Replies to Critics

1. The Goals and Methods of Social Ontology

Aydinonat and Ylikoski offer an insightful “bird’s eye” view of social ontology and its relation with social science. I agree with their claim that the project of UI falls somewhere in the middle between John Searle’s (1995) and Avner Greif’s (2006) projects. I also agree that some of its problems derive from its ambition to provide a general but at the same time explanatory theory of institutions. It is by no means obvious that such a synthesis is possible, and the proof of the pudding is in the eating—we cannot find out unless we try. Others (e.g., Geoffrey Hodgson 2018) have argued that the explanatory project should be kept separate from the conceptual task of defining institutions. This is not necessarily a matter of keeping philosophy separate from social science, but merely of recognizing that taxonomic definitions and explanations are different scientific tasks that we should better not conflate.

UI is built on the premise that keeping them separate is difficult, and on the “naturalistic” assumption that the best insights about social reality can be found in the theories of social science. Aydinonat and Ylikoski are allies in this respect. They are simply less optimistic than I am regarding the prospect of theorizing at the most abstract level about institutions. They start by saying cautiously that “if the type-token relation does not hold, there is much less reason to expect that abstract game-theoretic models can capture the functional essence of any particular institution.” They end up stating that there is no consensus on the cluster of activities that define marriage, and that economists’ functional definition of money is a “theoretically motivated stipulation” rather than the result of extensive empirical research.

Aydinonat and Ylikoski propose to shift the focus away from functions, and toward the social mechanisms that constitute the “generalizable core of social science.” This would have important consequences: it would mean that the study of institutions could deliver at best only generalizations that hold in limited historical contexts and along specific evolutionary paths. It would vindicate a familiar picture of social science (Greif’s, essentially) and of the limited usefulness of functional kinds for scientific purposes (as argued by Millikan 1999, for example).

As fellow naturalists, we tend to think that good ideas are vindicated a posteriori: again, the proof of the pudding is in the eating. I can only say that Aydinonat and Ylikoski’s alternative picture must be taken seriously. It calls for a detailed examination of the role played by functional categories in different areas of social science (anthropology and economics, for example), to determine to what extent they can identify genuine kinds (kinds that support rich inductive inferences, in the Mill-Whewell-Boyd sense). This is a worthy project: I will do my part but I hope that UI will arouse the interest of other philosophers interested in functionalism and social kinds.

2. Rules, Nash, and Correlated Equilibria

The explanatory core of the unified theory is outlined in the first part of UI. Wlodek Rabinowicz (2018) asks some deep and tough questions regarding the game-theoretic notions that I have used, in particular the concept of correlated equilibrium. But first of all, he worries about my general account of institutions as systems of rules in equilibrium. He is concerned that the account may be simultaneously too strong and too weak. It may be too strong if it excludes systems of rules that do not seem to be necessarily in equilibrium—moral rules are his example—and too weak if the theory mistakes a part for the whole: there are other aspects of institutions, like their material base, that do not fit the rules-in-equilibrium account neatly.

Is morality an institution, and does it fit the rules-in-equilibrium account? I confess that I tried to avoid morality as much as possible, while writing UI, for fear of disappointing too many readers. Morality is a controversial topic even for philosophers’ standards, and endorsing a particular view of what it is, how it works, how it is related with other institutional entities would have raised difficult questions that might have distracted the reader from the other contributions of the book. However, many clever philosophers will ask Rabinowicz’s question, or variants of it—so how about morality, or other non-instrumental forms of action? My discussion of normativity gives away some clues. In chapter 6, I notice that institutional rules are often normativized, and for good reasons. Norms stabilize behaviors that would otherwise be erratic, facilitating coordination in situations where predicting others’ behavior is crucial (see, for example, Michael and Pacherie 2015; Sugden (2004 [1986]): if I failed to show up on time for my lectures whenever I had a small reason to delay, my students would be justly annoyed and perhaps would also begin to show up late. A little work-ethics can easily take care of these issues. But even more importantly, norms help solve problems of cooperation—they help us to behave well in those situations where we are structurally inclined to behave in an antisocial manner. They do so by changing our motives—they give us reasons to be good to others, which can be modelled as incentives. How they do it—what kind of reasons they give—is a difficult and fascinating question that moral psychology is supposed to answer.

