Our interest here is in the fourth of these theories. It strikes us as particularly interesting because it differs significantly from what are perhaps the most familiar notions of ideology. While the most familiar notions of ideology are arguably those which focus on the process of how a belief is generated, Prinz and Rossi (2017, p. 341) “focus not on the process of belief acquisition, but on the formation of the meaning of the relevant concept.” Furthermore, unlike the second type, it does not, strictly speaking, presuppose the existence of any particular distorting form of power, even if power asymmetries “are likely to be part of the picture” (Prinz & Rossi, 2021, p. 121). However, we also think that the full potential of this novel theory of ideology has yet to be realized. As the theory currently stands, or at least, as we understand it, it focuses exclusively on how a narrative is used by certain people in a particular instance. If so, it does not rule out the possibility of addressing the disconnect by controlling or revising the way in which manifest and/or operative concepts are used. As a result, its conclusions are confined to a particular local case. As an example of this, Prinz and Rossi (2021) criticize the use of technocratic and moralistic narratives in debates about sovereign debt during the post-2008 financial crises facing Greece and other European states. They argue that the use of these narratives during this period was ideological because there was a substantial difference between the appearance and the reality of their use. If their argument is sound, it tells us that technocratic and moralistic narratives about sovereign debt were used ideologically in this particular context. It allows for the possibility that the use of these narratives could be plausibly regarded as nonideological in other times and places. Our interest, then, is in the extent to which it is possible to extend Prinz and Rossi's method of ideology critique, such that we can bridge the gap between: (i) P being used ideologically in a particular context and (ii) having reasons to regard P as in some sense pejoratively ideological across different contexts. Our main aim in this article is to show how it is possible to bridge this gap. Our basic idea is this: we can bridge the gap between the fact that a narrative is ideological in a particular context and the claim that it is ideological across different contexts by showing that it is resistant to various mechanisms that might otherwise prevent it from functioning ideologically. The crucial step between (i) and (ii) is thus to find some way of showing that P is resistant to the kind of reform or oversight that would bring manifest and operative concepts more closely into line with each other. Our idea might be best illustrated with a contrast case. Consider the medical profession. Let us grant, as seems plausible enough, that there are at least some medical doctors who are corrupt, incompetent, or in some other sense not what they purport to be. However, we might also think that there are various mechanisms—training, accountability, and so forth—that help ensure that such doctors are not the norm, and that the medical profession is on the whole reliable. To put it another way, the problem lies with bad doctors, rather than the profession itself. On our view, showing that a narrative is pejoratively ideological across different contexts amounts to showing that it is not like the medical profession. It means showing that there are no reliable mechanisms for keeping the narrative in check.2 Our proposed extension of this theory of ideology critique makes three main contributions to existing literature. First, it extends the critical tools of ideology critique by showing how it can be used to anticipate ideologically suspicious narratives before they arise. It thus addresses a crucial realist desideratum, namely, that of providing action guidance to real-world political actors (Galston, 2010; Horton, 2017). Second, we provide a short version of this kind of analysis for sovereign debt narratives. Extending the theory of ideology critique in this fashion would enable political actors in contexts other than that of post-2008 Europe to recognize that technocratic and moralistic narratives about sovereign debt would be ideologically suspicious if they were to be implemented in their own contexts. Third, we show that radical realism's commitment to fidelity to the facts is not inconsistent with certain forms of abstraction and generalization. Our proposed method for showing how a given narrative can be ideologically suspicious across different contexts is rooted in empirical considerations, rather than a priori speculation. Our method may not be confined to a single context, but it is by no means context-free. This essay proceeds as follows. In Section 2, we will examine Prinz and Rossi's discussion of ideology and sovereign debt more closely. We will try to show how their method of ideology critique effectively confines their criticism of sovereign debt narratives to their use in a particular context. In Section 3, we introduce a distinction between the ideological use of a narrative and its being ideologically suspicious. We explore the implications of this distinction and its relevance for broadening the scope of ideology critique. In Section 4, we outline our method for moving from a contextually specific theory of ideology critique to a broader theory that enables us to make claims about the epistemic status of a narrative across different contexts. In Section 5, we suggest some ways for applying this method to sovereign debt. Public debates about sovereign debt come in a great variety of forms, often focusing on the extent to which the state might need to limit public spending in proportion to its revenue or on the status of creditor claims. Here, we set aside the details of such narratives and concentrate on their main framing. Prinz and Rossi (2021, pp. 118–125) criticize two framings of narratives often invoked in these debates: technocratic narratives, and moralist narratives. Technocratic narratives hold that expert knowledge of economics reveals what a sustainable approach to sovereign debt would look like, and that this expert knowledge should be heeded. Moralist narratives hold that the failure to pay sovereign debt is immoral in much the same way that it is immoral for a person to break a promise. Prinz and Rossi argue that both technocratic and moralist narratives were used ideologically during the post-2008 European financial crisis because these narratives were not what they seemed to be. In each case, there was a substantial disconnect between the narrative's appearance and its reality. Our contention is that neither the moral nor the technical claim was epistemically tenable at the level of certainty at which it was uttered … nor was either claim in line with the actual practices of those putting it forward. (2021, p. 120) These disconnects enable Prinz and Rossi to conclude that the narratives were being used ideologically. However, there is a limitation to the scope of this conclusion. Prinz and Rossi's argument supports the claim that moralistic and technocratic narratives about sovereign debt are contextually ideological. That is, these narratives were used ideologically in a particular context (post-2008 Europe). However, this is not quite the same as demonstrating that the narratives are generally ideological, that is, pejoratively ideological across different contexts. Let us refer to claims that a narrative is contextually ideological as CI-claims, and claims that a narrative is generally ideological as GI-claims. To clarify what is at stake in the distinction between CI-claims and GI-claims, suppose it turns out that moralistic and technocratic narratives about sovereign debt are contextually ideological but not generally ideological. One might then acknowledge that they functioned ideologically in post-2008 Europe, but nonetheless insist that these narratives might play non-ideological or perhaps even eminently epistemically valuable roles in other times and places. Perhaps there are other contexts in which we can be confident that the narratives could be invoked without either the epistemic certainty disconnect or the hypocrisy disconnect (or some other disconnect altogether) being present. This possibility could be ruled out only if we have good reasons to think that the narratives are generally ideological. We want to emphasize that this is not necessarily a flaw in Prinz and Rossi's argument. Their argument that moralistic and technocratic narratives about sovereign debt are pejoratively ideological is substantial and important even if it is only advanced as a CI-claim. However, it has two significant limitations. First, a GI-claim is evidently stronger, and in certain respects, more informative, than a CI-claim. It informs us about the status of the narrative in multiple contexts, as opposed to just one. Furthermore, showing that a narrative is generally ideological might tell us something more substantial about why it has been used ideologically in a particular context. Second, and relatedly, a GI-claim provides clearer guidance for how the narrative in question should be approached and scrutinized in the future. This would strengthen the efforts of political theorists to provide practical, action-orienting guidance. Consider how Prinz and Rossi (2021, pp. 126–133) use their claim about the ideological character of moralistic and technocratic narratives about sovereign debt to argue for a further claim: “to counter the danger of ideology, sovereign debt must be politicized all the way down” (2021, p. 116). That is, they recommend that debate about sovereign debt be confined to narratives that do not present themselves as “above the political fray.” Insofar as this is a proposal for depoliticizing debt in all contexts, it is also, in effect, a proposal for rejecting moralist and technocratic narratives about sovereign debt in all future contexts. We think this proposal would receive stronger support from an argument making a GI-claim than one only making a CI-claim. Otherwise, one might argue that both the epistemic certainty disconnect and hypocrisy disconnect are no more than contingent features of moralist and technocratic narratives. Instead of proposing that the narratives are jettisoned altogether, one might instead propose that these two disconnects can be avoided. Why cannot we just recommend that the narratives be presented in a way that acknowledges uncertainty among experts? And why cannot we just recommend that proponents of the narratives employ them consistently? A stronger argument for jettisoning the narratives would require us to answer these questions. Doing so would amount to building a case for the claim that the narratives are generally ideological. Our main interest here, then, is in exploring how we might move from a CI-claim to a GI-claim. Over the next two sections, we will try to explain how this is possible. The first part of our strategy here is to make a distinction between the ideological use of a narrative and a narrative being ideologically suspicious. We have sufficient reasons to believe that a narrative is being used ideologically if we can conclusively demonstrate the existence of a disconnect between its appearance and its reality. In contrast, a narrative can still be ideologically suspicious even if the existence of this disconnect between the appearance and reality of its use cannot be conclusively demonstrated. In such cases, suspicion might arise because we have certain reasons for thinking that it is likely that a disconnect exists, even if we cannot conclusively verify it. It might also be because we have reasons for thinking that a disconnect would evade detection if it exists. If a narrative is ideologically suspicious, we cannot be confident that it is not