The paradigm case of coercion could be said to be when a robber stops you and says, ‘Your money or your life’. Coercion involves the restriction of freedom (reduction of options), which causes that person to do what she does not want to do. Coercion is wrong when it harms a person or fails to respect that person's autonomy. That is a conceptual analysis of coercion.
Even professionals working in bioethics (which includes medical ethics), including Leon Kass, misuse this term. Embryos cannot be coerced since they are not persons and lack freedom of will. But more importantly, future people cannot be coerced by the act of genetic selection or cloning. Imagine that IVF produces two embryos, Anne and Bob. The parents choose Bob because that embryo has perfect pitch (or is a clone). Later in life, can Bob complain that his parents coerced or limited his freedom by selecting him on the basis of having perfect pitch (or being a clone)? No—he owes his very existence (all his options and freedom) to their act of selection. Without assisted reproduction and selection (or cloning), he would not have existed. It is metaphysical fact that those who owe their existence to a reproductive act cannot be coerced by that act. Even more broadly, they cannot be harmed by that act unless it makes their existence so bad that their lives are not worth living.
This is an awful conceptual analysis, one that seems to ignore almost all philosophical work that has ever been done on the topic. The problems start right at the beginning. A paradigm case is supposed to be a relatively simple and straightforward example, so that it can serve as a general reference point for other things; it needs to have relatively few atypical features. But threats of any kind are not relatively simple and straightforward; while people generally recognize that some kinds of threats can sometimes be coercion, determining why and in what way threats are coercive has never been easy, because they necessarily bring in issues of emotional possibility and social expectation. Virtually all the obvious candidates for paradigm cases of coercion are physical: someone seizing you and physically making you do something, someone holding your nose to force something down your throat, etc. Even the simplest cases of threat (e.g., dangling someone off the side of a building and threatening to drop them if they don't tell you something) are more complicated than such cases, with all sorts of distinctive and atypical features.
The point would perhaps not bear pointing out, since paradigm cases are only important as reference points, were it not for the obvious thing that is not in Savulescu's analysis at all: the reason why physical cases of coercion are better candidates for paradigm cases than emotional cases is that they are more blatant examples of use of force. To get emotional cases in requires looking at how and in what way one person can be said to force someone else to do something simply by threatening, which brings in the psychological and sociological questions. (This is the only reason historically, for instance, that we take threat to be coercion and not something different even if related: people argued that certain kinds of threat were as much a case of forcing someone as physical forcing.) Savulescu's analysis is missing any conception of forcing at all, a fact that is glossed over by his choice of paradigm case. The analysis itself is so general that it doesn't require any force at all. Do you know how I can reduce your options in such a way as to cause you to do what you don't want to do? By doing just about anything at all, in the right circumstances. If I insult someone in whose good graces you'd prefer to stay, this might cause you to apologize to them on my behalf even if you don't want to do so. And if you claimed I coerced you, I would say (in a candidate paradigm case of rejecting coercion as an explanation), "Nobody forced you to do it." The vast majority of our actions reduce the options of other people; that's why coordination in the form of society is often necessary. The number of ways in which our actions can cause someone to do something that they don't want to do is truly legion, and very few of them involve anything that most people would recognize as coercive. And note in particular that Savulescu doesn't give any way of connecting how the reduction of options relates to the causing. If I reduce your options so that you still have several options left, but you stupidly think it leaves you only one option, in what way did I actually coerce you? Yet it fits Savulescu's account of coercion.
It's possible that Savulescu is assuming something like Nozick's account of coercion, which does take coercion to be primarily threat-based; but this is not a "conceptual analysis of coercion" but a substantive and highly controvertible position on the subject. Nor does this do anything to alleviate the obvious problem of overgenerality.
