/wires/ditching.html on the sourceforge.net. (For some reason I’m blocked from CalTech! Maybe they are having some problems.) What I want is the 1.0 manual.
Truman is not the only critic of avida or EC in general. Many evolutionary biologists are also. Not to mention, obviously, that many ECers have specific and substantive criticisms. Of both EC and biological theory! (I don’t know anything about Truman, so I don’t want to presume the usual “us vs. them” mentality. And I must admit I haven’t read his critique either.)
But I do notice, even right off the top, what is a common failure, repeated by RBH. Change can be neatly divided into adaptive, neutral, or deleterious. There is a consistent failure, and ultimately vain attempt to force discussion into the categories of traditional theory, when we make (either consciously or unconsciously) such a simple distinction. (It is a step in the right, adaptive, direction, even accepting such a ternary distinction. Only goes to show that “Darwinists” evolve by incremental steps.)
The very possibility that neutrality or degeneracy is an adaptation, and the basis of all other adaptations must be seriously considered and explored.
Is it a part of this program (avida)? And even if it as (as RBH indicates it is, maybe incidentally) will we understand it fully when we see it? Theoretical-spectacles?
I will also repeat my own very general caveat about programs: In this area the danger is that they become “self-fulfilling prophecies.” (Isn't that what programmers usually hope they are?) I can’t emphasize it enough, as biological research and theory has become increasingly computerized.
Darwin himself realized the problem(!). He adopted the term “natural selection” from the breeders, who used it to indicate their expectations of experimental results. Those results that were unexpected, did not conform to the conscious selections (expectations), or which they did not believe could be “consciously” selected, but were invariant or uncontrollable results, were denoted as the result of “natural” selection rather than “artificial” selection. (!Think about it!) I think we all know that what we consciously select can have unconsciously “selected” effects.
Gee, that sounds almost theological! Except that God, on the “principle of condescension” consciously chooses to be “unconscious.” Whereas we do not. Theology is so cool! (I am now discovering it. Anything new is cool.)
The only problem with theology is that there is no object for experimentation—other than ourselves! But it is an interesting “test case.” (And I have changed my opinion about this.)
All such “idealizations” in science are interesting test cases for how we perceive the world on the basis of (what kind of?) expectation, prediction, or imagination, as opposed to how it actually is. Gee, I will ask again: How is it that anything has evolved that is not perfectly consistent with the conditions of its existence? How did our minds evolve to imagine that things are different then they actually exist?
I think it’s very important to know, because I think it explains “design” and “evolution.”
Sorry for becoming a bullshit-philosopher. But if I could get my hands on the Adami et al manual maybe I could focus on the issues.
[I just wanted to add that I am not criticizing or even “making fun of” RBH. I know that RBH thinks seriously about the problems (Which many won’t even admit exist!) and I know he knows what he is talking about, and I rely on that. I am pointing to how “problems” are (theoretically) formulated informs expectations of “solutions” (and therefore the interpretation of results). Every question assumes not just an answer, but the correct answer, which “correct answer” depends on how the question is formulated. Aaaah! Damn!]
[ 22. April 2004, 15:07: Message edited by: Janitor@MIT ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 22. April 2004 16:47
J@MIT wrote quote:
But I do notice, even right off the top, what is a common failure, repeated by RBH. Change can be neatly divided into adaptive, neutral, or deleterious. There is a consistent failure, and ultimately vain attempt to force discussion into the categories of traditional theory, when we make (either consciously or unconsciously) such a simple distinction. (It is a step in the right, adaptive, direction, even accepting such a ternary distinction. Only goes to show that "Darwinists" evolve by incremental steps.)
If I add "as defined by the immediate selective environment" will that make J@MIT happier? ![[Smile]](smile.gif) (And, btw, that's what theory does: provide categories for phenomena. Absent theory, all we have is a bloomin' buzzin' confusion, a cosmic oddity shop filled with isolated and uninterpretable factoids.)
J@MIT further wrote quote:
The very possibility that neutrality or degeneracy is an adaptation, and the basis of all other adaptations must be seriously considered and explored.
Whether neutrality or degeneracy is the basis of all other adaptations is, I think, doubtful, though it is clearly the case that at (at least) two levels of analysis both are relevant to (potential, future) adaptation. Their very existence as contributors to population variability suggests that. Some wariness is necessary here, though: Focusing just on successful lineages is to consider a very biased sample. There are a heckuva lot of deleterious mutations that really are lethal.
J@MIT further wrote quote:
I am pointing to how "problems" are (theoretically) formulated informs expectations of "solutions" (and therefore the interpretation of results). Every question assumes not just an answer, but the correct answer, which "correct answer" depends on how the question is formulated. Aaaah! Damn!]
When I used to profess for a living I'd tell my students that the way a question is asked implicitly circumscribes a set of potential answers, and research on that question is directed at sorting through just that set of potential answers. Recasting the question redefines the set of potential answers and thus may push research in a different direction. Hence multiple rephrasings of questions results in sampling a larger set of potential answers, raising the likelihood that a (not necessarily "the") "correct" answer is even in the set. The only difference I have with J@MIT's statement in this respect is that there's no guarantee that a "correct" answer resides in the set of potential answers circumscribed by a given phrasing of a question. There are questions that are so ill-formed that they circumscribe a set composed of only "wrong" answers or even no answers - an empty set.
RBH
IP: Logged
|
|
Evan
Member
Member # 164
|
posted 22. April 2004 17:39
RBH writes,
quote:
There are questions that are so ill-formed that they circumscribe a set composed of only "wrong" answers or even no answers - an empty set.
This is an excellent point put in a way that I had never considered. Thanks.
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 22. April 2004 21:53
Technically, all Avida or any other computer program does is transform one precisely defined mathematical state into another precisely defined mathematical state. Until you define some explicit relationships among the components of the computer system, real world phenomena, and explicitly defined hypotheses, Avida can not provide analysis of either evolutionary processes or intelligent design processes.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 22. April 2004 22:32
warren_bergerson wrote quote:
Technically, all Avida or any other computer program does is transform one precisely defined mathematical state into another precisely defined mathematical state. Until you define some explicit relationships among the components of the computer system, real world phenomena, and explicitly defined hypotheses, Avida can not provide analysis of either evolutionary processes or intelligent design processes.
Yup. That's what the literature of modeling is full of. E.g, for an elementary discussion of dynamic models in this general area of inquiry see Holland's Emergence: From Chaos to Order, pp. 45ff.
Evan: Any time!
RBH
[ 22. April 2004, 22:37: Message edited by: RBH ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 23. April 2004 05:10
It is also useful to note that there is no known logical limit on ability of computer systems to transform or evolve from one precisely defined state to another. As far as is known, given any pair of a precisely defined starting point and an ending point, it is believed to be theoretically possible to define a computer program that will produce a transformation or evolutionary change from the defined starting state to the defined ending state. To my knowledge, no one is suggesting that IC or ‘can’t evolve’ are constraints or limitations applicable to mathematical systems.
IP: Logged
|
|
|
Printer-friendly view of this topic