being used ideologically, even though this is consistent with it not being so used. There are two important points that need to be made in order to make sense of this distinction. The first concerns how it maps onto the contextually ideological/generally ideological distinction. CI-claims typically hold that a narrative is being used ideologically. This is because they are based on observations of various facts about how the narrative has functioned in this particular context. We think Prinz and Rossi give us good reasons to think that moralistic and technocratic narratives about sovereign debt were used ideologically in the context of post-2008 Europe. In contrast, if any GI-claims are plausible, they cannot consist of showing that a narrative is used ideologically. This is because GI-claims are made without access to information about how the narrative is used (or would be used) in all relevant contexts. However, this is not to say that GI-claims are necessarily made in ignorance of how the narrative happens to be operating in certain contexts. The claim that a narrative is contextually ideological in context C1 might be an important premise in an argument for a GI—in fact, such claims do indeed form premises in the type of argument we will sketch in the next section. However, the argument for this GI-claim might not require awareness of how the narrative operates in C2, C3, C4, and so forth. Rather, it is meant to tell us something about how one might suspect that the narrative operates in other contexts, or would operate if it were to be used there. This points to a distinctive contribution of GI-claims: they can tell us something about the epistemic merits of the use of a narrative in a particular context even before it is used in this context. A GI-claim can therefore only be a claim that a narrative is ideologically suspicious. There might also be certain circumstances where a CI-claim is only a claim that a narrative is ideologically suspicious. One possible reason for this might be that our information about the functioning of the narrative in this context is incomplete, but nonetheless adequate for generating reasons for suspicion. However, given that CI-claims draw heavily on facts within the relevant context, we would suggest that they will tend to be claims that a narrative is used ideologically. We can summarize this in Table 2. The second point concerns the relation between the ideological use/ideologically suspicious distinction and the strength of the reason for rejecting the narrative. If X is known to be used ideologically, this provides the strongest reason for rejecting this use of X, since this entails that X has clear epistemic disvalue. In contrast, demonstrating that X is ideologically suspicious provides us with a reason that is not quite as strong, since it allows for the possibility that X is not actually being used ideologically. The fact that we might have good reasons to suspect X of being used ideologically does not necessarily entail that X is, in fact, used ideologically. Now if GI-claims can never be claims that a narrative is used ideologically, we might think this difference in the strength of the reason for rejecting the narrative undermines the rationale for looking for a way of making plausible GI-claims in the first place. In the previous section, we noted that one of the limitations of CI-claims is that they might not provide a particularly strong reason for rejecting the use of the narrative in contexts other than the one in which it has been found to be used ideologically. But, if GI claims cannot provide a particularly strong reason either, what value do they have? Our answer is that plausible GI-claims—that is, GI claims demonstrating that a particular narrative is ideologically suspicious—give us a stronger reason for rejecting the narrative across different contexts than any one CI claim. A CI claim gives us a very strong reason for rejecting the use of a narrative in a particular context, but a relatively weak reason for rejecting the use of the same narrative in other contexts. The latter reason would be no more than a weak form of induction: the fact that the narrative was used ideologically in one context gives us a reason to think that, if it is used elsewhere, it will also be used ideologically there. But as we saw in the previous section, this is because it might turn out that the features that cause a narrative to be used ideologically in a particular context do not obtain elsewhere. We can summarize this using Table 3. The top-right box represents attempts to inductively extend CI-claims to the use of the narrative in other contexts. The reason why we think this provides a weaker reason for rejecting the narrative in those contexts than a GI-claim is this: to the extent that an argument for a GI-claim can be rendered plausible, it is less subject to the contingent factors of particular contexts than an inductive argument stemming from a single CI-claim. The strength of the latter can be diminished by differences between C1 and other contexts, while the strength of the former cannot. Note, however, that this points to one of the challenges in making plausible GI-claims. It must be possible to show why a GI-claim is not vulnerable to differences in context. At this point, it is worth considering two related objections, each of which is based on the view that the purposes of ideology critique are adequately serviced by making a multitude of CI-claims, without the need for a GI-claim. First, one might argue that if one wants to know about the status of the narrative in another context, one can simply examine that context. For example, if one wants to know about the status of sovereign debt narratives in, say, the aftermath of the 1970s oil crisis, one can simply study that context instead. Second, one might also argue that the strength of the reason in the top right box of Figure 3 can be increased if we can make multiple CI-claims about