General Recommendation: Full Accept
Overall Summary and Thoughts:
This paper is clearly articulated and reviews a whole host of important literature on a complex subject, but does so in a way that emphasizes reader comprehension regardless of intellectual background. This paper meets and surpasses its initial aim as a tutorial of the causal relationships between multiple and oftentimes competing levels of biological (or “functional”) organization, and how they relate or compare to sociocultural contexts. The express purpose of this paper, to make Williams Rule more widely known to improve any positive change effort, is also accomplished in this paper.
Admittedly, a lot of the material presented in this paper went over my head, as I have difficulty applying biological and genetic principles to sociocultural contexts. The section that outlines several analogies was particularly important for me to wrap my head around the matter presented thereafter (e.g., the Monopoly analogy).
However, I am not entirely convinced that Williams’ Rule is as important to the overall structure and argumentation of the paper as the author assumes. It seems to me that Williams’ Rule is merely another expression of MLS theory and the ensuing causal relationships between multiple levels of biological organization. I don’t find reference to Williams’ work as necessary for the general and overarching argument of the paper: that MLS theory is widely applicable across disciplines and sociopolitical contexts and can be employed to inspire real change. Why must we invoke Williams’ Rule to explain such occurrences? To me, Williams’ rule appears circular in its reasoning. Of course, group-level adaptations would stem from group-level selection. Isn’t this merely a matter of semantics?
I am also having difficulty placing “Williams’ Rule” within my understanding of multilevel selection, because there is just so much more behind the process of higher-level selection itself: not only the initial evolution of the higher-level or emergent trait in question, but also the selective maintenance of said trait. For example, the literature has indicated four potential hypotheses that could explain the initial evolution of adaptive genetic mechanisms as emergent, higher-level adaptations (e.g., recombination, adaptive mutation, TEs, WGDs, etc.): (1) they may confer an initial benefit by increasing mean population fitness in the short-term (e.g., Wagner & Draghi, 2010); (2) they could initially evolve through genetic hitchhiking on another directly selected trait or “second-order selection” (e.g., Futuyma, 2017b); (3) they may arise from neutral or nearly neutral processes (Agrawal, 2006); or (4) they may initially arise within genomes as selfish genetic elements that propagate themselves at the expense of or neutral to the individual until environmental change occurs and the selfish element is domesticated by the host because it provides novel genetic variation (e.g., Brunet & Doolittle, 2015; Soucy et al., 2015). Thus, the central claim of Williams’ Rule, that “group-level adaptations require a process of group-level selection” is too simplistic to capture all the necessary stages of higher-level selection. Individual-level adaptations in one environment could hypothetically evolve into a group-level adaptation within another environmental context, and in fact, I think the sex and adaptive mutation (SIM, hypermutator mechanisms) literature suggests that such mechanisms hold an individual-level benefit in one environment, and a group-level benefit in another, meaning that they have been fine-tuned to manifest different advantageous outcomes dependent on the environment. I further question the assumption that higher levels tend to be undermined by selection at lower levels. In the sex literature, it is an established fact among experts that sex is largely deleterious at an individual level (again, dependent on numerous ecological, populational, and genetic factors) and largely advantageous at a higher-level of selection above the individual. Thus, I think there are cases in the literature where lower-level selection is actually undermined and often overpowered by selection at higher-levels, so I question the claim of the author that “Williams’ theoretical claim has withstood the test of time but his empirical claim has not (p. 4)”. Which level of selection dominates depends, primarily, on how well-adapted a population is to its surrounding ecological environment. Moreover, timescale choice is particularly relevant in considerations of interlevel conflicts. The author does, however, slightly retrace their steps in the following sentence on P. 4. Perhaps the antecedent sentence is too categorical, with its insistence that “Williams’ theoretical claim has withstood the test of time”.
With this being said, I do appreciate that higher-level selective processes may be undermined by lower-level events more often in sociocultural contexts. So, in this case, I suppose Williams’ Rule is an important framework to use and understood to inspire positive change.
Several points to consider, respectively:
· (p. 2) “Individual-level adaptations are frequently not good for the group, species, or ecosystem.” I think the literature supports the claim that individual-level adaptations are either effectively neutral at a higher-level or are beneficial at a higher level. Most multilevel adaptations, genetic evolvability mechanisms included, strike a compromise between both levels, and which level dominates at any given timepoint depends on how well adapted a species is to their environment.
