Reflections on Cohen’s Account of Manipulation

Robert Noggle, Central Michigan University, Philosophy (Anthropology & Religion)

ORCID 0000-0002-4682-3825

Assigned DOI: doi.org/10.3998/jpe.9953

Shlomo Cohen’s book is outstanding, and I have learned much from it. His account of manipulation is quite different from mine, but that doesn’t prevent me from agreeing with much of what he says. I’ll begin by explaining why that is. Then I’ll highlight what I think are Cohen’s most important insights. Next, I’ll discuss my main disagreement with Cohen. Finally, I’ll examine whether Cohen’s insights pose a problem for my account of manipulation.

A (Partial) Failure to Disagree

Cohen argues that “manipulation” cannot be precisely defined because it is a cluster concept: Cases of manipulation are more or less prototypical depending on how well the metaphor of “mechanification” applies to them. And that depends on the extent to which they display features like intent, nontransparency, getting into the target’s head, exploiting the target’s vulnerabilities, being bad for the target, and being different from background influences. Because these factors can exist in greater or lesser number and degree, the mechanification metaphor can be more or less apt. This makes the concept of manipulation fuzzy: There is a continuum ranging from prototypically manipulative to prototypically non-manipulative influences, with no sharp boundary between the two.

Cohen’s account classifies many more cases of influence as manipulation than mine does.

Consider this example:

Significantly inflated self-esteem … diminishes readiness to listen to criticism… Rational criticism is rebuffed, and so a manipulative approach to the overconfident person – for instance, by targeting the few vulnerabilities he does acknowledge – then becomes the most rational tool … (Cohen 2026, 8).

Thus, Cohen counts it as manipulation to puncture this person’s overconfidence and inflated self-esteem by calling attention to those actual flaws that he recognizes. Superficially, this resembles gaslighting, a form of manipulation that reduces the target’s confidence in her own judgements, and a manipulative tactic called negging, which diminishes a person’s self-esteem, typically by calling attention to some shortcoming. The Mistake Account classifies these tactics as manipulation only when they induce mistakes. However, in Cohen’s example, these influences don’t induce mistakes. Instead, they correct the mistakes of over-confidence and inflated self-esteem. Consequently, the Mistake Account would not classify this as manipulation (so long as the correcting is not done by inducing some other mistake). In fact, the example nicely illustrates how the Mistake Account classifies the same kind of influence as either manipulative or non-manipulative depending on whether it induces a mistake (the “Dual Use Phenomenon”). In cases like this, where the Mistake Account sees a non-manipulative counterpart to genuine manipulation, Cohen’s account will often say that manipulation has occurred.

This seems like a large disagreement. But appearances can be deceiving. This is because manipulation is an unruly concept, and there are multiple ways to tame it. This makes manipulation different from a less unruly concept like lying. Normally, if we specify all the facts in a hypothetical case, most people will be confident, and most people will agree, about whether someone lied. Efforts to define lying can thus proceed from a fairly well-defined set of data about when something is a lie and when it’s not. When a proposed definition of lying doesn’t fit this data, we have a strong case for rejecting the definition. However, the concept of manipulation is far more unruly. There are cases that almost everyone describes as manipulation. But there are also cases where people are uncertain, or where they disagree, about whether something is manipulation. And there are cases that some might hesitate to call manipulation without objecting to others using that word. Cohen’s account of manipulation as a cluster concept, without well-defined boundaries as to when it does or does not apply, captures the fact that we are sometimes uncertain whether to call something manipulation. It might also capture the fact that we sometimes disagree about whether something is manipulation, since people might disagree about whether a given influence has enough of the characteristics in the cluster for the influence to count as manipulation.

