Home » Intelligent Design » Dinnertime Design Detection

Dinnertime Design Detection

Last evening I was talking to a friend about how my dad had to learn morse code when he was in the navy, and he related a funny design detection story (not that he put it in those terms).

My friend had a cousin (we’ll call him Bill), and when he was a teenager Bill developed a nervous tapping habit, or so everyone thought.  One evening Bill’s family had an older couple over for dinner, and Bill was tapping away when both guests got red in the face and exclaimed “Bill!  What are you doing?”  It turns out Bill had been learning morse code and tapping on the table for practice.  The problem:  He was practicing with four letter words, and no one knew until the family invited two retired Western Union operators to dinner!

  • Delicious
  • Facebook
  • Reddit
  • StumbleUpon
  • Twitter
  • RSS Feed

8 Responses to Dinnertime Design Detection

  1. This post was designed to be amusing. I loved it! Great story.

  2. Actually, the story suggests that the design was not detectable by someone who did not know a) the language of the source and b) the coding system.

    Something interesting about English transmissions in Morse code is that they are statistically anomalous. The length of a codeword for a letter is related to the frequency of the letter in English text. The short codewords occur much more frequently in transmissions than the long codewords. Someone who knows neither English nor the code can detect this element of design in the sequence of “dits” and “dahs.”

  3. Oatmeal Stout,

    Conversely, the story suggests that the design was detectable by someone who knew a) the language of the source and b) the coding system.”

    What’s your point? What am I missing? I can’t seem to detect what it is your trying to say! Hurry! Let me know before I am destroyed!!

    Oops, too late. Maybe you can explain it to the next organism that arises. But, I doubt it.

  4. Oatmeal Stout,

    The content was not detectable by someone who did not know the code.

    That an agency was tapping was readily observable.

  5. Joseph @ 3:

    That an agency was tapping was readily observable.

    Yes. Not inferred, or detected.

  6. Diffaxial, Not detected? Really?

  7. OS:

    Design detection — you have proclaimed yourself an ID supporter, but show a pattern of supporting criticisms and misunderstanding of basic ID concepts.

    Guess what that supports, as a design inference [in the days when ACORN-type, Axelrod "astroturf" tactics are now commonplace]?

    GEM of TKI

    PS: the design inference does not claim — nor does it need to claim — a universal decoder capable of detecting any and all instances of functionally specific information. It does not even need to claim that it never makes negative errors: misidentifying as non-designed, cases that are. What it is focussed on, is that when the explanatory filters detect design, they do so reliably where we can check [and there is no credible mechanism tracing to chance and blind mechanisms of necessity], so that we have a good reason to trust them when we cannot make that check directly.

  8. What it is focussed on, is that when the explanatory filters detect design, they do so reliably where we can check [and there is no credible mechanism tracing to chance and blind mechanisms of necessity], so that we have a good reason to trust them when we cannot make that check directly.

    But we can only “check directly” when the design is a human process, which may be categorically unlike the inhuman process of supernatural design. Assuming that one category of design is like the other is assuming your conclusion.

    Nor, to my knowledge, do IDists “check” their methods of design detection, even against human designs. Has there been any study of the reliability of the methods of ID? Any double-blind tests, or research into error rates?

Leave a Reply