Social evolution and the individual-as-maximising-agent analogy
Introduction
Does natural selection tend to maximise something? Does it produce individuals who act as if they maximised something? These two questions, or maximisation analogies, have long occupied evolutionary theorists. Early formal results suggested that natural selection acts so as to maximise mean fitness (Fisher, 1930). In contemporary evolutionary theory, the success of kin selection has caused many to consider that natural selection makes evolutionary individuals act as if they maximised their inclusive fitness. However widely – and perhaps falsely – believed, such claims have been mitigated or not formally established. According to their current interpretation, Fisher's results do not show that we should expect natural selection to cumulatively increase mean fitness; kin selection has led to a number evolutionary analyses but not to a justification of inclusive fitness maximisation. General criticisms about adaptationism, whether due to genetic constraints or to social contexts, have also fueled doubts as to whether natural selection should really be expected to maximise something at all.
Such frustrating results have led to a renewed exploration of the maximisation claims, through a number of frameworks. First, there are two distinct claims or analogies to explore, depending on whether nature or evolutionary individuals may be seen as maximisers. Second, there are two broad ways in which to explore them. Some start from general evolutionary results, which they hope to apply to concrete situations. Others aim to work their way up from maximisation results in specific contexts to more encompassing ones. Is one method preferable? How could we hope to generally determine the domain of validity of maximisation analogies - what maximises what under which conditions and in which evolutionary contexts?
This paper has two related aims. First, it intends to describe some recent works on maximisation analogies and assess their limitations. Second, it focuses on a recently developed approach in order to critically discuss its fruitfulness when compared to a rival approach, derived from Hamilton's rule, which has recently undergone criticism. Most of the paper focuses on the specific case of social evolution and on the question about whether evolutionary individuals maximise their inclusive fitness – which is probably the most widely believed maximisation claim in evolutionary theory. More precisely, I assess Okasha & Martens recent local approach to the individual-as-maximising-agent analogy, its robustness with respect to interactive situations. I then defend the relative merits of a comparable global approach, arguing that it is conceptually on a par and heuristically advantageous.
The paper unfolds as follow. Section 2 discusses two possible understandings of maximisation and two possible approaches – global or local – for exploring maximisation analogies. Section 3 surveys various global approaches used to investigate the maximising analogies in the context of social evolution and underlines their current limitations. Section 4 then tackles local approaches – so-called individual-as-maximising-agent analogies – focusing in particular on recent work by Okasha and Martens, and uses their very method to suggest that the existence of a general function that individuals would maximise is unlikely. Section 5 discusses the issue of individual control, which leads to one way of re-establishing the relevance and heuristic advantage of the global maximisation approach. Section 6 concludes on the fruitfulness of global and local approaches for the maximisation research program in evolutionary biology and mentions one avenue for future work.
Section snippets
Nature and individuals as maximising agents
Natural selection may be said to lead to maximisation in different senses, which allow one to map existing approaches. A first useful distinction concerns the level at which the maximising analogy is located. Nature itself may be seen as a maximiser; alternatively, evolutionary individuals (members of Darwinian populations) may be seen as maximisers. These two options constitute two possible analogies through which evolution may be understood.
The nature-as-maximising-agent (NMA) analogy is
Global approaches
Over the history of evolutionary theory, the nature-as maximising-agent (NMA) analogy has been developed almost as soon as formal approaches emerged, starting with Fisher’s (1930) ‘fundamental theorem of natural selection’. According to a naive interpretation of this theorem, the mean fitness of a population always increases under the action of natural selection – that is, it is as if natural selection strived to maximise the average fitness of a population. This diagnosis proved puzzling
Utility transformations
Following Martens (2016), Okasha and Martens, 2016a, Okasha and Martens, 2016b put forward a new local approach to make the individual-as-maximising-agent analogy precise and to determine when it holds. Their method is as follows. When faced with a given context of interaction, we should follow three steps. First, we should work out its evolutionary characteristics (determine which traits will evolve under which conditions). Second, we should represent it in a game-theoretic matrix and
Individual maximisation and acceptable preferences
What we have seen so far suggests a limitation for the local approaches to the individual-as-maximising-agent analogy. As bottom-up approaches, barring possible general results over a wide class of games, they need to investigate interactive contexts separately. Even if positive results are obtained for particular games, nothing guarantees that they will extend to other ones. Moreover, the approach depends on basic intuitions regarding plausible utility functions. For instance, while Okasha &
Conclusion
Does natural selection act as if it maximised something, or produces individuals who do? Such maximisation analogies can be explored through global and local approaches, depending on whether we proceed from general results then applied to specific settings, or from local results then gradually extended to bigger classes of situations. Because some global results have only been obtained under restrictive conditions, or under general conditions that do not square well with maximisation analogies,
References (26)
An interpretation and proof of the fundamental theorem of natural selection
Theoretical Population Biology
(1989)The hawk-dove game played between relatives
Animal Behaviour
(1979)A first formal link between the Price equation and an optimisation program
Journal of Theoretical Biology
(2002)Optimisation of inclusive fitness
Journal of Theoretical Biology
(2006)Fisher's fundamental theorem of natural selection revisited
Theoretical Population Biology
(1997)- et al.
Limitations of inclusive fitness
Proceedings of the National Academy of Sciences of the United States of America
(2013) The grammar of society - the nature and dynamics of social norms
(2006)Hamilton's rule and its discontents
The British Journal for the Philosophy of Science
(2014)- et al.
Kin selection and its critics
BioScience
(2015) - et al.
A theory of fairness, competition and cooperation
Quarterly Journal of Economics
(1999)