Are you concerned that, although Big Brother is not watching you, your phone might be? Or are you troubled by the fact that the search engine you use while shopping on the Internet seems to know what you want before you order it or that the social media platforms you use appear to sense your mood? Cass Sunstein’s illuminating and engaging new book might help you understand what is going on. Although he does not engage in detail with the dark patterns and deceptive design embedded at the core of many of the apps, platforms and their facilitating AI infrastructure that now mediate our lives – Mark Leiser’s Dark Patterns, Deceptive Design and the Law (2025) does that – he nevertheless throws significant and harrowing light upon a fundamental issue. For Sunstein, along with others, thinks that the dark patterns-cum-deceptive design embedded in such technologies are most likely manipulative and, in the process of unpacking that thought, his book does three things. Part I elucidates and defends an account of the nature of manipulation; Part II applies that account to three scenarios (deep fakes, sludge and the Barbie problem); and Part III looks to the future. I concentrate mainly upon Part I here because it is the principal load-bearing section of the book.
Sunstein’s account of manipulation attempts, in the absence of “a unified account” (p. 18), to occupy some of the ground shared by numerous philosophical explications of the idea. His favoured elucidation is that “[m]anipulation is a form of influence, intended to affect thought or action (or both), […] not involving lies or deception, that does not respect its victim’s capacity for reflective and deliberative choice” (p. 19). While not entirely satisfied with this account – “it is a bit too broad” (ibid.) – he regards it as the best available, principally because it “captures many (not all) cases of manipulation, … [and specifically] cases that should concern those involved in policy and law” (ibid.). Developing an account of manipulation useful in such contexts, and which also captures “ordinary usage” (p. 13), is Sunstein’s main aim. He also holds that “[i]n ordinary language, the idea […] is invoked by people who are not committed to controversial views about psychological or philosophical questions, and it is best to understand […] [it] in a way that brackets the relevant controversies” (p. 20).
Sunstein claims that manipulation is not the same as coercion, “the two [being]… qualitatively distinct and each is bad, when it is bad, in its own way” (ibid.). It is, by contrast, closely related to lying-cum-deception, there being an “essential connection” (p. 26) between the notions. He grounds this claim by arguing that manipulation and deception function in the same way, both failing to “respect people’s capacity for reflective and deliberative choice” (ibid.) and creating a sense of betrayal and outrage in victims. The first part of that bundle constitutes the core of what is wrong with manipulation for Sunstein, it being unclear whether the sense of betrayal and outrage is merely a side effect of manipulation or a precondition for it. He sometimes characterises that wrong slightly differently, but presumably synonymously. Hence there is “a more-or-less identifiable category of actions [that] counts as manipulation; that those actions do not respect people’s capacity for agency; that those actions are often unacceptable for that reason; that those actions often lead people in directions that make their lives worse […]. Manipulators try to seduce choosers to do as they wish, and they often succeed” (p. 52).
If the wrong of failing to respect people’s capacity for reflective and deliberative choice is the same as that of failing to respect people’s capacity for agency, then there is a clear way in which it can be avoided even in the presence of manipulation. Consent eradicates the wrong, says Sunstein, provided it is explicit, although there are some circumstances in which inferred consent might also be effective (p. 44). This might lead one to think that explicit or transparent manipulation cannot be wrong, as, if one can see it, one can presumably either accept or avoid it. This is not quite Sunstein’s view though, for “[s]ome acts can be both manipulative and fully revealed to those who are being manipulated” (p. 45). He invokes the use of graphic, harrowing pictures that sometimes accompany health warnings (e.g. the images often reproduced alongside the warning that “Smoking Kills”) to support this claim. However, he thinks that, overall, “[c]overtness and hiding […] are common features of manipulation” (ibid.).
Even this very compressed statement of the content of Sunstein’s account of manipulation shows, I hope, that it is admirably clear, although it cannot do justice to his sharp, engaging and often amusing prose. His account is also constructed with a lightness of touch which will surely please general readers, although it may not satisfy those of a more analytical, philosophical bent. One thing that is less clear in Sunstein’s account is the fact that it is incompatible with some of its competitors, a result no doubt of his attempt to build upon shared ground. So, for example, while Sunstein commends Robert Noggle’s “mistake account” of manipulation, the two accounts look incompatible not just in their content but also with regard to their status. Sunstein, as we have already noted, thinks a unified account of manipulation – one which captures all significant instances of what we would regard as manipulation in ordinary usage – is unavailable. That, however, is exactly the kind of account that Noggle offers and defends.
Sunstein also suggests that his account of manipulation can be supplemented by another he admires, that offered by Anne Barnhill. Barnhill’s account, as Sunstein acknowledges, is broader than his own although that is not the key difference between them: Barnhill claims that manipulation does not extend to instances in which one is influenced in one’s own interest and Sunstein claims that it does. This again looks like a significant difference and, along with the observation about Noggle’s account, illustrates a more general point: that the differences and incompatibilities between existing accounts of manipulation are perhaps deeper and more important than Sunstein is inclined to think. And, if that is so, his effort to occupy the middle ground is problematic, there being genuine doubt as to whether it actually exists.
