Glen says: "I have to admit, this seems like a really difficult multi-objective selection method. "
Yes, yes, yes! But, to stick with the analogy, it is not in-principle more difficult than distinguishing chemical compounds. Admittedly, Chemistry had quite a head start as a formal science. However, if psychologists had their heads out of the rears, and had put in as much effort over the last 100 years into classifying the ways people interact with the world as chemists had put into classifying the ways chemicals interact with the world, the question wouldn't seem so intimidating. We would have achieved, or be close to, whatever psychology's version of the periodic table is (which I know is itself continuously up for re-conceptualization, but the basic one is still incredibly helpful). As for your more specific question, it is pretty easy to tell believers from fakers... so long as we exclude faker-recursion. That is, it is possible for a human to be a believer faking being a non-believer, etc. If we stick to the original two-option case, it is pretty easy - I submit - because we do it all the time. Specifying exactly how we do it is tricky only because the research hasn't been done. Check out any Daily Show coverage of the presidential debates. One of the best bits so far is the Fox News commentator who, after Romney's speech goes on for quite a while about how great it is that there were so many details, how this will really connect with voters and answer their questions, etc. Then, immediately after Obama's speech he goes off about how the speech included a lot of details, and that is sure to alienate voters. If we only saw the first speech, we might think that the commentator believes details are good, or at least that he believes viewers want details. After seeing the clips next to each other, it is clear that he was merely faking that belief as part of a larger pattern serving some other purpose. What are the varieties of ways in which we make these distinctions? It is a tremendously complicated, but ultimately tractable question. Eric P.S. This problem is of particular interest to one of the topologists on the list - Lee Rudolph - who just had a book on the subject release. I haven't read it yet, but I know it is (among other things) an attempt to apply modern, non-statistical, mathematics to this problem. That would include math that can adequately deal with discrete and non-discrete aspects, etc., which you point out we would need. Lee, can you give a more skilled plug? On Wed, Sep 19, 2012 06:29 PM, glen <[email protected]> wrote: > ERIC P. CHARLES wrote at 09/19/2012 02:54 PM: >> But Glen, when you talk about the infiltrator, or the person >> "paying lip-service", you are just appealing to a larger pattern >of >> behavior. > >Aha!! Excellent! So, tell me how to classify the patterns so that one >pattern is just lip service and the other is belief! If you do that, >then we'll have our objective function. I can develop an algorithm for >that and we'll be able to automatically distinguish zombies from actors. > Then we can begin building machines that try to satisfy it. > >> Agreeing with your assertion, "faking belief" looks different >> than "belief"... if you can see enough of the person's behavior >and/or see a >> close enough level of detail. > >The former, again, sounds like memory. The latter is something else. >It implies something about scale. We know actions are multi-scale >(anatomy, physiology, chemistry, physics). Is there a cut-off below >which we need not go? Genes? Chemicals? Or does the multiscalar >requirements for measuring belief extend all the way down? > >> a person who believes X >> and a person faking belief in X are distinguished by observing a wide >variety >> of ways in which the people interact with the world. > >So, in addition to memory and crossing scales, the measures are also >multivalent at any one instant or any one scale. > >> Also, for the record, one of the problems with using "moles" is >that it is very >> difficult to get people capable of participating in cultural practices of >these >> sorts over extended periods without becoming believers. The practices >become >> normal to you, the group becomes "your group", and even if you >can still turn >> them in/report on them/whatever you are supposed to do, you become >sympathetic. > >Uh-oh. This makes it sound like not only is there a multi-scale >problem, but there may also be a hybrid requirement. The mole either >continuously transitions from non-belief to belief or there's a >threshold. I.e. some parts of our classifying predicate will be >continuous and some will be discrete. > >I have to admit, this seems like a really difficult multi-objective >selection method. Building a machine that generates belief from a >collection of mechanisms, thereby satisfying the criteria, will be >exceedingly difficult, at least as difficult as artificial life and >intelligence. But this is what we have to do if we're going to continue >claiming that beliefs reduce to actions. > >-- >glen > >============================================================ >FRIAM Applied Complexity Group listserv >Meets Fridays 9a-11:30 at cafe at St. John's College >lectures, archives, unsubscribe, maps at http://www.friam.org > > > ------------ Eric Charles Assistant Professor of Psychology Penn State University Altoona, PA 16601
============================================================ FRIAM Applied Complexity Group listserv Meets Fridays 9a-11:30 at cafe at St. John's College lectures, archives, unsubscribe, maps at http://www.friam.org