The issue of harm and failure to respect as implying wrongness of coercion can be allowed to pass, although they are not part of a "conceptual analysis" of coercion, but points of substantive dispute. The next serious problem, though, is that there is no particular reason to think that Savulescu's analysis is complete -- either with respect to the analysis of coercion itself (we've seen that there are positive reasons to think it is not) or with respect to what makes coercion wrong. There is nothing wrong an incomplete conceptual analysis, except that Savulescu immediately goes on to treat it as complete, using it as a criterion for proper use and misuse of the term. But as long as it is incomplete, it cannot be ruled out that there are nuances that shift the border between appropriate use and misuse in ways that may not be obvious. (To take a simple example: Suppose someone gives 'killing a person' as a conceptual analysis for murder and then immediately starts criticizing other people for misusing the term. Is it not immediately obvious that, since the conceptual analysis is incomplete, treating it as complete will guarantee that the person gets the question, 'Is this a correct or incorrect use of the term 'murder'?", wrong in any number of cases?)
What is more, Savulescu's argument ends up not being consistent with taking his conceptual analysis as complete. Savulescu says, "future people cannot be coerced by the act of genetic selection or cloning", but nothing in his analysis implies this. The conceptual analysis gives two, and only two, conditions for something being coercion:
(1) It reduces someone's options.
(2) It causes them to do what they don't want to do.
Both of these can be met in the genetic selection cases. Certainly anyone's genetics might at some point in their lives cause them to do something they don't want to do; go to the hospital, perhaps. And there are perfectly straightforward senses in which your genetics can reduce your options. My genetics means that I don't have the option of doing things that require excellent vision. The only way Savulescu can get 'future persons' into the analysis at all is by reading (1) in a very, very particular way, one that is not at all obvious from his analysis, and is not at all an obvious candidate for how it should be read: it has to mean that coercion is reduction of the options that already existed (rather than normal options, or minimal acceptable options, or what have you). By that account, slave owners who arranged for slave babies were not coercing the slaves who grew up from those babies, because there was never a point at which the slaves in question had any option not to be slaves -- that option was ruled out from the beginning. This is not what anyone would have taken to be the obvious reading of condition (1), nor is it consistent either with our usual usage or the history of the term. If one doesn't take this extremely counterintuitive reading, neither (1) nor (2) rules out the possibility that someone can be coerced by means of the very act by which they are generated, nor do they imply that future people cannot be coerced by genetic selection or cloning.
I have no idea in what sense Savulescu's last sentence is supposed to be taken 'more broadly'; as his conceptual analysis of coercion states things, considerations of harm are not 'more broad' than considerations of coercion, merely overlapping. But even here Savulescu's claim is baffling: 'making someone's existence bad' is something we usually talk about in terms of harm, so it's an utter mystery how one would go about making someone's existence bad and yet not be harming them because you didn't make their existence so bad that life itself was no longer worth living. That's certainly not what anyone would usually say; it doesn't fit any of our paradigm cases of harm. And it doesn't follow from the conceptual analysis of coercion, since the conceptual analysis you would need is a conceptual analysis of harm, which we aren't given.
The same problems pervade the article. For instance, Savulescu says,
Concerns about coercion equally fail to apply to many acts of germline genetic enhancement. Engineering perfect pitch or increasing intelligence or giving a child a talent increases options and freedom. Coercion in such cases only exists if parents choose to then limit options.
But this doesn't follow from the conceptual analysis, either, which did not make any claims that the reduction of options had to be overall, rather than any sort of reduction of options. And it is quite obvious that one may simultaneously reduce options so that someone does what he or she doesn't want to do, while also increasing other kinds of options. Indeed, this happens in all sorts of coercive cases. A peasant who is kidnapped and forcibly made a eunuch to a king may have all sorts of options opened that did not exist before -- eunuchs have historically had all sorts of lines of power opened to them, so that someone might well have more options as a eunuch than as a peasant. But nobody would look at any of these to determine whether the eunuch in this instance had been coerced in the first place. Nobody thinks that it is relevant to the question of whether Ottoman Turks coerced young people into becoming Janissaries that the Janissaries then for all practical purposes ruled the Empire. If I torture you, or threaten you, or blackmail you into doing something that leaves you more free in the end, it's not obvious that the latter implies that I did not coerce you at all; and it is certainly not something that follows from Savulescu's conceptual analysis. There is a clear illegitimate shift here: the analysis considers only reduction of options, without telling us anything about how opening options would be factored in, and Savulescu in his example considers only opening of options, without considering the ways in which it would reduce options. There's no excuse of this, either; the people Savulescu is criticizing would certainly be focusing on the latter.