Turning a mere piece of advice (“do X”) into a norm (“you ought to do X”) may signal a change in the expectations of others—which as many philosophers and psychologists have noticed are very dear to us (for example, Bicchieri 2006; Brennan and Pettit 2004). It may also signal that there are particular reasons (of the non-prudential kind) to follow a particular course of action (Heath 2008). As I explain in UI, I do not think that these analyses are necessarily mutually exclusive: normativity (and morality) are so important for us that they are likely to take different forms, and to be supported by various mechanisms. My strategy in UI is to focus on the way in which the role of norms can be incorporated in the rules-in-equilibrium framework. The key move—which is by no means original—is to model norms as changes in the structure of incentives. This requires shifting between game-representations: a set of actions may be out-of-equilibrium in a game with material (or self-interested) payoffs but may be in equilibrium when normative considerations are incorporated in the game structure. ¹

Rabinowicz points out that while accepting moral norms may be an equilibrium, following the norms may not be one. The distinction between different games gives part of the solution: a set of rules may become an equilibrium after the incentives have changed. But I also think that we should not ask too much of morality: norms are often supported by material incentives (prizes and punishments). Punishments should not be too costly for the punishers or should be recoverable in the medium run. I have discussed some of these issues elsewhere, in relation to the recent surge of empirical studies of reciprocity and altruistic punishment (Guala 2012). Here—to answer Rabinowicz’s question—I would only like to say that morality is not best characterized as a particular kind of institution: it is an important element of many (perhaps most) institutions. No system of rules is likely to work smoothly without a modicum amount of normativity. That is why we have different systems of “business ethics,” “religious ethics,” “civic duties,” “sexual ethics,” and so forth (they are not always consistent, by the way). Humans are compulsive moralizers, and one role of a theory of institutions is to help understand why.

Rabinowicz is also concerned that the rules-in-equilibrium theory may be too weak. It seems unable to capture the fact that institutions are not just rules: there is also a material element in real institutions—such as the buildings that constitutes the University of Milan, the bodies that constitute its professors, and so forth. This remark is in line with a popular trend in recent social ontology, which has led some philosophers to focus on the non-human, material basis of social reality. ² By ignoring the material base, am I guilty of making a pars pro toto mistake?

Let me say at the outset that I do not have any objection to widening the definition of institution so as to encompass its material base. An advantage of the rules-in-equilibrium framework, in fact, is its ability to explicate the role that certain material entities (correlation devices) play in the functioning of institutions. But even in such cases, I think there is a sense in which the material (physical) properties of such devices are secondary, and the rules are primary for the functioning of institutions. Whether the light is green or blue, for example, does not matter much for the functioning of traffic lights. Whether the building of the philosophy department is located in the city center or in the outskirts of Milan does not matter much for my university.

This is a familiar point for those who subscribe to functionalism, multiple realization, and non-reductive materialism in the philosophy of mind and the philosophy of social science. And yet, it marks an important divide in contemporary social ontology. Unlike those who emphasize the material base of social entities, I do not think that explicating the ontology of institutional objects is the main business of social ontology. Social ontology should work in partnership with social science to understand the functioning of institutions in general. And how institutions function in general depends much more on human interests, conflicts, and the rules we devise to accommodate them, than on the material resources we use in the process. Another way to put it is this: most of the causal generalizations that explain how the University of Milan functions—or how universities in general function—hold independently of the material base. When the material base counts (think of a university’s real estate), it counts principally in virtue of its function (e.g., its financial value), rather than its physical properties. This seems to be true for many other institutions as well: monetary theory abstracts away from the material base of money, for example. Whether this hierarchy (rules first) holds universally is an interesting hypothesis, that would require a more extensive investigation. I offer it here as a stimulus for future research. ³

3. Correlation, Mindreading, and Simulation

Although UI is not a technically demanding book, the central game-theoretic concept on which the rules-in-equilibrium theory is built raises some issues of its own. Rabinowicz notices that the range of possible correlated equilibria is actually larger than I suggest, because in my informal examples I overlook probabilistic correlation devices that deliver private advice to the players. This widening of the equilibrium space is bad news on top of bad news, for it makes the problem of equilibrium selection even harder than it already was. As I say in the book, correlated equilibria cannot explain successful coordination. But then “what’s the point of this move to correlated equilibria in the first place?”

