Intelligent Design

Dr. Ewert Answers

Spread the love

A few weeks ago I solicited questions on Google Moderator as an experiment. Unfortunately, the experiment was not very successful as I only got a few questions. However, I will answer the questions that did get listed. During this time Dr. Thomas English posted some questions at the skeptical zone, and I’ll be answering those here as well.

DieB asks: What happened to the erratum of “The Search for a Search”

Looking at internet archive, it looks like I inadvertantly reverted the evoinfo.org website back to an earlier version sometime around December 2012 losing the erratum. I have put the erratum back in place.

DieB asks: Take Ω={1..100}, fitness function “distance to a target”. Now you have a search which finds the target {1} with probability 1, μ=δ_1. You say that μ=δ_1 models your search. What does this model allow you to predict, perhaps for the target {2}?

The search in question is highly biased towards 1, and if this search were the product of a higher level search, I can predict with that that higher level search must have been indirectly biased towards 1.

It doesn’t make any prediction about the search if were you to replace the fitness function with the distance to another point. The model is only intended to make prediction about higher level searches, it is not intended to make predictions about structurally similar searches. Those kinds of predictions are outside the scope of the problems it seeks to solve.

DieB asks: S4S: taking the average of active information (p. 477) is de facto using the geometric instead of the arithmetic mean. The geometric mean favors equidistribution, but do you think that this is the appropriate mean in this situation?

I should note that I’m not an author on this paper, and anything I say is simply my own opinion.

In this case, what’s being calculated is the expected value of active information. Any other calculation would not have been the expected value of active information, and describing it as the average active information would in the very least have been confusing if not simply incorrect.

It should be noted, furthermore, that this paper is not alone in taking the average of a logarithmic quantity. For example, information theory uses entropy which is the expected value of information calculated in a manner similar to that done in this paper. It is not accidental that the paper describes the quantity as active entropy.

However, one must be careful with the mean or expected value because it is all too often misinterpreted. It is common to mistakenly assume that half of something will be below average, but that is a property of the median not the average. Similarly, it might be tempting to conclude from a negative average active information that more than half of all searches will be worse than the baseline, but this would be inaccurate. In my opinion, the paper would have been better had it considered the median active information instead of the average.

Andy asks: At ENV you said: “The universe must have begun with a large amount of active information with respect to the target of birds” How did you determine birds were a “target”? What of the billions of other living and extinct species?

I find this question a little bit strange, as I devoted a paragraph to answering it in the very post being quoted here.

One might ask, why birds? Birds are, in this discussion, thus far the target. By “target,” we do not mean something for which the search is actively looking. Recall that the only requirement of the search is that it be representable as a probability distribution. The target plays no role in what constitutes a search; rather, the target only features in the context of measuring the active information in a search. The target is effectively the measuring stick. The choice of target is arbitrary, and I could have as easily chosen cities, paintings, beetles, cows, volcanoes, mountains, lakes, or crystals. The same conclusion applies to all of them: they show up far more often than chance would lead us to expect. Within a materialist framework, they have to be explained by a process biased in favour of producing them.

For the purposes of conservation of information and active information, the target can be anything you like. There is nothing special about birds that make them a target. You can pick anything you want to be a target.

Dr. English asks: What is the formal relationship between active information and specified complexity?

We have never developed any kind of formal relationship between active information and specified complexity. Both are directed at answering different though related questions. To illustrate, consider a poker player who manages to get ten royal flushes in a row. The other poker players naturally protest that he is cheating.

One defense he might offer is that while the probability of his hands was very low, so was the probability of every other hand. In fact, a royal flush is no less likely than any other combination of five cards. Specified complexity is intended to answer this kind of objection. It provides a justification for why we deem obtaining the royal flush an unusual event while an equally improbable combination of random cards would not be unusual.

An alternative defense would be to acknowledge that obtaining a series of royal flushes is unusual, but appeal to the process of shuffling. Somehow the shuffling process produced the royal flush. Active information is directed at this defense. It may be that the initial configuration and shuffling process somehow gave the player royal flushes with high probability. However, that doesn’t really explain the flushes, it merely moves it to the problem of having to explain why the initial configuration of the deck and the shuffling process gave him royal flushes.