· (p. 2) “The fact that special conditions are required for higher-level adaptations to evolve was not widely appreciated until George C. Williams wrote his book Adaptation and Natural Selection in 1966, which became a classic within the field of evolutionary biology (see Sober and Wilson 2011 and Agren 2021 for modern appraisals).” Depending on how the author is using the term “special conditions”—i.e., reflecting causal relationships and a complex causal field that is determined by necessary and sufficient causal conditions—then I don’t know if I can agree with this statement. Special conditions for group selection were at least considered by David Luck prior to Williams, see for more Borrello (2010) Evolutionary Restraints: The Contentious History of Group Selection. Univ. of Chicago Press, Chicago.
· (p. 2) “A statement this definitive deserves to be called a law or a rule.” In the philosophy of biology, there is a growing concern with the usage of the term “law” within a biological context. Ernst Mayr wrote it best, “Today, the word law is used sparingly, if at all in most writings about evolution. Generalizations in modern biology tend to be statistical and probabilistic and often have numerous exceptions. Moreover, biological generalizations tend to apply to geographical or otherwise restricted domains. One can generalize from the study of birds, tropical forests, freshwater plankton, or the central nervous system but most of these generalizations have so limited an application that the use of the word law, in the sense of the laws of physics, is questionable” (Ernst Mayr, 1982: 19). I only mention this to be cognizant of the massive literature that questions whether we really have any determinant properties in the biological sciences at all. In place of law construction, biologists generally construct causal models of biological phenomena in the form of representational concepts, theories, or causal mechanisms that are restricted in space and time (Sober, 1997; 2000; sensu “concepts” in Mayr, 2004). Such models are law-like in that they make predictive generalizations about what will happen if a certain set of conditions are satisfied by a system. However, unlike the phenomena studied in non-quantum physics, pinpointing the causal mechanisms behind biological phenomena is often a laborious task, as any one effect—especially at the level of the organismal phenotype—entails multiple interacting causes that act at multiple levels of organization. Thus, given the level of uncertainty and indeterminacy, causal models in biology tend to be probabilistic (e.g., the propensity interpretation of fitness), exception-ridden (e.g., the Price equation), and/or asserted ceteris paribus (e.g., the Hardy-Weinberg; for more, see Anjum & Mumford, 2018). A list of exhaustive conditions must usually be met for a causal model to be accurate and predictive, yet casual interferers can always intercede biological causation. Thus, I wouldn’t even make any mention of Williams’ Rule as a law. I think it is perfectly ok as a rule, but perhaps note the work on causal modelling in the philosophy of biology. In disciplines with a great amount of causal complexity, such as in sociopolitical contexts, then rules or causal regularities are more prone to exceptions. By the way, no topic in philosophy is more important for scientists than the literature on causation. If you would like to delve into the literature, might I suggest: (1) Illari & Russo (2014) Causality: Philosophical Theory meets Scientific Practice. Oxford University Press. (this is a great introduction to the literature). (2) Anjum & Mumford (2018) Causation in science and the Methods of Scientific Discovery. Oxford University Press.