The point is to try to explain how new institutions can create possibilities that did not exist before (possible collective behaviors, arrangements, allocations of tasks, and resources). Many collective problems that we face (such as those that can be modelled by the hawk-dove game) seem to offer only relatively bad options, at first sight. But with a little ingenuity, and the help of correlation devices, they can be turned into games that have more attractive equilibria. The notion of correlated equilibrium first of all helps us see how the trick is done, holding onto the basic insight that rules must be adequately incentivized. Moreover, it illustrates how during this process new institutional objects may come to life. Finally, it helps understand what differentiates human societies from the societies of non-human animals. Because they are largely pre-programmed, and because they lack symbolic representation and communication, most non-human animals do not have the opportunity to invent and play new conditional strategies as we routinely do. For all these reasons, I think, correlated equilibria are useful theoretical tools, whether we have a good theory of how equilibrium selection takes place, or not.

Rabinowicz also suggests that correlated equilibria may have two “cognitive” advantages with respect to simple Nash equilibria: because they rely on external coordination devices, which are not controlled by individual players, they may appear more prescriptive than ordinary (unconditional) rules. And because a single correlation device guides the actions of all the players, they may create an appearance of interdependence based on the existence of a common cause. I confess that I had never thought about these features before, but I find them rather appealing. We know independently that humans are not only natural moralizers but also spontaneous naturalists and essentialists. “Unloading” the responsibility of a social arrangement on an external material device may have advantages that are worth studying in more depth.

The cognitive foundations of institutions are discussed briefly in the “Interlude.” These chapters (7 and 8) were written in reaction to a question that I was often asked at conferences and seminars: what about collective intentionality? Is there any room for this concept in the rules-in-equilibrium approach? The Interlude presupposes two ideas: that collective intentionality is an intriguing hypothesis which, however, has proven to be rather impervious to empirical investigation; and that although it may be an interesting phenomenon, it has come to occupy an exaggerated space in the social ontology literature. My way to express these ideas was to illustrate how other mechanisms may play a foundational role in institutional behavior, without excluding collective intentions (and similar concepts) a priori.

Elizabeth Pacherie has recently made important contributions to the project of providing realistic cognitive foundations to joint action (e.g., Pacherie 2011, 2013), so I am particularly glad that she agrees with many of the things that I say in these two chapters. In her commentary, she encourages me to take a more radical stance regarding simulation (or “solution thinking”) and team reasoning: they are, she believes, “little more than notational variants” (Pacherie 2018). ⁴ Part of the reason is that team reasoning has more flexibility than I recognize: through a reconfiguration of the collective payoffs, all sorts of factors—including social norms and conventions—can be “written into” the decision matrix of a team. I think she is right: my account of normativity (chapter 6) licenses a similar flexibility with respect to the payoffs of individual players. Once this step has been taken, it seems inconsistent to be strict on the inclusion of norms in team payoffs. The upshot is that if every coordination game can be turned into Hi-lo, then solution thinking and team reasoning are on a par.

Her second suggestion is to recognize that team concepts and institutional terms play a similar role in the rules-in-equilibrium framework: both are essentially theoretical terms that promote coordination and cooperation in complex strategic interactions. Although they are in principle eliminable, they are useful, and in fact are ubiquitous in the folk psychology of complex societies. I am again grateful for this suggestion, and I see no reason to disagree: like many other institutional entities, teams may be conceived as essentially collections of rules. It follows naturally that team terms play the role of summarizing these rules and offering guidance to team members when they face problems of coordination.

4. Collective Acceptance, Dependence, and Fallibilism

Pekka Mäkelä, Raul Hakli, and S. M. Amadae (2018) offer a number of interesting reflections on collective acceptance theories, focusing especially on the extent to which they are compatible with the naturalistic perspective adopted in UI. A large part of their commentary is devoted to defend the acceptance view from the charge that it implies an antirealist and infallibilist attitude toward folk social theory. But before I get in the details of their arguments, I would like to clarify an important point: although in the book I often identify the collective acceptance program with the seminal work of John Searle (1995) and Raimo Tuomela (2007), I do not take these authors as paradigmatic advocates of antirealism and infallibilism. In fact, both Searle and Tuomela make explicit fallibilist and realist statements in their writings.