Dr. English asks: What is the formal relationship between active information and average active information per query? Does the conservation-of-information theorem apply to the latter?

English goes onto to claim that “Winston and his colleagues have measured only average active information per query”. This is incorrect. In “Climbing the Steiner Tree” we present values of active information without any use of any form of active information per query. “Conservation of Information in Search: Measuring the Cost of Success” calculates the active information for a number of examples. It is simply false to claim that we measure only active information per query.

Active information per query is defined to be the active information divided by the number of queries used to extract that information. We have commonly used the quantity of average active information per mean query which is the active information divided by the expected number of queries required to extract the information. English claims that the active information per query “seems unrelated to active information” but since the active information per query is simply the active information divided, its rather odd to claim they are unrelated.

Active information per query is a rate of information, much like watts are a rate of energy. In either case that conservation results relate to the total energy or information of a system, not the rate. But the conservation of information or energy still means that a system is constrained by the rate information or energy is put into the system. If you only 80 watts going into a system, you can’t constantly power a 100 watt light bulb. The watt-deficit is going to catch up with you sooner or later. Similarly, in order to obtain a system with a large amount of active information with a small amount of queries, the active information per query must be high.

Dr. English asks: Your “search” process decides when to stop and produce an outcome in the “search space.” A model may do this, but biological evolution does not. How do you measure active information on the biological process itself? Do you not reify a model?

One approach is to take the search space not to be the individual organisms, but rather the entire population of organisms currently alive on earth. Or one could go further, and take it to be the history of organisms during the whole of biological evolution. One could also take it to be possible spacetime histories. The target can then be taken to be spacetimes, histories, or populations that contain an individual organism type such as birds. Something related is developed in example 3.5. This avoids having to define a stopping condition that isn’t present in nature.

Another possibility is to model evolution as a process which halts upon finding the target, but distinguish between the active information derived from the evolutionary process itself and the active information contributed by the stopping behavior. The stopping behavior cannot induce birds to show up in the first place, it can only select them as the output of the search when they arrive. By looking at the number of opportunities for birds to arise, we can determine how much active information was added by the stopping process. It was shown in “Conservation of Information in Search: Measuring the Cost of Success” that the active information available from such a process is only the logarithm of the number of queries. Any other active information must derive from the evolutionary search itself.

Both approaches effectively end up adjusting for the number of trials. Getting a royal flush is improbable, but if you play five million hands of poker it is no longer surprising. Similarly, obtaining a bird is rendered much more probable given the number of chances for it happen in the history of universe. It is a very important point to keep in mind that we cannot simply look at the probability of the individual events but also the number of trials. However, even the vast resources of the universe is insufficient to render birds probable by chance alone. For birds to have been produced by an evolutionary process, the universe must have been biased towards producing birds.

24 Replies to “Dr. Ewert Answers

  1. 1

    I suggest you cross post this as a cross post at TSZ, as it was there that the Tom English’s question was posted.

    Not everyone who would like to discuss this with you is able to post here, including, I think DiEB, and several of us are reluctant to do so, given the deletion of the entirety of Aurelio Smith’s posts (apart from traces left in those you responded to).

    UDEditors: Elizabeth, if you don’t want to post your incoherent blitherings on these pages, then by all means do not. But this silliness of putting up post after post about how you are afraid to post here must stop. Only warning.

  2. 2
    Orloog says:

    Just checked https://www.google.com/moderator/#15/e=21afd2&t=21afd2.40

    You seem to have missed one question:

    “Still going? Great! Quick question: ?={H,T}, ?=?_H natural measure. In which sense is \overline{?} “natural” on M(?)?”

  3. 3
    larron says:

    Dr. Ewert, thank you for taking the time to answer the questions!