the use of the narrative, to a point where its strength might eventually exceed that of the GI-claim. The problem with the first objection is that it renders ideology critique less capable of action guidance. This is because it is confined to the study of past contexts. It does not allow for claims to be made about the status of narratives that might be used again in the future. The second objection, meanwhile, has two problems. First, since there are usually limits to our access to information about different contexts, we cannot always rely on the availability of a multitude of CI-claims in order to strengthen an inductive argument. Second, even if a multitude of CI-claims were available, the GI-claim would still enjoy a distinct advantage over the inductive argument: it can tell us something more specific about the underlying reasons why a narrative has shown a tendency to be used ideologically. With this in mind, our initial aim of extending Prinz and Rossi's theory of ideology critique beyond particular contexts can be reframed as an attempt to move from a CI-claim that a narrative is used ideologically to a GI-claim that the same narrative is ideologically suspicious. In the next section, we will try to show how this is possible. Let us start by assuming that we have a plausible CI-claim: that X is used ideologically in C1 in virtue of there being one or more disconnects between the reality and appearance of X in C1. Our aim here is to show how we can move from this claim to a plausible GI-claim: that X is ideologically suspicious across different contexts. Our strategy for doing this comes in two stages. First, we need to identify possible “deideologizing” mechanisms—that is, mechanisms that might be capable of overcoming these disconnects. Second, we need to show that these mechanisms are unlikely to succeed. If each of these two tasks can be accomplished, then we have shown that the possibility of differences in context does not prevent X from being regarded as ideologically suspicious. It would entail that any possible differences in context are not capable of generating effective mechanisms for overcoming these disconnects. Our aim in this section is to help clarify this idea of a mechanism. We can divide mechanisms into two broad categories. The first category, which we will refer to as safeguards, contains what are generally the most likely candidates for overcoming disconnects. Safeguards are mechanisms for controlling the reality of the narrative in such a way that it does not stray from its appearance. The second, less common category, can be described as narrative reform. It seeks to revise the appearance of the narrative in order to bring it into line with reality. If an effective safeguard could be applied to moralistic narratives about sovereign debt, it would generally prevent these narratives from being invoked hypocritically. If an effective form of narrative reform could be applied to technocratic narratives, it would reconceptualize the claim that a particular policy is conclusively vindicated by expert analysis in such a way that renders it somehow compatible with the absence of any expert consensus. We will first discuss two general problems when it comes to finding effective mechanisms. Following this, we will say a little more about safeguards and narrative reform respectively, and some of the more specific problems they may face. The first and most serious problem for finding effective mechanisms is that they must take people's real motivations and interests into account.3 An effective mechanism cannot propose a course of action for people who are entirely unmotivated to act in this way. In particular, a mechanism cannot be reducible to a proposal for different behavior on the part of those who use narratives. Consider, for example, the suggestion that the epistemic certainty disconnect can be overcome simply by having politicians and public actors who present the narratives in a way that does not create a misleading impression of a consensus among experts. Similarly, one might think that the hypocrisy disconnect can be overcome by having politicians who consistently follow principles, rather than doing so only when they find it expedient. Our view is that neither proposal is sufficiently serious to be counted as a mechanism worthy of consideration in the first place. One might as well propose to protect the security of banks by asking criminals (or owners) not to rob them. Note that this problem does not suppose that people who use narratives ideologically are somehow consciously intent on doing so. It may well turn out that if people derive some political or material benefit from using narratives ideologically, they may end up using them ideologically even if they do not explicitly wish to do so. Their desire to portray themselves in a good light may shield them from realizing what they are doing and saying (Geuss, 2017, p. 240). An effective mechanism must therefore not be reliant on the goodwill or self-awareness of those who use narratives. A second problem is that making sense of the effectiveness of a mechanism requires us to provide an error theory explaining why the CI-claim was plausible in the first place. If there is an effective mechanism for preventing moralistic and technocratic narratives about sovereign debt from functioning ideologically, why was this mechanism not effective in the context of post-2008 Europe? The upshot of this problem seems to be that a mechanism cannot be as straightforward as something like everyday common sense. This is not to say that everyday common sense cannot play a role in an effective mechanism, but that it is not sufficient. We will now turn to two mechanisms (safeguards and narrative reform) which, if successfully deployed, would stand a more credible chance to avoid ideological distortions. An effective safeguard will serve as a disincentive for ideological uses of narratives. It may do this by making