There is, further, an alternative account of manipulation that Sunstein himself offers in Chapter 6 of the book. He thinks that account captures more territory (p. 19) than, and is “different” (p. 11) from, the one offered in Chapter 1 which was outlined above. The Barbie problem that preoccupies Sunstein in Chapter 6 has this structure: “some person P engages in some action (for example, holds a party, serves alcohol, markets a product) that forces another person Q to make a choice (attend or not, purchase or not) that reveals information about Q that Q does not want to disclose, and that requires Q to decide whether to give a signal that Q would prefer not to give” (p. 116).
Sunstein wants to say that where Q’s dilemma does not arise by accident but is the result of P’s clever, mischievous or self-interested action, then there is manipulation. This might well run counter to the account of manipulation in Chapter 1, if we think that, in acting as they do, P respects Q’s agency-cum-capacity for reflective and deliberative choice yet still manipulates Q. Yet the fact that Q is “forced” to act in response to P’s action surely raises some doubt about that. My point, though, is not to highlight the difficulty in resolving this particular issue. Rather, it is to note that this is another account of manipulation seemingly incompatible with that initially favoured by Sunstein. And, furthermore, that when it comes to different accounts of manipulation, we may be looking at a fully contested space rather than a shared middle ground.
Sunstein’s solution to the problem of manipulation is to invoke a right against it. Given the wrong entailed by manipulation, and its apparent ubiquity, particularly in online environments, he holds that “there is a strong argument for a right not to be manipulated […] [which] [w]e should see […] as rooted in a principle of decisional autonomy, supported for different reasons by Kantians and welfarists, and operating against both private and public institutions” (p. 52). The right is a means of respecting “people’s capacity for reflective and deliberative choice” (p. 67; see also pp. 63 and 58). Sunstein does not invent this right de novo, as he thinks that in his own jurisdiction something like it is prefigured in US Government policy statements, particularly the Biden administration’s “Time is Money” initiative (p. 60). This seems a slender reed upon which to hang a serviceable right and not just because the pertinent parts of the policy are so general (“Americans should not be subject to confusing, manipulative, or deceptive practices online” (p. 61)). The vagaries of presidential politics also seem to have removed the right’s policy foundation.
Of course, it might be possible to elucidate the content of Sunstein’s right from his own discussion. He insists that it should cover “the most egregious practices, defined as those that are hardest to justify and that are most likely to impose real harm” (p. 56) and considers 14 instances of dark or deceptive patterns (pp. 56–57). But he offers no general conclusion here that could inform the content of the right not to be manipulated, save that when people’s capacity for reflective-cum-deliberative choice (the difference between the two, if there is one, is not unpacked) is undermined, then we are likely in the presence of manipulation. If the manipulation is significantly or really harmful, or likely to be significantly or really harmful, and also hard to justify, then there is perhaps a right against it. Were we to take this as a statement of the content and range of the right not to be manipulated, then it is indeed vague: how could we draft and hence enact that in legislation? The fact that we can raise that question casts some doubt on Sunstein’s claim to offer an account of manipulation that is particularly useful within legal and policy contexts.
Sunstein is clearly more interested in particular instances of manipulation than in delineating the general contours of a right against it. In this, he is no doubt influenced by the piecemeal, case-by-case process that was in part immanent in and subsequently flowed from Samuel Warren and Louis Brandeis’s invention of the common law right to privacy, which he cites approvingly (p. 50). Yet this emphasis on the particular seemingly undermines recourse to a general right: the advantage of the latter is surely that it offers a broad and thus parsimonious response to manipulation. So, for example, a general right against manipulation appears a more direct response to the many instances of deceptive design than piecemeal regulatory and policy efforts that tackle one instance of deceptive design at a time. While we might think that a general right not to be manipulated will short-circuit that process, in the sense that only the courts will be involved in upholding and applying the right, things could nevertheless become messier than we would expect. For the judicial process of delineating the right will surely be dynamic, insofar as the regulatory environment in which the right is applied is itself dynamic. In that case the general right will have to be applied by courts within changed and changing circumstances, which means we could well end up with a situation similar to that occurring in the application of the GDPR in the EU or, perhaps, not unlike that in which the right to privacy has been unpacked in the US.
Parts II and III of the book display the same brio and lightness of touch seen in Part I. Sunstein’s effort to put his account of manipulation to work with regard to deep fakes, sludge and the Barbie problem is illuminating, as are his suggestions for addressing the problems (a property right in one’s person for the first and, among other things, sludge audits for the second). The final part of the book looks to the future, examining how nudges might be non-manipulative and considers how AI might help us make better, non-manipulated choices. It is as illuminating and engaging as the rest of the book.
