Sunday, April 1, 2018

Fuzzy Boundaries

Fuzzy Selection Boundaries
One critique of multi-level selection theory is the biological reality (or unreality) of groups*.  Individuals are not fully entwined with the group the way individual cells are entwined with a parent organism. Lack of dependency and coordination challenges the reality of a full biological transition to the group level. The critique is that acting as if a fuzzy transition is "real enough" is just not appropriate. While humans outside of any parent group have significantly decreased fitness, they can still survive. Many evolutionists argue selectionist tools can't be legitimately applied to fuzzily cohered groups.

Boundaries are insufficiently clear, stark, or enduring.

Fuzzy Cultural-Ancestry (i.e. "race") Boundaries
Recently, anthropologists have critiqued geneticist David Reich's recent book arguing for the statistical and hence physical reality of statistical cultural-ancestry (i.e. "race"). One of the major complaints is that "race" boundaries are too fuzzy. Variation between populations is near meaningless compared to the variation within the whole population or even within statistically aggregated sub-populations.

This is another example where the complaint is that of insufficiently clear, stark, and enduring boundaries.

Fuzzy Religion Boundaries
One complaint with my (or anyone else's) necessary but not sufficient factor approach (i.e. 6 out of these 9 factors) for defining religion is that it produces boundaries which are too fuzzy. It also risks including groups which may pass a factor test but common sense would strongly suggest might not be religious (e.g. Amway, most zealous but non ultra fanatical comic fans, etc.)

How Many Socio-Biological Boundaries are Fuzzy?
The counter point to the insufficiently clear boundary critique is that some biological boundaries in the field of evolution are just not as solid as folk intuition might otherwise suggest.

For example, Massimo Pigluicci's seminal paper on An Extended Synthesis for Evolutionary Biology lists numerous levels of selection and multiple paths for genetic inheritance (what affects gene fitness).

 It is now clear that several levels of the biological hierarchy are, at least theoretically, legitimate targets of selection, from genes to individuals, from groups of kin to populations to species (Okasha, 2006)

The current emphasis in genomics research is no longer on the effects of individual genes or of single metabolic cas- cades but rather on the system-level properties of entire networks of gene products (Cork & Purugganan 2004; Wray 2007),  
Along similar, yet independent, lines it has been proposed that there are four, not just one, systems of inheritance affecting the evolu- tion of biological organisms (Jablonka & Lamb 2005): the standard genetic one, an epigenetic component (based on the inheritance of epi- genetic markers, e.g., methylation patterns), a behavioral one (e.g., imitation in some species of animals), and a symbolic one (limited, as far as we know, to humans). This has prompted a rethinking of previously basic concepts in evo- lutionary theory, beginning with the idea of a “replicator” (Szathma ́ry 2006), as well as re- newed efforts at empirical research exploring the extent and importance of heritable epige- netic effects (Chandler & Alleman 2008).

Gene-Culture Boundaries?
The incorporation of culture into evolutionary thinking muddies biological boundaries. This is why critiques of multi-level selection theory and other dual or multi-path inheritance models should not be discarded out of hand. The value gained by dealing with fuzzy edges doesn't necessarily compensate the lack of rigour it costs. Faulty logic may indeed create insidious stumbling blocks which are hard to remove.

On the other hand, reality seems to be all about fuzzy boundaries.

Part of the issue comes from the way people understand (or misunderstand) statistics.  For instance, is anything other than a 1-to-1 match for genes to "racial" group acceptable? I doubt it.  There are issues going from population level descriptions to individual level assumptions. Geneticists know not to do this. The public and many non-mathematically educated, humanity oriented, types may not. Further than this, many activists purposefully misconstrue this boundary so as to fire up righteous resistance. After all, if objective facts can be used nefariously.  In some people's minds, better to create taboos to ensure this can never happen.

In many ways this is reminiscent of academic purity arguments around fuzzy boundary contagion....

This is not to say that "objective scientism" on topics such as "race" / ancestors can't be offensive. It certainly can be extremely offensive. My own Sami / Lap heritage was definitely subject to this (albeit with the addition of "white dis-privilege").