· (p. 4) “Williams himself changed his mind for the specific traits of disease virulence and female-biased sex ratios (discussed in Sober and Wilson 1998, p. 35-50).” It is worth mentioning that Williams was also particularly confounded on the evolution of sex. See the back-and-forth between Williams and Maynard Smith in the book Evolution in Space and Time: “With the evolution of sex now seen as a major issue in evolutionary biology, Maynard Smith (1971a; 1971b; 1974; 1976; 1978) and G.C. Williams (1971; 1975; 1978; 1980; 1988; with Mitton, 1973) spent the next decade attempting to find a theoretical model that would fit for the evolution of sex. At first, their mathematical models worked within the confines of the individual selection assumption and attempted to conserve the reductive foundations of Darwinian theory. But repeated attempts to find an immediate individual benefit were either inaccurate or inapplicable to natural populations (Dagg, 2016). This led Williams (1971: 161) to concede that “sexual reproduction must stand as a powerful argument in favor of group selection,” and Maynard Smith (1976: 257) to bemoan “one is left with the feeling that some essential feature of the situation is being overlooked” (interesting sidenote: later reflecting on this time, Maynard Smith [1998] noticed the errs of firstly conserving the theoretical traditions of Darwinian theory, as he would eventually accept sex as one of the best examples of higher-level selection because of the evolvability benefit that it confers to species). (p. 44)”
· One interesting and related sidebar: “From a different perspective, the fact that there is variation amongst individuals within populations points to the possibility that this could explain its initial evolution or even short-term selection for evolvability. When individuals vary in some respective trait, there is an increased causal efficacy of selection in the short term. Natural selection can operate more quickly on individuals within a population than on a population itself, hence why multiple levels of biological organization may not only be a happenstance of biology, but also an adaptation in and of itself. I would call this idea global evolvability—the notion that evolvability mechanisms operate at multiple levels of biological organization to streamline evolution at every level. Indeed, if evolvability operated at just the individual level, or just at the populational level, this would be less advantageous than if it operated at multiple levels with less causal constraints between levels (i.e., upwards and downwards causation). This is not to fall prey to teleology by claiming that evolution indubitably selected for global evolvability because it was advantageous for higher-level biological entities. But there is some evidence to suggest that multilevel causal interactions do exist. I suspect that most higher-level properties of populations, species, or taxa first evolved through individual-level selection operating within populations until they were selected over longer periods and/or following extinction-type events (i.e., macroevolutionary transitions).” (Distin, p. 102) Thus, when we are talking about MLS causation, it becomes doubly important to note the timescale at which we are viewing the evolutionary phenomenon in question.
· (p. 5) “Evolution is all about relative fitness”—I perfectly understand what you mean here, applying relative fitness at multiple levels of biological organization, but I think when we focus so much on relative fitness rather than absolute, we are overlooking the important history of why evolutionary biologists typically chose relative fitness as the most common metric of selection: “Reducing evolution to fundamentally a process that produces changes in allele frequency over time is thus problematic because we are forced to accept some dangerous assumptions about causation (e.g., monocausality) that limits our causal explanatory scope to within populations. Moreover, limiting our evolutionary scope to within populations thus sets specific abstract conditions under which to view the causal processes of evolution (Pigliucci, 2008a), which oversimplifies the causal dynamics of each respective process. However, early 20th-century technological limitations also caused population and evolutionary geneticists to be confined to performing within-population analyses. It was nearly impossible to study the extent of genetic variation between populations using Mendelian methods—which only became possible after the advent of proper genome sequencing techniques in the 1980s. Evolutionary research was therefore executed almost exclusively by studying allele frequency changes within populations rather than between populations (Nei, 2013). This is one reason why relative fitness is the most common metric of natural selection used in population genetics and mainstream evolutionary biology, compared to the measurement of absolute fitness that is often used in ecology (Bell, 2017). These technological limitations and simplifications ultimately restricted our causal models of evolutionary dynamics to within populations over shorter-time scales” (Distin, 2023, p.23).
· (p.5-7) Perhaps offspring size is another useful analogy to use. Marshall et al. (2008) showed that unpredictable environments might select for greater “bet-hedging” (i.e., evolvability) strategies, whereby mothers conceive offspring of variable size to increase the chances that some of these offspring are suited to the prevailing conditions. Cameron et al. (2021) thus recently demonstrated the effectiveness of using a multilevel approach to explain why such variability in offspring size exists.
· One minor spelling correction: “harrassed” to “harassed” on page 6.
· Perhaps another useful analogy: In my view, level “conflict” arises when a new trait and/or feature emerges in a given environmental context, and the environment subsequently changes, thus revealing the conflict. This is to say that new traits could likely first evolve under cooperation, but become a conflict following environmental change and/or if the trait is not “honed” to both types of environmental contexts over time. For a real-world example, human wars are a phenomenon of level conflict, with cooperation ensuing within groups, but with sufficient conflict between groups. Following “environmental changes” or simply peacemaking, the conflict will dissipate. The United States currently has no higher level of organization, no shared identity, thereby causing between-group conflict. When the higher level is fully formed, then there should be no lower-level conflict. Level “harmony” thus arises from compromises made between the various levels, which will ensue following sufficient time and selective pressure = “trait honing”. Conflict ensues when selection is acting at both levels, simultaneously. Cooperation ensues when selection is acting only at a higher level. Level conflicts will tend to lead to cooperation between levels, evolutionary stable strategies, or equilibriums reached (i.e., “compromises”).