    Oorlog, the last question was a latecomer. It refers to The Natural Probability on M(?) at DiEbLog. Perhaps Dr. Ewert will address it, too….

  4. 4
    larron says:

    It doesn’t make any prediction about the search if were you to replace the fitness function with the distance to another point. The model is only intended to make prediction about higher level searches, it is not intended to make predictions about structurally similar searches. Those kinds of predictions are outside the scope of the problems it seeks to solve.

    That is very strange: an algorithm which works for the Traveling Salesman Problem for the 10 biggest cities in the U.S. is “outside of the scope of the problems” when it is used to solve the TSP for the 10 biggest Canadian cities?

  5. 5
    Orloog says:

    Could someone unblock DiEb?

  6. 6
    larron says:

    Winston Ewert, in the paper A General Theory of Information Cost Incurred by Successful Search you describe searches using initiator, terminator, inspector, navigator, nominator, and discriminator. Isn’t a search described by this entities which uses the given fitness function “distance to the target” to find the target effectively – whether it is {1} or {2} – structurally the same for both targets and not only similar?

  7. 7
    Mung says:

    Thanks Dr. Ewert. I particularly appreciated the section on active information and specified complexity.

  8. 8
    SimonLeberge says:

    Could someone unblock DiEb?

    And legitimize ex post facto Dr. Ewert’s response to him in a forum from which he’s excluded? A similar remark applies to the expelled Dr. English.

    To get an idea of why meaningful discussion seems unlikely, check out the full sentence that begins with Dr. Ewert’s quotation, “Winston and his colleagues have measured only average active information per query”. People with doctoral degrees generally know to use ellipsis (…) to indicate that they’re omitting text.

  9. 9
    Mung says:

    Perhaps Elizabeth Liddle is yet another sock puppet of Alan Fox and so he has actual reason to be worried about having more of his posts deleted.

    Trying to understand the argument:

    You deleted all the posts of a sock-puppet, so now all legitimate posters should worry too. But Why?

  10. 10
  11. 11
    Dionisio says:

    Dr. Ewert,

    Are the questions highlighted in the below linked post somehow related to this thread or off topic? Thank you for any comments on this.

    http://www.uncommondescent.com.....ent-566910

  12. 12
    Zachriel says:

    Orloog: Could someone unblock DiEb?

    Interesting blog. Thanks for the link.

  13. 13
    DiEb says:

    1) I’ve asked Barry Arrington to unblock my account – if you read this, it has worked! Thanks everybody!

    2) There isn’t much interest in the technical aspects of EIL’s articles on neither side of the aisle, while the philosophical and theological implications are discussed with great furor. The German term for this is Luftschloss.
    4) That said, I was surprised that no one of the pro-ID side took advantage of this opportunity to ask questions! Granted, I can live without Joe discussing infinity….

    3) I hope that Winston sticks around to address some follow-up questions!

  14. 14
    Mung says:

    DiEb,

    Great! So “Aurelio Smith” was a sock-puppet of Alan Fox.

    Does the fact that the posts of “Aurelio Smith” were deleted give you reason to believe that your posts will be deleted?

  15. 15
    Orloog says:

    Mung, what a relief! “Aurelio Smith” got his level two banning (block + deletion of comments) for being a sockpuppet, and not because of abuse and cyberstalking or worse, as KairosFocus feared:

    Where, too, I am confident that where level two banning was applied at UD, it would be for serious cause comparable to why a gentleman would request that a guest immediately get up from the living room sofa, collect his belongings then exit the house and not return to the premises — so, it is instructive that, even after several times where I have had to point out such material context you seem unwilling to acknowledge the seriousness of the persistent problem of abuse and cyberstalking or worse; where you know or full well should know you are dealing with a target of that stalking.

  16. 16
    kairosfocus says:

    F/N: Pardon a tangential remark. I note briefly for record; it seems that there remains an unwillingness to face the problem of cyberstalking and abusive commentary. Where, sockpuppetry is of course notoriously a typical part of the problem. It is high time that abusive and enabling behaviour are cleaned up, and no, we are not forced to put up with abusive behaviour, stalking and the like as the price extracted for attempted discussion. G’day. KF

  17. 17
    SimonLeberge says:

    Mung:

    DiEb,

    Great! So “Aurelio Smith” was a sock-puppet of Alan Fox.