By contrast, I focus on what we might call the “core cases” of manipulation—those that most people confidently call manipulation. My goal is to identify a feature that (1) is present in the core cases, (2) distinguishes them from influences that most people confidently agree are not manipulation, and (3) helps explain why the influences present in the core cases seem bad. We might then build a definition of manipulation from this feature. This kind of project is neither intended nor likely to be perfectly consistent with each and every use of the term “manipulation.” That’s because the unruly nature of the concept makes it inevitable that any such feature will be absent in some cases that some people might call manipulation, and present in some cases that some might not call manipulation. How should we respond to these mismatches?

One option would take them as evidence that the proposed definition is faulty. After having drawn this same conclusion about many proposed definitions, we might conclude, as Cohen does, that the project of defining manipulation is hopeless. A second option is aligned with a philosophical methodology called reflective equilibrium. It involves determining whether a proposed definition provides a sensible explanation of what manipulation is and why it seems bad. The more sensible the explanation, the more willing we should be to adopt it and abandon uses of the word that don’t fit with it. Consider this analogy: Long ago, the term “fish” applied to whales. When modern taxonomies re-classified whales as mammals, we faced a decision: Reject the new definitions for failing to match how we use the words; or adopt the new definitions because they offer a more sensible way to categorize different creatures. Sometimes, it’s worth revising how we use a word to fit a definition that explains and categorizes things in a sensible way.

Obviously, I favor the second option. I believe that the Mistake Account offers a sensible enough explanation of manipulation’s nature and moral status that we should let it guide how we use the word, even though it might differ somewhat from how the word is sometimes used.

This approach is well suited to certain useful tasks in applied ethics. When faced with cases where there is uncertainty or disagreement about whether manipulation occurs, we can use the presence or absence of the feature common to the core cases of manipulation to settle the question. For example, several years ago, traffic engineers painted lines across the pavement of Chicago’s serpentine Lake Shore Drive. Mostly, these lines are equidistant, but they get closer together near a curve, creating the illusion of increased speed. That illusion, which encourages drivers to slow down before a curve, has probably saved lives. But is it manipulation? If the Mistake Account is correct, the answer appears to be, “yes.” The design creates a mistaken impression of increasing speed. So, the Mistake Account classifies it as manipulation. Surely, it’s morally justified manipulation, but still manipulation. Thus, this approach can help us answer practical questions like: Does this or that advertising or sales tactic, or online influence, or nudge, do something relevantly similar to what happens in one of the core cases of manipulation? And when there is uncertainty or disagreement about whether an influence is manipulation, this approach provides a principled way to settle it.

The topic of manipulation raises many fascinating questions. The one I personally find least fascinating is: “What is the right way to define ‘manipulation’?” Given the concept’s unruliness, there may be no single right way to define it. Much depends on one’s goal. Mine is to tame the unruly concept of manipulation by isolating a feature that matches our most uniform and confident applications of it, so that we have a principled way to decide when something is manipulation and when it’s not. Cohen’s goal is a deeper and more profound exploration of the human condition. This requires painting with a broad brush rather than sketching with a fine-tipped pen. Cohen’s broad brush paints as manipulation many more influences than my fine-tipped pen does. But that’s less a disagreement than a function of different projects. Consequently, if I put aside the question of the “right” way to define “manipulation,” and substitute “influence” when Cohen says “manipulation,” I agree with much of what Cohen says.

Cohen’s Insights

The observation that we are not pure Kantian rational agents, who weigh up reasons in a detached, emotionless way, is now commonplace. Cohen works to establish the corollary: We will not do well at getting along together if we limit ourselves to some idealized Kantian form of rational persuasion that appeals only to detached, emotionless logic, devoid of any non-rational influences.