· (p.6) “Because within- and between-group selection were both strong forces, the result was to maintain individual differences among the males. Aggressiveness was being maintained by within-group selection and passiveness was maintained by between-group selection.” Commensurate with the empirical literature on evolvability, since most genetic evolvability mechanisms show a dual-functionality—they are balanced between short- and long-term interests. DNA/RNA polymerases promote replication fidelity in stable environments, and replication infidelity in capricious environments. Sex exhibits both robustness and evolvability. Taken together, conflicts only arise in the short-term when significant time or selective pressure has not “honed” the trait in question. However, in the long-term, selection has likely honed the trait or its tried, failed and purified. Again, this is why the timescale is perhaps the most salient variable to consider in conversations of multilevel causation.
· (p. 7) “when the conditions specified by Williams’ Rule are met.” What conditions are these? I don’t think the author ever presents the actual conditions necessary for group selection (e.g., if a trait is causally connected to the fitness at any level of biological organization in a given environment).
· (p. 9) “Even single multicellular organisms, which are the quintessential example of an adaptive unit, have not entirely suppressed disruptive within group selection such as cancer and meiotic drive genes.” Keep in mind that evolutionary senescence at an individual-level can be considered an adaptation at higher-levels (or, at the very least, there isn’t enough selective pressure maintaining past “Medawar’s Curve”). Perhaps there isn’t an evolutionary advantage at a higher-level to completely suppressing cancer or meiotic drive, which aligns with the literature on hypermutation, since there also isn’t an advantage in the long-term to evolve completely perfect replication fidelity. This is a good paper on programmed life spans and evolvability: Mitteldorf J, Martins AC (2014) Programmed life span in the context of evolvability. Am Nat, 184: 289-302.
· (p. 10) “it was confined to the study of genetic evolution for most of the 20th century”. In addition, technological limitations caused population and evolutionary geneticists to be confined to performing within-population analyses for much of the 20th century. It was nearly impossible to study the extent of genetic variation between populations using Mendelian methods—which only became possible after the advent of proper genome sequencing techniques in the 1980s. Evolutionary research was therefore executed almost exclusively by studying allele frequency changes within populations rather than between populations (Nei, 2013).
· (p. 11) “A related narrative is that indigenous societies are inherently cooperative and respectful of their natural resources”. Historically speaking, pre-Neolithic indigenous or hunter-gathering societies were more likely to be cooperative, and nearly all modern indigenous societies are more respectful of their natural resources relative to modern, industrialized societies. See for more Nelson MK & Shilling D (2021) Traditional Ecological Knowledge: Learning from Indigenous Practices for Environmental Sustainability. Cambridge University Press.
· (p. 12) “For example, the first native American tribes to gain access to guns and to master horseback riding used their power advantage against other native American tribes (Gwynne 2011; Silverman 2016).” Same situation with the Māori in New Zealand. There was always inter-group violence. This is the benefit of MLS theory, since we can now comprehend the existence of species-level adaptations such as sex, implying that there are adaptations built into our genome that are there to benefit us as an entire species.
· This seems to be a major part of the story that was left out (likely for good reasons such as flow of the paper or limited scope): The mathematical models of Fisher and Haldane in the early twentieth century focused on within-species natural selection, and thus their models could not envisage adaptations at a level of selection higher than the individual (Nei, 2013). G.C. Williams (1966) was perhaps the most vociferous opponent of higher-level selective explanations, altering attitudes towards them for decades to come. He thought the idea of group selection was tenable, likely because of his interest in sex, but thought of it as a relatively “weak” evolutionary force, since individuals have a faster rate of turnover than groups. He instead saw group-level features as fortuitous group benefits, or features that incidentally benefit groups, but selection did not cause their initial evolution. For example, Williams thought there was an evolvability-type benefit behind the evolutionary maintenance of sex, but he thought this was unlikely the reason why sex evolved in the first place. In his eyes, the higher-level advantage of sex did not evolve out of higher-level selection, but rather as an incidental side effect (spandrel) of individual-selective processes. Species-selection theory was thus promoted by Stanley (1975a;b) as an alternative to group selection theory, marketed as a selective explanation for higher-level features of biological species that did not conflict with the time constraints of the individual selection assumption. Group selection failed because it worked within the time-constraints of individual selection, focusing on how lower-level individual interactions create emergent variation and differential ‘fitness’ at a group level. Species selection, in contrast, was built on the idea that species themselves are ‘individuals’ and therefore evolve in a similar fashion to individual selection, yet over greater geological time scales. What thus led to the gradual acceptance of species selection theory over group selection theory was the simple distinction that it made over evolutionary timescales (Goldberg, 2014).