    Does the fact that the posts of “Aurelio Smith” were deleted give you reason to believe that your posts will be deleted?

    So is DiEb the sock-puppet of larron, or is larron the sock-puppet of DiEb? He’s gone out of his way to make the correspondence clear.

    Then again, larron might comment, “I’m not DiEb,” with DiEb adding, “I’m not larron.” That would make a fool of me, wouldn’t it? Then again, again, what if I tell you that I am DiEb’s sock-puppet’s sock-puppet?

  18. 18
    DiEb says:

    DieB asks: What happened to the erratum of “The Search for a Search”

    Looking at internet archive, it looks like I inadvertantly reverted the evoinfo.org website back to an earlier version sometime around December 2012 losing the erratum. I have put the erratum back in place.

    That’s nice to hear, thank you, no conspiracy after all 🙂

  19. 19
    DiEb says:

    @Mung: Having your all your posts deleted isn’t any longer an unprecedented event. I have the habit to save my edits, so, for me, not much will be lost. But I’m more concerned about getting blocked without any official and correct reason, having KF speculate about my crimes…

    @SimonLeberge: *LOL*

  20. 20
    DiEb says:

    DieB asks: S4S: taking the average of active information (p. 477) is de facto using the geometric instead of the arithmetic mean. The geometric mean favors equidistribution, but do you think that this is the appropriate mean in this situation?

    I should note that I’m not an author on this paper, and anything I say is simply my own opinion.

    Well, it’s an informed opinion, as you are one of the people most involved with the authors.

    In this case, what’s being calculated is the expected value of active information. Any other calculation would not have been the expected value of active information, and describing it as the average active information would in the very least have been confusing if not simply incorrect.

    Please take a look at this: Five Years of “The Search for a Search” There, I give an example where this “average active information” leads to an incorrect and confusing result.

  21. 21
    DiEb says:

    DieB asks: Take ?={1..100}, fitness function “distance to a target”. Now you have a search which finds the target {1} with probability 1, ?=?_1. You say that ?=?_1 models your search. What does this model allow you to predict, perhaps for the target {2}?

    The search in question is highly biased towards 1, and if this search were the product of a higher level search, I can predict with that that higher level search must have been indirectly biased towards 1.

    It doesn’t make any prediction about the search if were you to replace the fitness function with the distance to another point. The model is only intended to make prediction about higher level searches, it is not intended to make predictions about structurally similar searches. Those kinds of predictions are outside the scope of the problems it seeks to solve.

    Addressing this answer will take some time, I’ll address this later. Will Dr. Winston Ewert be around?

  22. 22
    DiEb says:

    And then, there is the question:

    “Still going? Great! Quick question: ?={H,T}, ?=?_H natural measure. In which sense is \overline{?} “natural” on M(?)?”

    I’d appreciate an answer to this question: your construction is elegant, but why should it be natural?

  23. 23
    SimonLeberge says:

    DiEb:

    “Still going? Great! Quick question: ?={H,T}, ?=?_H natural measure. In which sense is \overline{?} “natural” on M(?)?”

    Evidently you’ll have to carry on the discussion with raw LaTeX expressions, e.g.,

    I_\boxplus = \frac{I_\Omega}{\overline{Q}}.

    Of course, if you want to write \mu(T) < 1, you have to enter “&lt;” to get the “<” symbol. Funny thing about this stuff is that it’s a lot easier to write than to read, even for readers who know how to write it.

  24. 24
    DiEb says:

    roma locuta causa finita? Winston, I’m sorry that this experiment was not as successful as you hoped, but I’m afraid it will be even less a success if you don’t intend to follow up….

    OTOH: if you think that you are finished with the questions, then it is really of no importance which venue you have chosen, UD or TSZ.

Leave a Reply