Cohen supports this claim with a panoply of richly described vignettes that will seem familiar to anyone who regularly interacts with actual humans. Their protagonists are people like us and those around us: People with insecurities and anxieties; people prone to inflated egos and hurt feelings; people who are guarded and avoidant, fearing rejection or commitment. Cohen’s vignettes show, time and again, how misguided it is to imagine social life proceeding entirely by rational persuasion, and how trying to do so would be ineffective and counterproductive. This is not because the people in Cohen’s vignettes are deeply flawed or profoundly dysfunctional; Cohen’s point is that even high functioning humans have foibles that can lead to misunderstandings, hurt feelings, missed connections, and other forms of “social friction.” Social friction requires a lubricant to keep the gears of our social life from grinding to a halt. That social lubricant is (what Cohen calls) manipulation. It consists of a myriad of ways of being less straightforward but more sensitive, less direct but more effective, less coldly rational but more warmly human. Cohen’s demonstration of how ordinary social life, as lived by actual humans, thrives on influences other than rational persuasion provides deep insights into the human condition.

Also insightful is Cohen’s linking of manipulation to mechanification. It’s easy to conflate this with the more common idea that manipulation treats a person as a mere object. But treating someone as a mechanism does not ignore that person’s rational agency in the way that treating someone as a mere object does. Instead, mechanification operates on and through a person’s rational agency, co-opting it for one’s own purposes. This co-opting is far better captured in the metaphor of mechanification than the metaphor of objectification.

Disagreements

Treating someone like a mechanism might seem like a bad thing, and perhaps the thing that makes manipulation bad. But Cohen provocatively claims that neither mechanification nor manipulation itself is inherently bad. On Cohen’s view, manipulation is not generally pro tanto bad (though certain instances of it are bad).

By contrast, I believe that manipulation is generally pro tanto bad. Here I must add a qualification that I make in my book (pages 170-171) but omitted earlier for the sake of brevity. I believe that there are special cases where manipulation is not even pro tanto bad. These are cases where the target has consented to the manipulation or voluntarily undertaken some activity in which manipulation is allowed and expected. This fact is more about consent than about manipulation, because consent often completely nullifies pro tanto wrongness, as when my agreeing to a boxing match nullifies the pro tanto wrongness of punching me in the face. But where consent is absent, my view is that manipulation is always pro tanto bad.

Cohen denies this, and his denial does not seem to have anything to do with consent. Instead, it is based on manipulation’s role in remedying social friction. This role, Cohen suggests, makes it implausible to regard it as generally pro tanto bad. He writes that a key part of “the relevant phenomenology of human relations associated with manipulation…has not received adequate attention. Once it does, I don’t think we are left with good-enough reasons to view manipulation as pro tanto bad” (Cohen 2026, 5-6). This “relevant phenomenology” involves the role of manipulation in remedying social friction. More specifically, Cohen writes that “the social lubrication functions of manipulation are so integral to and important for the success of human relations and social cooperation as to make it implausible to view manipulation as bad at its core” (Cohen 2026, 6). This claim, in turn, is supported by those wonderful vignettes I discussed earlier.

Now I agree that manipulation (on either Cohen’s account or mine) can often remedy social friction. But this fact is not enough to show that it is not pro tanto bad. Of course, some remedies are not pro tanto bad. An antibiotic (without significant side effects) that is a remedy for a bacterial infection is not pro tanto bad. But for severe, antibiotic-resistant infections, the only remedy might be amputating an infected limb. Surely, the loss of a limb is pro tanto bad. It is something to be avoided whenever possible and regretted even when it is necessary to prevent something worse. It can be a remedy despite being pro tanto bad for the simple reason that sometimes something bad is the only remedy for something worse. So, the fact that manipulation is the remedy for social friction does not, by itself, establish that it is not pro tanto bad. The question is whether socially lubricating manipulation is more like an amputation or an antibiotic. Is it completely benign—not even pro tanto bad—like an antibiotic? Or is it the lesser evil, like an amputation, a pro tanto bad thing to be chosen only when it is the lesser evil, and even then with reluctance and regret?

The answer almost certainly depends on what kinds of manipulation we are using to remedy the problem of social friction. The expansiveness of Cohen’s account of manipulation classifies a vast array of influences as manipulation; I suspect that many of them are more like an antibiotic—not pro tanto bad—than like amputation. But this very fact helps explain why one might want a narrower definition of “manipulation,” which applies only to forms of influence that are pro tanto bad.

