Home » Intelligent Design » Sober and Irreducible Complexity

Sober and Irreducible Complexity

It has come to my attention that Sober claims to be using Behe’s definition of irreducible complexity.

Behe defines irreducible complexity as:

A single system which is composed of several interacting parts that contribute to the basic function, and where the removal of any one of the parts causes the system to effectively cease functioning”.

We can test this with structures like the flagellum through knockout experiments that remove bits of the structure so we can observe whether it continues to function. Thus the IC hypothesis makes predictions that can be tested.

Sober predicts an intelligent agency that created the irreducibly complex structures in nature must itself be irreducibly complex. Can someone tell me how to go about removing bits of the structure of this irreducibly complex intelligence to see if it is indeed irreducibly complex? If no one can then Sober’s hypothesis is untestable, unfalsifiable pseudo-science. It doesn’t pass muster even as philosophy as demonstrated by my previous analysis and it certainly doesn’t qualify as science when its hypothetical conclusion can’t be tested in any conceivable way.

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

4 Responses to Sober and Irreducible Complexity

  1. Not to be contrary, but is it possible that once Artificial Intelligence is developed (this is conjecture, but I’m guessing we will sooner than later) we can then theoretically “remove” bits of intelligence? Like removing certain routines and modules? What do you guys/gals think?

    (I’m a pro-ID advocate, btw, but I feel the need to play Devil’s advocate in this situation…)

    How would that reveal any information about the designer(s) of irreducible structures in living things? -ds

  2. Indeed, Berlinski makes the correct argument in Mere Creation: things which produce things which are inaccessible to Darwinian pathways are themselves inaccessible to Darwinian pathways.

  3. I’m just guessing at a possible path of falsification. If it can be shown that our AI intelligence can: 1) produce IC structures, and 2) this AI intelligence is NOT IC, then Sober’s premise would be falsified. (There would exist a non-IC intelligence capable of creating IC.) So then it wouldn’t be untestable….

    I don’t think Sober’s assessment is philosophically sound, but I think we may be able to falsify the specific claim concerning intelligence and IC, since there is a conceivable way we could test it.

    Of course we wouldn’t be testing to see if the Designer(s)’ mind was IC, but I don’t think we’d need to. We’d just need to show one exception to the rule to show that Sober’s assumption is false. Just my thoughts.

  4. Atom, I don’t think we can use AI to falsify Sober’s premise. We WOULD be testing to see if the Designer’s mind is IC. Forgive me it this is jut too obvious, but wouldn’t WE be the designers of AI? Even if AI becomes fully sentient and self-replicating, as the Singularity Watchers predict, it could still be traced back to us. And,even if we someday discover that we’re in an eternal feedback loop and that our own AI designed life on earth (I’m sure there’s a better verb tense for this in Greek)–we still must contend with the IC of our own minds. I think. :)

Leave a Reply