Interdependence and Causation: An African Philosophical Challenge to Lewis's Counterfactual Analysis — Epoche B2
Causal Density Is Not Constitutive Relationality: Ubuntu, Menkiti, and the Later Lewis The objection and what it would need David Lewis's counterfactual analysis is often said to fail for systems whose parts are deeply interconnected — ecosystems [1] , economies, communities. The reasoning is that the analysis asks what would have happened had one event not occurred, and in a tightly woven system removing one strand changes everything, so there is no nearby world in which the rest stays put and the comparison is uninformative. This essay argues that the objection as usually stated confuses two different things. One is causal density : many factors, feeding back on one another, so that the removal of any one has wide effects. The other is constitutive relationality : the identity of an item being partly made up of its relations, so that the item cannot be specified apart from them. Only the second would trouble Lewis [2] , because only the second makes the counterfactual antecedent unspecifiable. The first is a difficulty about knowing what would have happened, not about whether there is a fact of the matter. Two consequences follow, and the essay defends both. First, the ubuntu literature does not supply constitutive relationality in the strength required, and the reason is internal to that literature — Kwame Gyekye's criticism of Ifeanyi Menkiti's account establishes it [3] . Second, where causal density is the real problem, Lewis himself supplied the remedy in 2000, and it is a better fit for interdependent systems than the 1973 analysis it replaced. The thesis would be false if the relational personhood defended in African philosophy did entail that a person's act cannot be specified independently of the community [4] , or if the influence account failed on densely connected cases. The 1973 analysis and the trouble it ran into In 'Causation' Lewis defines causal dependence between two distinct events c and e by a pair of counterfactuals: had c occurred, e would have occurred, and had c not occurred, e would not have occurred. Causation is then the ancestral of that relation — c causes e when a chain of causally dependent events links them. The chain device exists because of pre-emption, which is the objection that has driven the whole literature. Suppose an assassin shoots and a back-up assassin stands ready to shoot if the first does not. Had the first not fired, the victim would have died anyway, so the second counterfactual fails and the analysis wrongly denies that the first shot was a cause. Lewis's 1973 answer is that the actual process runs through a chain of intermediate events, each dependent on the last, whereas the back-up's chain never got started. The answer works for simple cases and not for others. Where the back-up is cut off only at a late stage, or where several sufficient processes run to completion together, no chain of dependence links the actual cause to the effect, and cases of this kind accumulated for twenty-five years. The interdependence objection is a member of this family, though it is rarely recognised as one. When it says that no single contributing factor was necessary because others would have taken up the slack, it is describing pre-emption by redundancy. It is not a new difficulty arising from holism; it is the oldest difficulty in the literature, wearing different clothes. What ubuntu does and does not assert The stronger form of the objection makes an ontological claim rather than a claim about redundancy: that in a relational world the relata themselves are not separable, so 'had this act not occurred' does not pick out a determinate alternative. African philosophy is invoked at this point, usually through the Nguni saying that a person is a person through other persons. It is worth asking what philosophers who have defended that saying take it to mean, because they disagree with each other, and the disagreement settles the question. Ifeanyi Menkiti gives the strong reading. Personhood, on his account, is not conferred at birth but acquired: the community makes the person a person, through a process of incorporation marked by ritual and by the assumption of obligations, and a newborn is not yet a person in the full sense. Personhood is processual and communally conferred, and the community is prior in the order of being, not merely in the order of upbringing. Kwame Gyekye's reply [5] , in Tradition and Modernity , is that this proves too much. If the community confers personhood, then those who fail to acquire it — the infant, the disgraced, the one who never takes up the obligations — lack the standing that grounds their claims on others, and the moral content that made the doctrine attractive drains away. Gyekye's moderate communitarianism holds instead that the individual has an intrinsic core: a capacity for evaluation and for revising inherited values, which is what allows a member of a community to criticise it. Communal relations shape the person profoundly and do not constitute the person's identity all the way down. Thaddeus Metz's attempt to give ubuntu a principled formulation for moral theory points the same way [6] . What he extracts is a normative claim about which relationships have value — roughly, that one should honour relationships of identity and good will — not an ontological claim that persons are exhausted by their relations. Read through its own internal debate, then, the tradition supplies causal and formative density in abundance, and constitutive relationality of the required strength only on the strong Menkiti reading, which Gyekye rejects for reasons that have nothing to do with defending Lewis. The later Lewis Meanwhile the difficulty that is real — density, redundancy, absence of any single necessary factor — has an answer that the interdependence objection generally ignores, because it addresses the 1973 paper as though nothing followed it. In 'Causation as Influence' Lewis replaced the all-or-nothing test with a gr