Earlier, I suggested that my disagreement about how to define “manipulation” is less important than it may seem. But that does not mean that it is not important at all. The point at which Cohen says that manipulation is not generally pro tanto bad is when I do start worrying about definitions. The reason why I worry is that the term “manipulation” carries a negative moral connotation. This makes it different from a more neutral word like “influence.” “Jones influenced me” is a report. “Jones manipulated me” is typically a complaint. Calling something manipulation casts it in a negative light in a way that is much different from calling it influence. Rather than seeing this connotation as a data point to be explained, Cohen seems more interested in explaining it away by suggesting, in effect, that manipulation only seems bad because the most prototypical instances of it are bad (Cohen 2026, 4) or because it’s “frequently” bad (Cohen 2025, 119).

My approach to manipulation is revisionist in that it requires us to be ready to give up some initial judgments about whether something is manipulation. Cohen’s view is far less revisionist in this way. But it is revisionist in denying the commonsense idea that manipulation is generally (pro tanto) bad—an idea which I think is the correct explanation of the term’s negative connotation. The cost of Cohen’s less revisionist claims about what counts as manipulation is a more revisionist claim about its moral status.

This strikes me as a high cost. Characterizing manipulation as not typically (pro tanto) bad robs the concept of its utility as a way to call out influences that are (pro tanto) bad—ones that should be avoided absent a very good reason, such as consent or the fact that they are the least bad way to prevent something much worse. Indeed, the very fact that social life is so full of friction only heightens the importance of being able to name and shame certain forms of social lubrication as typically (pro tanto) bad. This does not mean that manipulation—even on this narrower definition—can never be justified. It just puts us on notice that there had better be a good reason to resort to it. Sometimes, such reasons are present; sometimes manipulation is part of something entered into willingly, and sometimes it’s the only way to prevent something even worse. But there is value in marking off a class of influences as morally disreputable, to be used only with a good reason. Indeed, if we didn’t have a concept of morally disreputable influence, we would need to invent one. But we do—it’s “manipulation.” Nor do we need to strip away its negative moral connotation to create a more neutral term for influence, for we already have a neutral word for this as well—it’s “influence.” So, while I don’t think that it is incorrect to define the term “manipulation” as expansively as Cohen does, I think that there are strong pragmatic reasons not to.

Potential Challenges to the Mistake Account

The forgoing assumes that we can define “manipulation” so that it applies only to influences that, absent consent, are pro tanto bad. Not surprisingly, I think that the Mistake Account does this. But now a new question emerges: Does the Mistake Account succeed? The Mistake Account claims that, absent consent, manipulation is always at least somewhat bad, though sometimes it is only trivially bad, and sometimes it is justified as the lesser evil. Do any of Cohen’s vignettes involve a non-consensual influence which (1) the Mistake Account classifies as manipulation and (2) is not pro tanto bad? If so, that would spell trouble for the Mistake Account.

There are two reasons for thinking that the answer is no. The first is that it’s independently plausible to think that (non-consensually) inducing a mistake is pro tanto bad. The second is that it seems likely that the influences that Cohen sees as not pro tanto bad will fall into one of three categories, none of which pose a problem for the Mistake Account: (a) influences that the Mistake Account does not classify as manipulation; (b) manipulation whose pro tanto badness is nullified by consent; and (c) manipulation whose pro tanto badness is outweighed by the bad consequences it prevents.

We have already examined one case of influence which Cohen sees as not pro tanto bad, but which the Mistake Account does not classify as manipulation—the example where the influencer corrects mistakenly high self-esteem and over-confidence. Here is another:

“anxieties… diminish responsiveness to reason, and therefore the capacity to cooperate rationally. Quelling anxieties to restore cooperation cannot… be done through reasoning. Soothing manipulations geared toward distraction and changing of mood are the way to connect across the anxiety barrier” (Cohen 2026, 7).