Definite things can still be probabilistic. Anyone who has taken a quantum course knows this. Reconstructing probabilistically determined entities is never going to be a 1-to-1 enterprise. Perhaps the statistical socio-physics literature has produced some work here. I'm not sure, and to me, engaging Luddites on this front tends to boorishness.  The two sides have different utility functions. Both are justifiable. They just represent different Haidt teloses.  Neither should be universally mandated.

While one may be more "real" than the other, they need to exist in tension. Without this, you tend to generate some rather nasty secondary effects (hyper-rationalism which increases the chance of eugenic like perspectives emerging or taboo based anti-scientism / Ludditism / socially approved "facts").  This is partly why I support learning how to live with and work with "religious" ways of knowing. Societal coherence (& survival) requires it.

Here is one physical analogy of the situation. You have an individual item you want to test (cultural/ethnic ancestry, religiousness, group level). It produces some of the same signals as the category you are to test. Let's say you notice a -1 charge. Is it an electron? What about a Tau lepton? (Both have this charge)

You test some properties. It has a half integer spin. What is it guaranteed to be?  (Both have this spin)

At this point you can't conclude. Let's assume other measures are unavailable. You might guess which one it is based upon probabilities. Electrons are more plentiful than Tau's. Has the particle's reality changed?

I think arguments boil down to a couple of scenarios:

  1. Treating the "particle" as set of probabilistically composed options (i.e. a X% chance of an e- and (1-X)% chance of a T-.
  2. Treating them an a single indistinguishable entity that has an inescapable chance of error. (i.e. its always an electron, we're just wrong X% of the time)
  3. Treating them an an indistinguishable entity that, if used, produces an inescapable chance catastrophe.
  4. Treating the system as a single entity (which has the average properties of its entities, say a mass which is the average of the e- & T-)

My own naivete tends to see the anthropological critique of "race" / statistical ancestry as an example of case 4 thinking. I suspect genetic evolutionists see things as a case 1 scenario.

I suspect another major issue with fuzzy boundaries in social science work is the issue of scale.

Scale is rather arbitrary. For example, what is the ideal scale for "race" / statistical ancestry? While it is easy to assume anthropologists are naive of quantitative methods for phylogeny, I doubt that is really the case. I suspect they mainly reject it because of error concerns, utility calculations, and the secondary effects this might have in how they and their systems value people.

And so, while regression can optimize the scale of sub-population groupings, it is always going to be somewhat arbitrary. This is because the scale of distinct lineages varies across the world. People who were genetically isolated (to varying degrees) did not produce groupings of the same size. This makes scales rather arbitrary. Are you looking for distinguishable phenotypes or distinguishable markers? Migration genetics seems to look for anything that produces distinguishable populations so as to track movement. That does not necessarily correspond to distinguishable phenotypical differences. It, of course, may, but this is of course probabilistic.

The heritability and distinctiveness of culture is, I suspect, the nail that seals the fate of anthropological interpretation. If cultural difference is not significant and is so malleable so quickly, "race" can become as trite as assuming the colour of clothes you wear one day to another says anything about who you are. Of course, many people, like Amazon, Facebook, etc. would suggest the clothes you wear on any day enable them to probabilistically say a lot about you. But if you went from a goth to a cowboy to a prep, have you really changed?  Process philosophy starts rearing its complicating head...

However, if cultural difference is significant, and long-lasting enough, then the anthropological interpretation is sealed, albeit in a different direction.  I think gene-culture work, such as the classic lactose problem illustrates the value of this interpretational frame. Similarly, I think the evolutionary synthesis represented by Pigluicci's classic paper, heightens the role of cultural factor influence on genetic phenotypes enough to severely weaken the anthropological interpretation.  However, I don't think it destroys it.

I think both positions are needed for the tension they produce. But as Reich says, the balance is definitely shifting - as painful as that may be some. It is one more thing that seems to be forcing science and society to confront fuzzy boundary issues on sensitive topics.

Here's a sample of the sacralization of the topic. This suggests rational conversation is all but impossible. A reconciliation to Haidt's two teloses seems all but inevitable.

*This is a critique most germane to the MSL1 formulation. MLS2 posits collectives reproducing collectives.

No comments:

Post a Comment