I agree with Professor Adil Haque’s description of the ICRC’s bifurcated position on voluntary human shields. I have accordingly added a note in my original post which now refers to his Letter to the Editor and this reply.
It might also be valuable for readers to have a more elaborate account of the ICRC’s bifurcation. According to the ICRC’s Interpretive Guidance on Direct Participation in Hostilities (DPH), there are two types of voluntary human shields:
Category 1: Voluntary shields who are DPH (i.e., lawful targets):
This class includes civilians who (voluntarily and deliberately) physically block or impede an attack on a military target and may also include civilians who shield a military target in such a way as to undermine the physical capacity of the attacker to “identify” or “destroy” the object.
In its explanation of category 1, the ICRC Guidance states: “This scenario may become particularly relevant in ground operations, such as in urban environments, where civilians may attempt to give physical cover to fighting personnel.”
Category 2: Voluntary shields who are not DPH (i.e., not lawful targets):
This class includes civilians who (deliberately and voluntarily) position themselves in or near a military target to protect it from attack (through use of their legal immunity from attack), but who do not physically hinder the attacker’s capacity to identify or destroy the target. In other words, their actions would make it more legally difficult (but not more physically difficult) for the attacker to destroy the military target due to the increase loss of civilian life.
In its explanation of category 2, the ICRC Guidance states: “[I]n operations involving more powerful weaponry, such as artillery or air attacks, the presence of voluntary human shields often has no adverse impact on the capacity of the attacker to identify and destroy the shielded military objective.”
For an important debate on the validity of the ICRC’s bifurcated view—see an article by Prof. Michael Schmitt (pp. 732-35) criticizing the Interpretive Guidance and an article by Nils Melzer defending the Guidance in response to Prof. Schmitt (pp. 869-72).
Here are the relevant passages from the Interpretive Guidance:
Voluntary human shields: The same logic applies to civilians attempting to shield a military objective by their presence as persons entitled to protection against direct attack (voluntary human shields). Where civilians voluntarily and deliberately position themselves to create a physical obstacle to military operations of a party to the conflict, they could directly cause the threshold of harm required for a qualification as direct participation in hostilities. This scenario may become particularly relevant in ground operations, such as in urban environments, where civilians may attempt to give physical cover to fighting personnel supported by them or to inhibit the movement of opposing infantry troops.
Conversely, in operations involving more powerful weaponry, such as artillery or air attacks, the presence of voluntary human shields often has no adverse impact on the capacity of the attacker to identify and destroy the shielded military objective. Instead, the presence of civilians around the targeted objective may shift the parameters of the proportionality assessment to the detriment of the attacker, thus increasing the probability that the expected incidental harm would have to be regarded as excessive in relation to the anticipated military advantage. The very fact that voluntary human shields are in practice considered to pose a legal – rather than a physical – obstacle to military operations demonstrates that they are recognized as protected against direct attack or, in other words, that their conduct does not amount to direct participation in hostilities. Indeed, although the presence of voluntary human shields may eventually lead to the cancellation or suspension of an operation by the attacker, the causal relation between their conduct and the resulting harm remains indirect.
Depending on the circumstances, it may also be questionable whether voluntary human shielding reaches the required threshold of harm. The fact that some civilians voluntarily and deliberately abuse their legal entitlement to protection against direct attack in order to shield military objectives does not, without more, entail the loss of their protection and their liability to direct attack independently of the shielded objective. Nevertheless, through their voluntary presence near legitimate military objectives, voluntary human shields are particularly exposed to the dangers of military operations and, therefore, incur an increased risk of suffering incidental death or injury during attacks against those objectives.
 This view was generally shared during the expert meetings (Report DPH 2006, pp. 44 ff.; Report DPH 2008, pp. 70 ff.).
 During the expert meetings, this scenario was illustrated by the concrete example of a woman who shielded two fighters with her billowing robe, allowing them to shoot at their adversary from behind her (Report DPH 2004, pp. 6 f.).
 See Art. 51  (a) AP and, for the customary nature of this rule in international and non-international armed conflict, Customary IHL, above N 7, Vol. I, Rule 14. For the relevant discussion during the expert meetings, see Report DPH 2004, pp. 6 f.; Report DPH 2006, pp. 44 ff.; Report DPH 2008, p. 70.
 While there was general agreement during the expert meetings that involuntary human shields could not be regarded as directly participating in hostilities, the experts were unable to agree on the circumstances in which acting as a voluntary human shield would, or would not, amount to direct participation in hostilities. For an overview of the various positions, see Report DPH 2004, p. 6; Report DPH 2006, pp. 44 ff.; Report DPH 2008, pp. 70 ff.
 See also Art. 51  and  AP , according to which any violation of the prohibition on using civilians as human shields does not release the attacker from his obligations with respect to the civilian population and individual civilians, including the obligation to take the required precautionary measures.
 See Report DPH 2004, p. 7; Report DPH 2008, pp. 71 f.