False false dichotomies

Jed Harris wrote, in one of his otherwise very insightful comments:

(Incidentally, there probably is no viable distinction between cognitive structure and content.)

This statement is true, in that there is probably no distinction that I can write, that Jed can’t come up with a counter-example to.  Much as I can’t write a definition of "game" that Wittgenstein couldn’t come up with a counter-example to.

But the statement was used to imply that distinguishing AI architectures by reliance on content vs. learning is nonsensical.  If that were so, knowledgeable people would be confused when Eliezer (or Lenat) says Cyc emphasizes content more than other architectures do.  They aren’t.

Some more-popular false false dichotomies:

  • Nature vs. nurture (e.g., genetic or instinctual vs. learned behavior):  We’re told that there’s no true distinction between them, since "nurture" can only occur when expected by "nature".  I like Paul Bloom’s reply (paraphrasing), "There’s something wrong with a theory of mind that says that a knee reflex and word learning are the same sort of thing."
  • Race:  We’re told that race is a "social construct" because, for any particular genetic criteria you set to determine who is in a race, someone can be found who looks to us like they belong to that race, yet doesn’t satisfy your criteria.
  • Gender:  There are people naturally having characteristics of both sexes; people whose phenotypic gender is different from their genotypic gender; and people who’ve had sex-change operations.  Therefore, there is no gender.

You probably knew where I was going with this when you saw the Wittgenstein reference.  Every word in our languages breaks down when you apply enough pressure to it.  A word encodes a statistical regularity.  Applicability in all cases is not required.  Forbid us from using words that aren’t precise, and we’d be unable to talk at all.

GD Star Rating
loading...
Tagged as:
Trackback URL:
  • http://hanson.gmu.edu Robin Hanson

    Yes dichotomies need not be exactly definable to be useful. But we don’t want to accept just any old sloppy dichotomy. So the question is: how do we tell when they are too sloppy?

  • http://shagbark.livejournal.com Phil Goetz

    It depends on the context. A dichotomy might be valid for one purpose, but invalid for another purpose. We shouldn’t accept or reject them once and for all.

  • frelkins

    @Phil

    on the context

    Am I following you? You are thinking of the T L-P from section 4? You are specifically pointing to section 4.063, “An analogy to illustrate the concept of truth?”

  • http://hanson.gmu.edu Robin Hanson

    Technically, I think you want to say CYC emphasizes content over architecture more than other projects do.

  • http://jed.jive.com/ Jed Harris

    Thanks to Phil for the kind appraisal of my comments. I’m not sure he and I disagree. I certainly agree with his general point that no distinctions stand up to unlimited pressure, and I very much appreciate his mention that “A word encodes a statistical regularity”. Bravo! This is under-appreciated.

    To further the discussion I should clarify some issues from my perspective.

    I didn’t mean that the distinction is nonsensical and I hope none of my other points depended on such a claim; if so I’d need to fix them.

    The example of knee reflex and word learning is useful in a couple of ways. First, of course I agree there are more genetically determined aspects of bodily reactions, etc. So the idea of a spectrum rather than everything together in a lump is good.

    But consider word learning, especially in the context of Eliezer’s points about exchanging high level thought patterns between differently structured AIs. Obviously the kinds of words humans can learn are very constrained by their cognitive, auditory and articulatory architecture. Conversely, that architecture evolved to a considerable extent just so they could hear, understand, learn, generate, and use words. So in a more various world we’d have to say “human words” since there’d be lots of kinds of words humans couldn’t learn.

    (There’s a lot more here I’d like to say about the larger social context influencing the words we can understand, but now is not the time.)

    So what I was trying to do with that (rather awkward and too often repeated) expression was to acknowledge that content and structure co-determine each other. In that light, “no viable distinction” is the wrong way to put it. Maybe I should have said:

    Incidentally, there’s probably no way to specify cognitive structure and content independently, since they co-determine each other.

    Thanks very much to Phil for motivating improvement here.

  • http://shagbark.livejournal.com Phil Goetz

    Am I following you? You are thinking of the T L-P from section 4?

    Nope. I just meant that a dichotomy like “northern vs. southern attitudes” might be a good one when talking about the US Civil War or pancake restaurants; less good when talking about gun control; and of little use when talking about Newcomb’s Paradox. So, no one should say “nature vs. nurture is a false distinction”; they should say that it’s a poor distinction for the topic under discussion.

  • Peanut Gallery

    See http://en.wikipedia.org/wiki/Family_resemblance for some passages in Philosophical Investigations where Wittgenstein talks about games and family resemblance.

  • http://shagbark.livejournal.com Phil Goetz

    Oh – Thanks, Peanut. I misunderstood freklins’ question.

  • Josh Witten

    “Race: We’re told that race is a “social construct” because, for any particular genetic criteria you set to determine who is in a race, someone can be found who looks to us like they belong to that race, yet doesn’t satisfy your criteria.”

