About the Unreal

The erudite and petulant Ignatius Riley [22] from Confederacy of Dunces and the elegant Scarlett O’Hara from Gone with the Wind [23] are fictional characters. Mythological narratives too often contain reference to fictional characters, such as Homer’s depiction of Achilles and Patroclus [24], or fictional locations, such as Midgard in the Eddas [25]. Fiction understood here may also concern real characters, as in the case of War and Peace, which is set in Moscow. Or it may treat of hypotheticals such as “What if Nelson had lost to Napoleon at the Battle of Trafalgar”, or of counterfactuals such as “If Oswald hadn’t killed Kennedy, someone else would have” [26]. There is also speculative technology, such as Ray Bradbury’s matryoshka brains [27] or H.G. Wells’ anti-gravity cavorite that enabled faster-than-light travel [28]. In each example, we see putative reference to some manner of fictional entity that is assumed to have never existed.

Given this wide range of fictions and of fictional entities, it will be helpful to identify a paradigmatic example on which to focus our attention. Superman, for those unfamiliar, is a fictional character represented in various media. When one hears putatively accurate assertions such as “Superman is strong” there is an implied context in which the expression is to be understood, not to be confused with the real world which we inhabit. Superman is strong in representations within, say, comic issues published by DC Comics. And it is indeed true in the actual world that a Superman character is described in these publications as being strong. To evaluate the claim that “Superman is strong” one must relativize the expression to some seemingly fictional context for evaluation.

Many digital twins represent physical assets, though some simply serve as prototypes that prescribe how to create a future physical asset. So-called “digital twin instances” mimic physical products to which a representing digital twin remains linked throughout the life of the product, while “digital twin prototypes” specify what must be done to produce a physical product meeting the specifications of the prototype [29]. Similarly, blueprints representing buildings, pharmaceutical chemicals, or vehicles, provide guidance for how certain products can be created, and this is so whether or not they are eventually created. Such guidance need not be focused exclusively on the creation of products, however. Something similar applies also to training protocols, education plans, legal codes, congressional bills, and so on. For example, the Tax Cuts and Jobs Act of 2017 prescribed a lowering of taxes on corporations and individuals in the US [30] which did not take effect until 1937. Failed legislation seems to sharpen the issue further, as one might plausibly ask what the 1926 Child Labor Amendment to the US Constitution is about, given that amendments require 38 states to ratify, and as of 1937 only 28 states had done so [31]. Throughout these examples there is a prescription in play reflecting how someone or some group desires the world to be, whether that involves a change to the way the world is or maintenance regarding how they want the world to continue being.¹

Our paradigmatic case will center on Samantha, who works for the car manufacturer Honda and is tasked with drafting and promoting specifications for new vehicle models for some coming year. When Samantha creates a blueprint detailing a new Honda Civic model to be produced in 2035, there is an implied prescribed context against the background of which her work should be understood. Samantha is not, we may assume, describing anything that presently exists, but is instead outlining a blueprint for some entity intended to satisfy certain requirements once manufactured; she is prescribing how she intends the world to be. Expressions Samantha makes about the intended output, such as “The Honda Civic SLS 2035 will prove itself to be my best work” are thus evaluated against this special context.

Simulation involves the use of physical, mathematical, behavioral, or logical representations of some system in the interest of simulating how that system may behave under various circumstances [33]. Such representations often concern systems that may or may not exist, such as historical systems, where our knowledge turns on events incompletely recorded, or biological processes which occur at scales too small for direct observation. In other cases, the simulated systems may be entirely theoretical, such as a system devised to study the formation of galaxies under alternative physical constants [34]. Engineers devising simulations for, say, red teaming exercises [35] often do not want the simulated events to occur. Indeed, they are often preparing strategies for preventing the simulated events from coming to pass as well as planning for what steps to take in the event they do. In either case, evaluations of expressions borne out of simulation exercises, such as “Respond with nuclear capabilities” must be evaluated against some special unreal context, as in the case of fiction and blueprints.

As a paradigmatic example, consider a team of cybersecurity experts engaged in red-team exercises regarding potential vulnerabilities in a network. Discourse concerning simulations of this network are to be interpreted within the special context in which certain threats and defense postures may or may not manifest.

There are of course many other similarities and differences across these cases. For example, our everyday cognitive attitudes and conversations involve the uses of expressions which involve engagement of imagination. From another direction, fiction neither describes how the actual world is nor prescribes how one wants the world to be. In contrast, blueprints are often prescriptions for how one wants the world to be, while simulations are often descriptions about how the world could be. Simulations do often

involve prescriptions also, as when they are used to inform decisions about what would have to be done should certain conditions obtain.

