some ideas on "What is an explanation?"

Explanation is about answering questions of the form "why is B the case?".

I can see 4 different notions:

retrodiction: (i.e. inverse prediction)
if A predicts B, and B happens, then knowledge of A explains it.
Problem: although a barometer may predict rain the next day, knowledge that the barometer was predicting rain does not explain the rain (on some definitions/accounts of explanation). This notion is probabilistic but not temporal.

causal explanation (stricter than retrodiction): if A is known to cause B, after B happens, finding out that A is the case will answer "Why is B the case?". I think this is pretty uncontroversial. (although one should never say this to philosophers: I can already imagine you guys giving me a classical scenario). This notion is temporal and probabilistic.

derivation:
If B can be derived logico-mathematically from A, then finding out that A may answer the question "Why is B the case?". This notion is neither temporal nor probabilistic.

analogy:
Finding out that there exists an A analogous to B might satisfy your curiosity about "Why is B the case?".
If we only had this notion, we could never explain the explainer because of infinite regress. (Never mind that we always have to have some axioms that we take for granted). It seems to suggest coherence networks instead of deductive ones where there's a top and a bottom. This notion is probabilistic but not temporal.

Why the sex ratio in sexually reproducing species at reproductive age is 1:1

from the odd pages of a philosophy paper titled "Prediction":

[About Fisher's explanation]
Too briefly, the explanation is that (assuming equal parental cost to
produce offspring of either sex, ignorance of offspring quality, and setting aside
complications), no matter what the mating system, a parent will spread more copies of its
genes by producing offspring of the less numerous sex: since every successful mating
requires the genetic contribution of exactly one member of each sex, members of the less
numerous sex are more easily able to obtain multiple successful matings, be more choosy
about mates, or enjoy whatever reproductive advantages members of that sex enjoy in the
mating system. Even if a species' mating is structured such that a few successful males
do all the mating and most males do not mate at all, when males are less numerous than
females a parent will do better on average by producing sons with a proportional chance
of being one of the lucky few than daughters who are guaranteed to mate. Therefore, it
pays to produce members of whichever sex is more rare, ensuring strong selective
pressure against any mechanism that favors producing male offspring over female or vice
versa

In other words, the genes are using a 50/50 mixed strategy in the game of "choose the baby's gender". (remember game theory?) (note: this "choice" is always made through the sperm: the choice of mix between X-sperms or Y-sperms, so any genetic disposition to produce more males or females must be passed on to male babies)

In particular, if male children died more than female children, we should expect that more males will be born.
Likewise, if male foetuses+children died more than female foetuses+children, we should expect that more males will be conceived. I would like to see actual statistics on this... or maybe we could "postdict" the infantile death rates of previous centuries, if we observed a small deviation in the number of conceptions in some native populations today.

I used to think that we should expect to have more females (because one male can fertilize many females), but this argument convinced me.

Of course, if we did have an imbalance in the sex ratio, the parents' genes would need enough time to "know" about the sex ratio before the countervailing pressure made any effect (it's not clear how many generations would be needed for the right mutations to appear). In particular, genetic adaptation will not solve China's female-deficit anytime soon: it will take many generations.

But Fisher's argument doesn't account for population selection:
(1) populations with more females could have more children, or (2) a greater female-to-male ratio could promote peace. Both mechanisms would promote population growth, but the selection effect is on the level of the whole population, which is much weaker than individual selection: in such populations, a male child's expected number of children will be higher than a female child's... so grandparents that make more boys will be more successful. This will tend to bring the ratio back to 1:1, even in a growing population.

But it would be interesting if we found that individuals from some recently "exploded" populations (think poor countries) were found to be genetically more likely to produce more daughters than sons. It would have to be because some segments of the population were growing faster by producing more adult females than adult males (possibly due to a sudden decrease in female infant mortality or, a cure of male-child-promoting diseases, such as we might expect from the Chinese population if they suddenly cure hepatitis B, after having enough generations with it to adapt their child-sex-production-ratio).

These arguments can get pretty involved, and I'm feeling insecure about my reasoning here. Do you see the appeal of formalizing such arguments into a logical system? What about the appeal of diagrammatic representation of these hypotheses and speculations as an argumentation map, instead of a 1-dimensional text full of nested parentheses?

These sort of arguments count as qualitative reasoning, don't they?

This argument in particular, due to being so meta (not atypical in arguments about evolution), where individual genes are being selected on the basis of the genes of the whole population, makes me want to formalize it in something like epistemic logic.

The Logic of Evolution

(this should go on My Notebooks eventually)

Why do kin help each other?

The answer is: evolutionary selection pressure towards kin-helping behaviour.
Individuals who accidentally developed genes that promote kin-helping behaviours caused their genes to spread more than individuals who didn't have such genes. So more babies with kin-helping genes were born than babies without kin-helping genes (because the kin helped were more likely to have the same kin-helping genes).


But why would I have a strong preference to helping my brother, when all humans have 99.9% of their DNA in common with me? (Why should I have a strong prefererence to helping my fellow humans when monkeys already have 98% in common with me?)

Humans used to compete among each other for limited resources (is this correct?), so helping everyone equally would not have made any appreciable difference to humanity, whereas helping your kin would have promoted your family's genes, at the expense of other genes in the pool.
(Are such behaviors different in populations where help is not reproductively zero-sum population-wide, by virtue of their environment?)

Also, it's not clear which genes are selfish, and whether some are more selfish than others. The kin-helping gene only promoted itself, so in that sense it's selfish.


I'm not comfortable with these answers. I'm ok with my assumptions, but I'm not comfortable with the fallibility of my own reasoning here, and the "selfish gene" metaphor isn't that clear me.