Searle, to be sure, is a tricky case: in some paragraphs (e.g., Searle 1995, 33-4), he makes antirealist pronouncements about social facts, while in others, he says things that point in the opposite direction (e.g., Searle 1995, 47). But putting exegetical issues aside, the fair thing to say is that the acceptance framework probably has the resources to support both realism and antirealism. So it is not surprising to find a prominent metaphysician like Amie Thomasson developing an antirealist, infallibilist version of the acceptance view, while other philosophers—like Mäkelä, Hakli, and Amadae—defend a fallibilist and realist one. Since I do not want to get bogged down with arguments about what Searle “really meant,” in UI, I focus mainly on Thomasson’s (2003) and David Ruben’s (1989) work, which incidentally offers a sharper target than Searle’s ambivalent writings. ⁵

I say this not just for the sake of accuracy, but in order to explain what role the critique of infallibilism and antirealism has in the overall strategy of UI. The main goal of chapters 11 and 12 is not to criticize the collective acceptance approach. Infallibilism, antirealism, and more generally antinaturalism are worthy targets on their own—regardless of whether they follow from the collective acceptance view or not. Their rejection in UI therefore is not meant to trigger a modus tollens-type of argument leading to the rejection of the acceptance view. ⁶ As Mäkelä, Hakli, and Amadae explain in their paper, the acceptance view probably has the resources to survive such a challenge, because if properly articulated it may be ultimately compatible with realism.

This does not mean that I endorse the collective acceptance view. As I explain in chapter 8, I do not think that collective attitudes (intentions, etc.) should be placed at the center of social ontology. The reason is that, quite simply, there are many social institutions that do not require collective acceptance in order to exist (racist institutions like apartheid are my favorite examples, but there are others). If we add to this the well-recognized fact that many social phenomena (like racism itself, recessions, and inflation) are non-institutional, we ought to regard collective intentionality and its relatives as certainly interesting, probably useful, but ultimately non-essential building blocks of social reality. ⁷ Although Mäkelä, Hakli, and Amadae do not address this point directly, I suspect that some disagreements among us derive from it. The sufficiency of individual intentional action, and of preference-belief explanations in particular, is a controversial thesis that many philosophers reject. ⁸ While I am not convinced that they do not suffice, I do not exclude other forms of intentionality either. Time (and more research) will eventually tell us what is the case.

Because it provides a useful grammar to represent the construction of institutional objects, the acceptance view has made the ontology of such objects a central topic of investigation in contemporary metaphysics. Mäkelä, Hakli, and Amadae for instance notice that I have little to say on how the imposition of a social function (say, traffic regulation) onto a physical object (say, a light) leads to the creation of a social object, suggesting that it is a limitation of the rules-in-equilibrium theory compared to collective acceptance accounts. This is a fair critique if you think that the constitution of institutional objects is primary in social ontology—but as I said I do not believe that. One of the points that I have tried to make is that institutional objects can acquire their special status without the imposition or collective acceptance of anything. While the “X counts as Y” formula and its variants make a lot of sense for the imposition of names or the introduction of institutional concepts, it is ontologically unnecessary because there are many coordination devices to which we do not assign names. To use my favorite example of the Nuer-Dinka game (Guala 2016, chapters 4-5; Guala and Hindriks 2015), whether two tribes have invented an institutional name or concept for the border that separates their territories, or not, is of secondary importance. That it functions as a border, coordinating the grazing of cattle, is what really matters. The function is fulfilled by changing people’s behavior, a point that Mäkelä, Hakli, and Amadae emphasize, and with which I fully agree. My only concern is to stress that collective acceptance per se does not seem to be necessary, assuming that acceptance is understood in an interesting, robust fashion.

The central and most interesting part of the paper by Mäkelä, Hakli, and Amadae in my view concerns the formulation and the implications of the dependence thesis, within the framework of the acceptance view. The argument in section 3.1 seems perfectly sound to me: Searle’s theory is essentially a theory of tokens, and Mäkelä, Hakli, and Amadae’s proposed version of the dependence thesis for types, with the distinction between K_G and K_g (respectively, the kind-concept for a specific group G, vs. the various kind-concepts of the various groups g), makes room for the possibility that the latter may be sometimes “rudimentary, mistaken, and far removed from the proper kind ‘money’ that social science tries to discover” (Mäkelä et al., 2018). The acceptance view therefore is not necessarily at odds with realism and fallibilism.