Expressions about fictional entities are not directed to the future, but blueprints and simulations are. Blueprints are often about desired or desirable outcomes, whereas simulations are often about undesirable outcomes and subsequent planning in the event such outcomes are realized. Simulations are thus characteristically associated with probabilities or likelihoods. While these differences are important, it remains the case that expressions found in all the mentioned cases are connected by relativization to some special context.²

Another theme worth highlighting is that expressions stemming from each case are often interpreted in natural language as referencing some instance or some individual. For example, the expression “Superman wears a red cape” suggests reference to an individual called “Superman.” There are some who talk as if there were some entity denoted by “Superman”, for example they hold that the referent of this expression is a concept, or some other mental particular [36, 37]. But it is not a concept that wears, or is held to wear, a red cape. There is however an aspect of such expressions which does indeed involve reference to entities that exist in the real world, namely to entities existing at the level of types rather than of instances. Expressions such as “Superman wears a red cape” are about familiar, actual entity types such as Wearing³, and Redness, and Cape, where an actual entity type is a type which is or has been instantiated. And there are numerous examples of all the mentioned types in this the actual world.

These remarks apply just as well to implicit reference in fictional sentences, such as “Superman is from the planet Krypton”, where there is reference to types such as Planet, Rocky, Watery, and so on. There is however no reference to any entity denoted by the term “Krypton”, because there is no such entity. Similarly, simulations about red teaming exercises involve references to types such as Bot Network that have actual instances; an instance of Blueprint for a planned Honda Civic SLS 2035 may refer to Steering Wheel and Portion of Metal, instances of which, again, exist.

We can thus accept that each of our examples (comics, blueprints, and so forth) are about something even where they involve no reference to specific instances in the actual world. Indeed, it would be quite odd, we claim, to assert that Samantha creates a blueprint for a Honda Civic SLS 2035 instance, prior to any such vehicle being manufactured. Such a commitment opens the door to questions such as “To which instance does the blueprint then refer?”

There is another sense in which the actual world harbors much more than we can faithfully represent than is achievable through fiction, blueprints, or simulations. For in each case the information artifacts in question exhibit loci of indeterminacy [16]. There is no answer to the question “What was Superman’s grandmother’s eye color?” because this topic was never addressed in the Superman narrative. If Superman existed, however, this question would have an answer, and this regardless of whether it is or is not known by anyone. As diligent as Samantha is in construction of a blueprint for her new model, there will be dimensions of variability between what is specified in her plan and any actual vehicles produced. Red- and blue-teaming simulations, similarly, cannot hope to

characterize all unknown unknowns. The contexts in relation to which they are defined are necessarily incompletely specified.

To illustrate our proposal, we describe a recursive recipe for decomposing Information Content Entities that are not obviously about anything that exist into logical combinations of actual classes. Many of the ingredients have been described already, but we introduce the remaining ingredients here.

CCO provides three sub-properties of is about designed to reflect different attitudes agents may bear in regard to the relations holding between Information Content Entities and Entities which they are about. These object properties provide a foundation on which to represent paradigmatic examples of our cases, without requiring reference to problematic instances. These sub-properties⁴ of is about are:

The information content of a newspaper article describes some current event, just as an accident report describes some accident. A blueprint serves as a model for some product, just as a professional code of conduct serves as a set of rules for anyone acting in the corresponding professional role. The content of a photograph represents the photographed entity, just as the content of a transcript represents the verbal interaction transcribed. The sense of “isomorphism” in the definition of represents is relative to the type of entities involved. For example, the arrangement of Napoleon’s body parts in a painting by Jacques Louis David was meant to reflect the actual arrangement of these parts in Napoleon’s body.

With respect to our cases, we maintain that fictional Information Content Entities are best understood as describing some logical arrangement of classes and object properties, ultimately in terms of actual classes. Blueprints, on the other hand, are best understood as Information Content Entities prescribing some arrangement of classes and object properties; and simulations are best understood as representing, insofar as, were the simulated phenomena to exist, so too would a relevant isomorphism.

With these ingredients in hand, we now turn to our recipe. To provide an ontological representation of an entity that did not, does not and perhaps will never exist:

Regarding 3: In OWL2DL, there are only three viable options for relating the relevant Information Content Entity to what it is about, by asserting: universal constraints (all x is about only y), existential constraints (all x is about some y), or a direct instance-to-instance (specific x is about specific y) relation. We can put aside the last, since it would require introducing a “dummy” instance which the relevant Information Content Entity would be about. We can also put aside the existential constraint, which amounts to asserting the relevant Information Content Entity is about some, though no particular, instance. This again runs counter to the intuition that in fiction, blueprints, and simulation relevant Information Content Entities are not necessarily about any given instance, since there often is no such instance. This leaves us with the universal constraint, which amounts to asserting that the relevant Information Content Entity is about only an instance falling within a class.