ideological uses appear as lacking credibility. It may also do this by boosting the credibility of non-ideological uses, since this allows them to be more clearly distinguished from ideological uses. Here it should be noted that it is an important feature of ideological uses of narratives that they are not perceived—at least initially—as completely lacking credibility.4 Otherwise, people recognize the narrative for what it really is and there is no actual disconnect between its appearance and reality. If we claim that New York is located in the middle of the Sahara Desert, it would be a little odd to say we are peddling an ideology. We can identify two broad categories of safeguards. The first is “accessible success metrics.” If a narrative involves the assertion of claims that are straightforwardly vindicated or falsified, then this can prevent it from being used ideologically. For example, it might be argued that the reliability of medical doctors can be somewhat assessed by the extent to which the patient responds to the prescribed treatment. This arguably plays some role in preventing their profession from becoming something that might be called ideological (i.e., they do not actually do any work that improves people's health, in spite of their claims to do otherwise). The use of assessable success metrics as mechanisms faces two main difficulties. First, it is not always clear what counts as success. Even in the medical profession, it might be argued that a certain prescribed treatment may have “succeeded” in giving the patient a fighting chance at recovering from an illness, even if the patient did not ultimately recover. Differences in ideas of success may be even more pronounced when it comes to partisan politics—one person may look at the results of a particular policy and pronounce it a resounding success, while another may look at the same results and pronounce it a failure. Second, if we bear in mind the distinction between correlation and causation, it is not always clear what the results of a particular course of action actually are. We might suppose that the fact that the patient recovered well demonstrates that the prescribed treatment succeeded, but it might also be the case that the treatment had no meaningful effect and the patient recovered regardless. One might think these considerations make accessible success metrics largely unavailable to medical expertise, and perhaps all other forms of expertise-based narratives more generally. If so, we might think that all such narratives are ideologically suspicious. Our view is that the general lack of success metrics does indeed generate an element of ideological suspiciousness for all expertise narratives. However, this suspiciousness can be understood in scalar terms, and may be kept sufficiently low in order to justify not describing them as ideologically suspicious in a binary sense if other effective safeguards can be found. A second type of safeguard, perhaps more amenable to making clearer scalar judgments about the merits of different expertise narratives, can be called shorthands. These include certain red flags, which increase our suspicions that a narrative is being used ideologically, or signs of credibility that diminish our suspicions. Perhaps the most obvious example of the former is the lack of transparency about methods or reasoning, while full transparency might be an obvious example of the latter. A medical doctor's credibility might be enhanced by making efforts to explain her diagnosis and/or prescriptions to her patients, while a doctor who hands pills to her patient without explanation might be regarded as suspicious. In other cases where a narrative involves some appeal to expertise, it might be possible for nonexperts to use additional shorthands to evaluate the reasoning of experts (Goldman, 2001). There might be more shorthands that extend beyond transparent reasoning. For example, a doctor who does not mention the cost of potentially expensive medication might be raising a red flag, while a doctor who takes the time to ask the patient for a thorough medical history might be considered more reliable as a result. The difficulties facing shorthands seem to fall into two main categories. The first concerns the limitations of people's capacity to implement them. Consider, for example, the suggestion that people can assess the quality of their doctor's instructions by carefully studying the latest medical research in online academic journals. The second concerns people's willingness. Perhaps it is not overly difficult for people to scrutinize claims that a particular medical position enjoys universal expert support by doing a quick search for dissenting expert views. But even if this is so, this can function as an effective safeguard only if people are actually willing to do this research. It may be that these difficulties are not usually substantial impediments to the role of shorthands in giving credibility to certain expertise narratives, including that of medical expertise. However, it is important to note that the ability of people to implement shorthands regarding expertise narratives may often be dependent, not only on their own cognitive skills, but also on the willingness of experts and other political elites to make information available to them. This lack of willingness on the part of experts and elites may severely diminish the effectiveness of shorthands, as we will shortly see in Section 5. Narrative reform is meant to address disconnects between the appearance and reality of a narrative by adjusting its appearance. If the disconnect relates to concepts that are part of the narrative, then this would require altering how people understand the concepts in question, in such a way that better fits with how the concept functions in reality. If the disconnect relates to the claims presupposed by the narrative, this would mean altering the content of these claims so that they are no longer counterfac