    This is misstating the genetic problem. “Races” (more appropriately “geographic ancestry,” as this is what the genetic data predict) are defined genetically in probabilistic terms based on allele frequencies. As a result, individuals create semicontinuous distributions between clusters of genetic similarity. Furthermore, traditional concepts of race imperfectly correlate with the genetic data. Traditional racial groups are, in part, a “social construct.” The concept of racial groups, in general, is flawed because the groups cannot be precisely defined, not because of individual exceptions.

    Sex, on the other hand (gender is now being used in the literature to connote an individual’s perception of their male/female identity), can be precisely defined as male (1 X & 1 Y chromosome) and as female (2 X chromosomes), even though those two groups do not include all individuals.

  • http://shagbark.livejournal.com Phil Goetz

    “Furthermore, traditional concepts of race imperfectly correlate with the genetic data.”

    Imperfect correlation is OK. That’s what I’m saying.

    “Traditional racial groups are, in part, a “social construct.””

    In some ways, such as “black” meaning “at least one black ancestor” in the US, whereas it doesn’t mean that in the Caribbean. But race is not a meaningless term. Lots of medical conditions are highly correlated with race in the folk-taxonomy phenotype-based concept of race.

    “The concept of racial groups, in general, is flawed because the groups cannot be precisely defined, not because of individual exceptions.”

    And the point of my post is to say that the fact that groups can’t be precisely defined doesn’t mean the concept of racial groups is a bad one.

  • Tyrrell McAllister

    Two points on the race issue.

    First, calling race a “social concept” is not to deny its reality. It’s not to deny that there is a useful, even if very fuzzy, dichotomy between people of race X and people not of race X. That distinction is real, and it really matters.

    Second, it nonetheless seems that race is not a valid concept within the science of genetics. This just means that when geneticists carve up humanity into populations in the way that is useful for them, their partitioning does not coincide with the partitioning induced by the social concept of race. Their partitioning can’t even be usefully viewed as a refinement of the racial one. That is why race shouldn’t be thought of as a scientific concept.

    There is some statistical correlation between the social concept of race and the scientific concept of a genetic population, but the correlation is too weak in practice. There’s probably also a statistical correlation between genome and favorite food, but that doesn’t make favorite-food surveys an acceptable substitute for genetic tests.

    That, at least, is what I gather is the mainstream view among population geneticists. But I also gather that there isn’t a unanimous consensus.

  • TGGP

    Ernst Mayr, who came up with the commonly used definition of “species” in biology, says race is just as valid a concept.
    http://www.gnxp.com/MT2/archives/001951.html
    “Population” is just another word for “race” in genetics. A euphemism, if you will. When discussing non-human species, scientists have no problem using the term “race”. The comments thread to this post discuss the two terms and their usefulness.
    http://www.gnxp.com/blog/2008/11/kenan-malik-and-kerry-howely-on-race.php

  • Josh Witten

    “Imperfect correlation is OK.”

    The imperfect correlation of race with geographic ancestry (what the genetic variation actually determines) hurts the ability to use racial groups to predict phenotype.

    “Lots of medical conditions are highly correlated with race in the folk-taxonomy phenotype-based concept of race.”

    True for rare recessive disorders, but frequently over interpreted. These correlations are almost exclusively the result of founder effects/bottlenecking/genetic drift, not selection. Quantitative traits, which are the majority of disease phenotypes, do not correlate well. Preventative medical interventions taken based on race for these conditions tend to be very low risk and are prescribed inefficiently (many treated people needed to see one beneficial outcome), such as diet, exercise, and statins. Race provides little predictive power outside of the extremes.

    The difference between race and sex is that exceptions from the male/female dichotomy are rare, genetic abnormalities. While there are clusters of geographic ancestry, the exceptions in the continuum between these clusters are normal and the genetic variation is consistent with random, genetic drift.

    “Ernst Mayr, who came up with the commonly used definition of “species” in biology, says race is just as valid a concept.”

    Absent data, Mayr’s opinion is irrelevant regardless of previous contributions or his acknowledged brilliance. Furthermore, he did not have access to the expansive sequence data we have today. I would not presume to guess how these data would affect his opinion.

    “”Population” is just another word for “race” in genetics. A euphemism, if you will.”

    I won’t. Race is a sloppy term that some equate to population. It does not go in the other direction. Population has explicit meaning. Race is one potential way to define a population. Race is a sub-optimal tool for defining discrete populations and adding efficiency to predictions.

    “When discussing non-human species, scientists have no problem using the term “race”.”

    Personally, I do not know scientists (which puts both of our conclusions somewhere between all scientists and none) who discuss “races” of non-human species. You will hear the term “ecotypes,” which implies adaptation to local environmental conditions. Just because scientists use the term does not make it valid.