Importantly, this option neither asserts nor implies that there is such an instance. It requires only that, if there were such an instance, then it would have to be an instance of a class resulting from the logical combination of actual classes and object properties. It is this combination of classes that a relevant instance of Information Content Entity will be about. It is, moreover, only this combination that expressions putatively about such an instance will in fact be about.⁵ One might, at this point, balk, given that we seem to have simply replaced “dummy instances” with “dummy classes”. Not quite. What we have done is avoid introducing expressions putatively referring to instances but which in

fact do not denote anything, by appealing to logical combinations of classes that themselves consist of instances that exist in the real world.

In every case, classes and object properties must be unpacked into actual classes and object properties that have only actual classes as relata.⁶ Object properties are, in this recipe, ultimately explicated in terms of the classes which are their domains and ranges. We now apply this recipe for each of our paradigmatic examples.

Because we are dealing with fiction, we leverage the describes relation in our formalization. For simplicity, we introduce a subclass of Information Content Entity (ICE) called Fictional Description, as well as various classes and object properties where needed, at least one of which has no obvious parallel in the actual world. We then have, assuming a standard list of features associated with the character Superman:

At the completion of our recursive application of the recipe, we end with “Superman” being defined ultimately in terms of classes and object properties that have actual instances. For each string putatively referring to some non-actual entity, such as “Krypton”, or to some non-actual relation, such as “fires eye lasers”, we decompose to classes all of which have instances, such as Rocky Quality or Laser.

Even so, one might wonder what to do if, rather than fires eye lasers, the pertinent object property would be fires laser having a class Eye Laser as its range, a class which has no actual instances. In this case, we would simply leverage the classes Eye and Laser to define the class Eye Laser, which would of course have no instances but would be decomposable into actual classes. This approach can be applied equally to object properties such as fires eye laser with range Eye Laser.

It has been suggested in [37] that such a recipe falters when fictional characters are represented as engaging with entities that violate constraints of imported ontologies. Following the example of [39], Superman might encounter a ghost which – one might argue – should be classified as a Person having material parts and yet also as an Immaterial Entity having no material parts. In BFO, the class Material Entity is disjoint with the class Immaterial Entity, so such a classification would result on our approach in an inconsistency. As a flat-footed response, note that while authors

might label ghosts as persons, that does not mean ontologists should follow this labeling. Ontology engineers must not let themselves be tricked by mere labels. This is to say that, if there are examples of ghosts classified as persons, then it is plausible that the authors do not mean by “person” an entity with material parts. For such cases, we might introduce a class expression such as “ghost person”, situated outside the material entity hierarchy.

This will only take us so far, of course. Authors of fiction need not abide by ontology best practices. We might envision an author who creates a fictional character that is explicitly a ghost that has no material parts and yet also, simultaneously, has material parts. This is a logical contradiction. Even then, however, there need be no problem for our approach. The relevant hierarchy could be extended to accommodate, perhaps with a subclass of Continuant each instance of which has continuant part some Material Entity and has continuant part some Immaterial Entity. Our hypothesized ghost description would then be about only instances of the Continuant subclass so defined, though this class would remain empty. For such a case, the above recipe will not terminate, which is also as it should be. There are descriptions of logical inconsistencies in the world, but they are not about any instances in the world.

In explicating the blueprint case, we leverage the prescribes relation in our formalization and introduce a subclass of Information Content Entity called Blueprint.

Moreover, any reference within Blueprints to other Blueprints can be characterized by following our decomposition recipe as illustrated with fiction above.

There is, however, a remaining question over the relationship such a pattern bears to prescribed entities once they exist. Applying the proposal for digital twins as defined here [42], we say that the Blueprint initially only prescribes that instances be created, though as yet no specific instance of the relevant class exists. Once an instance is created on the basis of the Blueprint, we say that this Blueprint still prescribes new instances to be created but also represents the instance that has already been created. The axioms governing represents in CCO would then entail that the Blueprint in question is a special type of Information Content Entity, namely, a Representational Information Content Entity. This class is not disjoint from its sibling – the domain of prescribes – which is Directive Information Content Entity. In other words, CCO allows for Information Content Entities to both prescribe and represent simultaneously.

In explicating simulation, we leverage the represents relation in our formalization, as well as a new type of Information Content Entity called Cyber War Game Simulation Representation.

One might worry that such hypothetical scenarios could represent contradictory events or impossible strategic responses—for instance, deploying centralized and decentralized defensive tactics simultaneously. Such hypothetical contradictions present no inherent difficulty provided that our representation refrains from enforcing unnecessary mutual exclusivity among these scenario elements, following suggestions similar to those discussed above with respect to fiction.