· (p. 12) “Adam Smith’s metaphor of the invisible hand, which implies that the unregulated pursuit of lower-level interests robustly benefits the common good. This has been the guiding metaphor of neoclassical economics for the last 70 years and Williams’ Rule reveals it to be profoundly false (Wilson and Gowdy 2014)”. Yes, but didn’t Adam Smith also argue for social enterprises that benefit the common good at a collective and public level, such as education? If my memory serves me, he argued that certain industries should not be guided by the invisible hand. Just something to consider.
· (p. 12) “Something more is required to explain the evolution of such a prosocial behavior and that “something” is the existence of many such groups, variation in the frequency of the prosocial behavior among the groups, and the differential contribution of the groups to total evolving population.” Isn’t this “something” merely referring to the environmental context that selects for prosocial behavior? Again, I’d like to point to the salience of the environment for emergent traits. This is the part of the story that has always been left out, or at least, misunderstood to its fullest extent.
· (p.13) “Initial attempts to model the evolution of altruism, no matter how they were framed (e.g., group selection, kin selection, reciprocity) tended to make a core set of simplifying assumptions.” Similar to what happened in the sex literature. To start, early population genetic models made many simplifying assumptions that abstracted away from the natural complexity of evolutionary systems. Early models viewed the evolutionary dynamics of populations as infinitely large, sexually reproducing, and randomly mating with Mendelian heredity (Millstein & Skipper, 2007; Otto, 2009; Nei, 2013). However, natural populations are not infinitely large, they often nonrandomly mate, rates of sex have been observed to vary, and genes clearly do not follow perfect Mendelian patterns (e.g., epistasis, incomplete dominance, etc.). Therefore, such dangerous assumptions made it difficult to model certain evolutionary phenomena that did not fit within the stringent conditional clauses of these early models, such as sex (Otto, 2009; Meirmans & Strand, 2010).
· (p.13) As Henrich (2004, p. 16) put it: “Thus, a great deal of theoretical work shows that genetic group selection will only lead to substantial levels of altruism when groups are very small, migration rates are quite low, and the intensity of selection among groups is high compared to the intensity of selection within groups.” Does not take into account the primary factors underlying the manifestation of higher-level or emergent properties: timescale and environmental variability.
· (p. 15) “The reason that biological and human social systems often fail to behave adaptively is due to agents within the system pursuing their own adaptive strategies, to the short-term detriment of other agents and the long-term detriment of the system as a whole.” Reminded me of this quote by Dobzhansky, “A species perfectly adapted to its environment may be destroyed by a change in the latter if no hereditary variability is available in this hour of need. Evolutionary plasticity can be purchased only at the ruthlessly dear price of continuously sacrificing some individuals to death from unfavorable mutations” (Dobzhansky, 1937: 126-127) and this one by Mather: “Failure to achieve an adequate balance spells either its own (the individual) doom, on the one hand, or that of its descendants, on the other. Existing organisms must therefore have descended from those which had most adequately balanced the advantages of fitness and flexibility in the past” (Mather, 1943: 44).” The literature on the evolution of sex is a perfect example in this regard, as multiple studies have confirmed that the outcrossing or genetic mixing seen in sexual species is largely advantageous (but advantageous at a higher level of selection) relative to asexual counterparts. Sex causes evolutionary rescue over longer timescales and frequent spatial changes relative to asexual modes, implying that asexuality is not an effective method in the long term (also dependent on numerous factors and largely biological kingdom in question).
· (p.16) “The welfare of the whole earth system must be the ultimate unit of selection. Any other target of selection runs the risk of becoming disruptive at higher scales.” Since selective pressure is a necessary ingredient for selection, it is our job to illuminate the selective pressures applied at the level of the whole earth. Most modern issues transcend national boundaries and affect the entirety of our species (e.g., climate change, future pandemics, MAD). Only through an understanding of how these selective pressures may cause extinction in the long term, can we do something about it in the short term. We must create an evolvable yet robust society, similar to the genetic structure of most living organisms.
