The discussion of the "mere means" principle that Parfit locates as the second part of Kant's Formula of Humanity is one that I have traced from the various accounts of it given from 2002 to 2008, culminating in the posting here. In this posting I'm going to cover the description of it that is given in the final published version which is in Chapter 9 of On What Matters.
The published version of the discussion of the mere means principle is very little changed on the previous formulations that Parfit gave of it. As with the earlier formulations he opens by describing the mere means principle as stating that it is wrong to treat anyone merely as a means and distinguishes between treating as a means and treating merely as a means. The latter involves some kind of objectification being added to the treating as a means and, due to this element of objectification, neglecting any sense of the "well-being and moral claims" of the other. As with earlier cases Parfit considers a peculiar objection concerning slave-holding and as with previous formulations uses this objection to arrive at the second mere means principle which does not merely say that it is wrong to treat someone merely as a means but also to "come close" to doing so. The "coming close" involves not entirely neglecting the well-being and moral claims of the other but instead giving too little weight to them.
This addition to the initial simpler formula is next mitigated by inclusion of a discussion of ways in which we avoid treating someone either merely as a means or come close to doing so. The ways in which this can be done are by either relating to them in a way which manifests that our behaviour is "governed or guided" by some relevant moral belief or concern or, apparently more challengingly, we would choose to bear a great burden for this person's sake. However, Parfit's apparent way of indicating ways of avoiding such conduct is rather slim in its scope since, as he admits, it is often unclear whether the first of these conditions is met. Similarly, the second way seems not to rule out that we might be treating someone merely as a means since the sacrifice we might be prepared to make for them in the given case need not prevent us from treating the person in question merely as a means right now.
Parfit does, however, offer a better point when he suggests that whereas treating someone as a means may only refer to intentional actions, that treating them merely as a means requires, instead, understanding of underlying maxims of conduct. This point is the first useful element of his analysis but is immediately followed by a complication that leads his analysis towards a conclusion that is not initially obvious (at least not for anyone who hasn't read Parfit's previous treatments of the question). The complication is a distinction that is introduced between acting in a way that treats someone merely as a means and regarding someone as being merely a means. The reason given for introducing this distinction is, however, an odd one since the case that is meant to justify it patently fails to do so. This is the case of gangsters who, whilst regarding others merely as means, may, for example, pay for their coffee rather than steal it simply because it is less trouble. Here, however, the question concerns not Parfit's alleged distinction at all, but rather one about whether certain kinds of action have moral worth.
Another case meant to illustrate the distinction also in fact tells us something else. This is the case of someone who marries a rich, aged person in order to inherit their money but does nothing, during the period of the rest of their life to harm them. Here the suggestion is that the actions in question, judged merely as actions, do not constitute a wrong but, this being said, they are actions that are not morally worthy. However the introduction of reference to questions of harm is given weight since harm being inflicted would automatically brand the actions in question as wrong. The example is under-specified however since, unless we assume that acts of deception are not themselves wrong, which would be a big assumption, we can hardly say that the actions undertaken here are ones that really meet the test of the mere means principle.
Parfit next formulates a third version of the mere means principle which directly brings in reference to criteria of harm as part of what demonstrates that we have treated someone merely as a means or come close to doing so. However, introducing this notion of harm as something extra is problematic. After all, it involves the assumption that there is something further required other and above the reference merely to treating someone merely as a means and this has not been established by Parfit's account.
Parfit next combines the mere means principle with the principle of consent that he took to be the first part of the Formula of Humanity to give an overall description of this formula. That would give the result that we treat people merely as means when we treat them in such a way that they could not rationally consent to our so treating them. This leads Parfit back to his trolley type cases which lead him to the conclusion that there is a basis, given his view of rational consent, to prefer some people to others in cases of moral dilemmas given that those not favoured would be capable of giving this rational (not actual) consent themselves. The reasoning in question is of a clear consequentialist type and is certainly a counter-intuitive view of the way the "mere means" principle might be thought to work!
In recognition of this point Parfit looks again at the mere means principle and assesses what he terms the "standard view" of it. On this view, which is related by Parfit to his third conception of the mere means principle, and hence builds in reference to harm straight from the start, it would be a violation of the mere means principle to treat people without their consent as a means of achieving some aim and thereby harming them. Notably, Parfit's description of the "standard" view refers to actual consent. However, Parfit states three immediate objections to this formula as given. The first is that harming people as a means may not involve harming the people whose consent is in question. This is true though not evidently to the point of the use of the "standard" view as a reply. Whilst there are ways of treating people as means which are not directly harmful to people whose consent we might need and which might harm others there would, on this criteria, still be a question about the consent of these others. It is also true, as Parfit points out, that we might be treating consenting parties as a means without treating them merely as a means but this is again beside the point since, in that case, we would be doing nothing to violate the "standard" view as stated. Finally, Parfit claims that in treating them merely as a means we need not be acting wrongly. This is the only point of Parfit's that needs to be considered, though, given it involves a different standard of wrongness to the "standard" view, it requires this other standard to be given a defence.
In order to give this defence Parfit reaches again for a kind of trolley case. In doing so his point is to suggest that cases of lesser harm could be "rationally" consented to even if not actually consented to as a way of ensuring an overall better outcome. However such an argument involves a number of important assumptions. Firstly, it requires us to accept that outcomes are the ways of measuring the question of what is involved in treating someone merely as a means. It thereby illicitly rules out other ways of assessing the question of what is involved in treating someone this way. Secondly, in invoking lesser harm it steps back from some of Parfit's earlier cases which stated that even death could be rationally consented to, so questions of commensurability of harms seem to undercut even the possibility of reference to agency on his full view which raises substantial questions. Thirdly, and finally, the argument for a different view of the wrong is not settled by examples which build in disputable assumptions since the examples effectively beg the question asked.
Subsequent to rejecting the "standard" view on the basis of his appeal to examples Parfit looks at another view of the mere means principle to which he voices objections. Onora O'Neill and Christine Korsgaard refer to deception or coercion as involving treating someone merely as a means and Parfit replies to this by using an example to show that deception could be adopted for an altruistic end that ultimately benefited someone other than oneself. This is certainly plausible though what Parfit does not here discuss, given the narrowness of his counter-example, is the way such benevolent treatment can quickly be the basis of tyrannical treatment of others. The guideline against deception might not be, as O'Neill and Korsgaard's statements suggest, completely absolute, but still, there are some very good reasons to take it as a general rule. The final part of the chapter returns to investigating further cases loaded in terms of being considered primarily in relation to harm.
There is little in this chapter likely to persuade one to modify the view of the mere means principle since nothing in Parfit's argument really leaves behind the appeals to question-begging principles and cases. In this respect, despite the modifications and extensions of this discussion that occur in the course of the evolution of Parfit's view, there is remarkably little real progress in the nature of this view.