Mäkelä, Hakli, and Amadae’s next argument is aimed at showing that causal dependence and ontological dependence are not mutually exclusive relations: X’s being K may be ontologically dependent on our acceptance of X as K, and our acceptance may cause X to be K. Here, we carry the risk of talking past each other: my main reason for distinguishing between causal and ontological dependence is to inquire whether there is some type of dependence that may be a threat for the naturalistic, realist project—given that, apparently, causal dependence does not pose any specific problem. Ontological dependence and its relatives (like constitution, grounding) have spurred a small industry of articles and books over the last decade. In this literature, “ontological dependence” is typically taken to mean non-causal dependence. This is the reason why I focus on it in chapter 11: if we agree that causal dependence is not a problem for realism, the only plausible remaining option for an antirealist like Thomasson or Ruben is to invoke non-causal (ontological) dependence, whatever that may be. In UI, I argue that it does not work either, and Mäkelä, Hakli, and Amadae agree. The only difference is that they take it as good news for the acceptance view, while I am rather indifferent in that respect: as I said, I believe that the collective acceptance view has other limitations, and its relevance for social ontology does not depend crucially on whether it implies antirealism and infallibilism, or not.

Finally, a point about method: I am inclined to take acceptance theorists to be engaged in the sort of “Strawsonian” work that Mäkelä, Hakli, and Amadae describe at the end of section 3 of their paper. The main service of such work is “to vaguely point to social phenomena to be understood and explained in scientific terms,” to use their own words. Although this project has independent interest, it means that some philosophers must scale down their ambitions rather drastically, compared to what they say when they are in their most belligerent spirit. Both Searle and Tuomela, for example, seem at times to conceive social ontology as providing new, better foundations for scientific research, rather than as an underlaborer of science. For example,

I believe that where the social sciences are concerned, social ontology is prior to methodology and theory. It is prior in the sense that unless you have a clear conception of the nature of the phenomena you are investigating, you are unlikely to develop the right methodology and the right theoretical apparatus for conducting the investigation. (Searle 2009, 9)

Much in the same spirit, Tuomela (2002, 7) says that philosophical investigation “is meant in part to critically analyze the presuppositions of current scientific research and . . . to provide a new conceptual system for theory-building.”

Such ambitions would be justified if philosophers had successfully identified a new, non-causal relationship linking mental representations with social entities, or if they had discovered a mechanism for the creation of institutions that social scientists ignore. But they have not, in my view, and for this reason what they say can be largely translated into the language of contemporary social science. To provide a “unified framework” was one of the tasks of UI, and its success will depend, I hope, on whether the unification will turn out to be useful.

5. The Riddle (or Muddle) of Social Kinds

Rico Hauswald is a keener metaphysician than I am, and I am grateful for the suggestions he makes in his commentary. The distinction between tokens and types has already been the target of criticism (see, for example, Epstein 2016, as well as Aydinonat and Ylikoski 2018) but Hauswald adds further reasons to doubt it. As a matter of fact, for the sake of my argument I just need a taxonomy of different “kinds of kinds,” so I would be happy to draw the distinction in terms of historical versus functional kinds, for example. Hauswald’s similarity-based proposal seems to do the job: while functional kinds only require similarity of function, historical kinds require also common descent.

Hauswald (2018) gives an insightful analysis of the ontology of kinds. His first point is that our use of kind terms underdetermines the identification of the boundaries of real kinds. If tigers (by hypothesis) are the only predators that exist (or the only predators known to us), the appearance of a new species of predators calls for an arbitrary decision: we ought to decide whether “tiger” refers to all predators or to a particular subgroup only, and this decision cannot be determined by the way the world is. This is right, but I see the issue of kind identification (and related issues of kind manipulation, reform, change) as a substantial rather than merely a linguistic matter. Therefore, I find the solution that he indicates rather attractive: some kinds are more natural in the sense that they have more explanatory and inductive power. Of course we can recognize the existence of different “kinds of kinds,” and how we use terms like tiger or marriage for referential purposes is not so important, as long as we recognize that the kinds are different. Once this is understood, we can police our language and dismiss bad arguments based on verbal ambiguity. This is what chapters 13 and 14 are meant to achieve.

