I got my doubts when I read an article by Christian Barry et.al. who put forward the idea that there is yet a third category, namely enabling, and that the bipartite doing-allowing distinction is not exhaustive. In order to substantiate their thesis the authors do not found their point only on theoretical arguments, but they test it in an experimental setting. Moreover, they don’t consider the distinction in general but they investigate only the distinction between doing and allowing harm. In their investigation they use four cases:
Push: A cart stands at the top of a hill. John pushes it. The cart rolls down the hill and injures Tom, who is sitting at the bottom of the hill.
Stayback: A cart is rolling down a hill. John could put a rock in the way of the cart that would stop it, but he does not. The cart rolls down the hill and injures Tom, who is sitting at the bottom of the hill.
“Push and Stayback are straightforward in the sense that their classifications under the traditional doing-allowing distinction have been uncontroversial among philosophers”, so Barry et.al. “Philosophers generally agree that Push and Push-like cases are cases of doing harm, and Stayback and Stayback-like cases are cases of allowing harm” (p. 68). Actually I don’t fully agree with this remark, but for this blog, I can ignore my criticism.
Two other cases used by the authors in their investigation are:
Interpose: A cart accidentally starts rolling down a hill. Tom, who is sitting at the bottom of the hill, will not be injured by the cart if he can get out of the way of its path. John puts a rock on the ground. The rock stops Tom, and he is injured by the cart.
Remove: A cart is rolling downhill towards a point where there is a rock that would bring it to a stop. John removes the rock. The cart rolls down the hill and injures Tom, who is sitting there.
While Push and Stayback are cases of doing and allowing, Interpose and Remove exemplify enabling, so Barry et.al. In order to explain this, they refer to a study by Barry and Øverland who introduce two factors that separate enabling from doing and allowing: “Relevant action, the first factor, obtains if the question of how an agent is relevant to some harm refers to some action of theirs.” (cf John’s pushing of the cart by which Tom is injured in Push.) “The second factor obtains if there is a complete, intact causal process initiated by the agent’s action that links this action to the harm.” (The cart pushed by John injures Tom in Push) (p. 70). Both factors are present in Push and absent in Stayback. However, in Interpose and Remove the first factor is present but the second factor isn’t: John’s action has an impact on the result of the cart rolling down (namely that Tom is injured), but the relation between John’s action (moving the rock) and Tom being injured is indirect. Therefore we say that moving the rock enables that Tom gets injured and that’s why enabling is a special category next to doing and allowing.
Is this idea also shared by non-philosophers? In order to investigate this, the authors presented several versions of the Push etc. cases to a number of test persons. In a first experiment testees were asked to classify cases as doing or allowing harm. The result was that Push cases were clearly seen as doings and Stayback cases as allowings. However Interposing and Remove cases scored somewhere in between: About half of the testees saw them as doings and the other half as allowings. This suggests, so the authors, “that the traditional bipartite doing-allowing harm distinction cannot capture Interpose and Remove cases.” (p. 77). That enabling can be seen as a separate category was confirmed in a second experiment. Now the test persons were asked to categorize cases as “doing harm”, “allowing harm” or “enabling harm”. It appeared that Interpose and Remove cases were clearly more often seen as enablings than Push and Stayback cases (which hardly were). A third experiment investigated the question whether enabling harm is normatively distinct from both doing harm and allowing harm by asking the testees in different test versions whether John should compensate Tom for the injury. Also now the Interpose and Remove cases emerged as a special category, though not so clearly as in the other experiments.
I think that the conclusion of Barry et. al. is convincing that there is not a bipartite distinction between doing and allowing harm, but a tripartite distinction between doing, allowing and enabling harm. This result has not only theoretical consequences. It is only practical, for it appears that enabling harm is a specific normative category. This means that it has to be allowed for when talking about questions of guilt, responsibility and compensation. What remains is the question whether the tripartite doing-allowing-enabling distinction applies only when someone is harmed, or whether this tripartition generally replaces the dichotomy of doing and allowing.