The anxiety that Cohen envisions here seems clearly excessive and thus mistaken. Consequently, the Mistake Account implies that reducing it is correcting a mistake rather than inducing one, and so would not count as manipulation (at least insofar as the correcting is not accomplished in a way that induces some other mistake). Here is one more:

“Indifference…frustrates social engagement and cooperation.… Surmounting the indifference barrier may well necessitate manipulative shaking up – for example, by presenting things in more shocking and sensational ways” (Cohen 2026, 7).

Since it is not a mistake to care about what really is important, influence that replaces mistaken indifference with appropriate concern induces no mistake (so long, again, as the replacing is not done in a way that induces some other mistake). Hence, the Mistake Account doesn’t classify it as manipulation.

Now let’s consider influences that the Mistake Account would classify as manipulation, but where the pro tanto badness is nullified by consent. Cohen writes that the “large set of positive manipulations that enhance the good life… include manipulations that are enjoyed for their own sake…” (Cohen 2026, 6). Cohen is correct in observing that manipulation is sometimes entertaining. Stage magicians and pickpockets both engage in manipulative misdirection. A novel and a conman’s story about the unfortunate Nigerian prince both induce empathy that is mistaken because its target does not exist. The difference, of course, is consent. As I noted earlier, consent can nullify the badness of many otherwise (pro tanto) bad things, including manipulation. Most examples of entertaining or playful manipulation are parts of practices to which we at least tacitly consent. This does not demonstrate that manipulation is not generally pro tanto bad. It just demonstrates that consent (even if tacit) often nullifies something’s badness.

It's reasonable to suppose that parties who voluntarily embark on romance might at least tacitly consent to some harmless manipulation that may be part of the enterprise. Thus, the woman in Cohen’s dating example probably does nothing wrong when she engages in harmless manipulation to help advance a relationship that both parties want. Notice, though, that this tells us less about the moral status of manipulation than the moral magic of consent. Romantic manipulation undertaken on an unwilling subject is certainly not benign; it’s stalking or harassment. The fact that consent may keep romantic manipulation from being pro tanto bad means that we cannot use the non-badness of such consensual manipulation to establish a more general claim that manipulation is not pro tanto bad.

A final kind of case is where the influence that is classified as manipulation by the Mistake Account is pro tanto bad, but justified by the circumstances. Such cases range from easily justified trivial manipulation to serious manipulation that is justified only when it prevents something terrible. Examples of the former include a teacher feigning enthusiasm for a dull subject, or using gentle peer pressure to encourage a shy friend to join at karaoke. Cohen’s story of the spouse with a crush on her husband’s best friend (Cohen 2026, 9) lies at the other end of the range. Misleading one’s spouse about a potential threat to their marriage seems, to me at least, like a clear example of something that’s pro tanto bad. But it may be justified to prevent destroying a marriage over feelings that are irrational and temporary. If manipulation is the right choice, it is not because it’s not pro tanto bad; it’s because its badness is outweighed by the fact that the alternative is worse.

Obviously, this is only a small sample of Cohen’s examples. But perhaps it’s enough to indicate why I am optimistic that the Mistake Account can meet the challenge they pose. While I agree with Cohen that non-rational influences are not always pro tanto bad, none of his examples provides a clear counter-example to the claim that influences that the Mistake Account classifies as manipulation are (absent consent) pro tanto bad

Conclusion

Let me close by (re-)emphasizing that, despite my disagreement about terminology, I find Cohen’s book to be a fascinating, insightful exploration of the human condition, our nature as social beings, and the various forms of influence that enable us to get along together.

References

Cohen, Shlomo. 2025. The Concept and Ethics of Manipulation. Cambridge University Press.

Cohen, Shlomo. 2026. “How to Think about the Basic Moral Status of Manipulation.” Journal of Practical Ethics Forthcoming.