Hauswald’s argument about the historicity of institution types is particularly insidious for the argumentative strategy of these chapters. My critique of traditionalism presupposes unchanging types, but I agree with him that the evolution of “anthropological fundamentals” is more than a mere possibility—it is very probably happening, and it blurs the boundary between tokens and types of institutions. (Redrawing the boundary in terms of functional and historical kinds would just sanction the existence of this problem, without pointing toward a solution.)

Here is my best shot: at any particular point in time, functions tend to cluster in a specific way. Because the solutions of various coordination problems are related to each other, the space of possible institutions is probably limited. Hauswald is right that “future developments [of the anthropological fundamentals] are basically unpredictable” and therefore “we cannot have a theory that tells us which marriages can exist” in general. This makes us similar to those chemists who were initially building the table of elements with only partial and imperfect grasp of the fundamental laws of particle physics. But given specific background conditions—the anthropological fundamentals that hold here and now—we can use our knowledge of psychology, biology, and social science to make informed scientific guesses about the space of possibilities. It is not true that absolutely anything goes: some institutional arrangements are just not feasible given the way things are at the moment or will be in the short term.

Is this an acceptable solution to “Caligula’s problem?” Perhaps one day human beings and their politics will be so different that a horse will be able to fulfil the functions of a senator. But if we adopt the view from here, we can only envision a possible scenario in which the Senate, for example, retains its ceremonial functions and loses every decisional role. In that case, Caligula’s horse might perhaps perform some of the ceremonial functions. Then we would have to decide whether the difference between the classic functions of senate and the functions of Caligula’s equine senate is wide enough to classify them as different types of institutions, or whether we should consider the equine senate as a particular instantiation of the general functional type senate. Boyd’s theory of homeostatic property clusters does not have the resources to answer this question, as Hauswald points out. Boyd (2000) says that the way in which we draw the boundaries between kinds is determined at least in part by the general theoretical framework (the concepts, categories) that we use, and the sort of inferences that we want to make. While realism is not compromised—the categories do reflect real clustering—Boyd’s concession introduces an element of conventionalism that seems to stand in the way of the solution that I pursue in UI.

Perhaps this is alright, and we should just accept the (partial) conventionality of kinds. Such a conventionality does not make the identification of kinds arbitrary, but merely relative to the background theoretical frameworks that we use, and to the inferences that we intend to make. I find it difficult to figure out how the equine senate would work, and why we would like to encompass it under the general umbrella term “senate” (what sort of inferences this classificatory move would facilitate). In more realistic cases, such as the case of gay marriage, the advantages of including the new contractual relationship under the umbrella of the old institutional term seem obvious enough—the network of functions is thick and the loss of explanatory power once we include the possibility of same-sex marriage appears to be minor. ⁹

Realism licenses appraisals made in terms of “explanatory loss” and “inferential gain.” Although historically contingent clusters of functions may not support “yes-or-no” judgments regarding the identity of institution types, they should still allow us to make comparative judgments regarding their degrees of similarity and inductive potential.

I would like to finish these replies with some meta-philosophical considerations. I was aware, while writing UI, that the book would have to face a diverse audience with different standards of evaluation. So it is not surprising that a dedicated metaphysician would find some aspects of the book underdeveloped or—worse—just sloppy. The main task was to demonstrate by means of example that a naturalistic inquiry can be fruitful in social ontology, providing spectacles that more metaphysically inclined philosophers could use to see a familiar landscape in a new light. That a naturalistic approach is fruitful does not mean that it has no limits, however. I am happy to concede that a question such as “is marriage a type?” is not a matter for social scientific investigation. But looking at how scientists conceptualize marriage—looking at the role that such a concept plays in anthropology, sociology, jurisprudence, and ordinary discourse—does help us distinguish between the different things that we call “marriage.”

Hauswald suggests that my use of intuition as a source of evidence and validation does not fit my professed naturalism neatly. On this point I disagree: intuitions are used in UI as sources of prima facie evidence, or as heuristic tools that point us in the right direction when we theorize. The reason they can do this is that our ordinary concepts and intuitions often reflect genuine distinctions between real kinds, or between real causal mechanisms. Given that folk sociology helps us navigate complex social environments, it would be surprising if our intuitions were systematically wrong. But of course they are sometimes wrong, and for this reason intuitions and folk categories are not the ultimate authority in social ontology. They provide hypotheses or hunches, that ought to be investigated scientifically.