What's more, the whole thing begins to be quite comic. Savulescu by the end of the article has complained that (1) professionals working in bioethics and medical ethics often do not use the term in the way his conceptual analysis would require; (2) professionals working in law often do not use the term in the way his conceptual analysis would require; and (3) in everyday life, in matters where they have to deal with problems in medical ethics, people very often do not use the term in the way his conceptual analysis would require. Which shows that he does not quite grasp what a conceptual analysis does: it analyzes the concepts people use by beginning with how they use it and seeing what that requires about the concept itself; and while people may use concepts confusedly or inconsistently, a proper conceptual analysis will be able to explain why they would be confused. (Nor can Savulescu claim to be doing conceptual analysis in some other, weird way; he explicitly appealed to a paradigm example, and that is the whole point of a paradigm example in conceptual analysis: typical use.) The evidence is not sufficient to determine the analysis, but it is essential for doing it. Savulescu's conceptual analysis not only is defective in itself, it apparently floats free from almost all actual evidence about the concept. If your conceptual analysis cannot explain how almost everyone except yourself uses the term in question, you have not actually analyzed the concept that you are claiming to analyze.
Savulescu admittedly insists that he is not giving a full defense; but he does presuppose throughout his entire argument that this is a good, even if not best, analysis. But this is one of the worst conceptual analyses I've ever seen, and certainly the worst I've ever seen published in a professional journal. You could give a bright undergraduate the article and they could point out obvious problems with it. And a bad conceptual analysis doesn't establish what he claims -- "good philosophy—which may well include good counter—arguments to all my philosophical arguments in this paper—is essential to understanding and deciding bioethical issues—and law". I do in fact agree with this claim, at least. But Savulescu fails miserably at showing it. His conceptual analysis being a mess, it doesn't show that conceptual analyses would do any more than sow confusion. And his argument is not appropriately structured, even if his conceptual analysis were any good, for showing that it is "essential" to understanding and deciding bioethical issues and law, rather than say, merely occasionally a useful supplement. Indeed, nothing Savulescu says establishes any relevance to law at all; for all he shows, the 'conceptual confusions' he identifies that get involved in law might well just be necessary evils given all the many things law has to balance. There seems to be a rather large contingent of bioethicists who do not seem to grasp the elementary point that law, as a practical discipline, often has to operationalize, approximate, and rig up extrinsic qualifications and restrictions to work in the first place. Going around pointing out 'conceptual confusions' in the law is an idiot's game; the question is how one would eliminate them in law while still meeting all the goals that must be met by the law -- and that is sometimes very, very far from an idiot's game.
That's all I have to say about that, although the thing that Savulescu says in the article that had me hit my head on the desk -- literally, and I do mean literally -- was this:
David Hume famously described this ‘fact–value’ or ‘is–ought’ distinction. One of his greatest contributions to ethics was to observe that values cannot be read straight off natural facts. To do so is what GE Moore described as the naturalistic fallacy. Science and ethics are completely different kinds of enterprises.
Full discussion of this would take another long post. But setting aside the fact that 'fact-value' and 'is-ought' are not the same distinction, Hume manifestly does not think that "values cannot be read straight off natural facts"; he's a moral sentimentalist, which means that he thinks moral matters concern facts about moral feelings of approval and disapproval, which are natural facts (just not necessarily natural facts, depending on how the phrase is used, about the thing approved or disapproved). To put it in terms other than Hume's own, he's a naturalist about ethics. To claim that science and ethics are completely different kinds of enterprises on the basis of Hume's discussion of the way in which ethics can be part of the Science of Man is completely wrong, and there hasn't been much excuse for such a use of Hume in decades. I wasn't even born yet when people were pointing out that this sort of thing is a caricature of Hume, and not even a very good one. Alas, I fear that I will be pointing out that it is a caricature of Hume on my deathbed, be it many decades down the road.