The rapidly growing literature on artifacts revolved mostly around finding non-trivial artifact essences, while dissenting voices pointed out the plurality of artifact kinds and raised legitimate concerns about the applicability of any essence for artifacts and artifact kinds. I call the first endeavor artifact essentialism and the latter artifact anti-essentialism. Both essentialists and anti-essentialists, implicitly or explicitly, share the same assumption: that there is only one legitimate artifact concept that we can profitably use. I call this view artifact concept monism. I argue that the current state of artifact essentialism cannot provide an extensionally adequate and definitionally coherent overarching concept. The extensional and definitional problems I point out led some anti-essentialists to give up on classificatory aims and others to doubt the primacy of metaphysics on the topic of artifacts. In this paper, I aim to offer an alternative to artifact concept monism. I call my view artifact concept pluralism. I argue that artifact concept pluralism provides a better framework to deal with the problems artifact essentialism faces. Furthermore, it enables us to bring metaphysical and epistemic considerations together without giving up on the classificatory aims and requiring a significant revision in our taxonomical practices.

That said, this paper’s main methodological leaning is clear: practices come first. According to David Davies (2004), an ontologist of art should not put forward metaphysical principles before examining the practices closely; art practices impose a “pragmatic constraint” on metaphysical accounts. As Davies (2004, 18) describes this pragmatic constraint, “Artworks must be entities that can bear the sorts of properties rightly ascribed to what are termed ‘works’ in our reflective and critical and appreciative practice…” Similarly, in this paper, I assume that artifact practices impose a “pragmatic constraint” on the metaphysics of artifacts. This does not mean that artifact practices are the final arbiter of our best metaphysical account; rather, our rational reconstruction of the output of the relevant practices determines our metaphysical accounts. However, as artifact practices are (even) less uniform than art practices and given the problems current monistic accounts face, I argue that a responsible form of pluralism is needed to account for artifact practices.

Following Kathrin Koslicki (2008, 201), I take kinds as “taxonomic classifications under which particular objects may be grouped based on shared characteristics of some sort.” Accordingly, an artifact concept is what singles out the relevant characteristics required for artifact-kind membership. Artifact concept monism assumes that there can only be one way of grouping entities under artifact kinds, and thus, it assumes that there is an overarching artifact concept. Artifact concept pluralism rejects this assumption. I construct a model of artifact concept pluralism following Christy Mag Uidhir and P.D. Magnus’s proposal on the art concept pluralism. According to Mag Uidhir and Magnus (2011, 91–92), there are at least four art concepts; in other words, there are four ways of grouping art objects, and each way of grouping has its own strengths and weaknesses. Mag Uidhir and Magnus (2011) draw their art concept pluralism on the model of species pluralism. According to species pluralism, there are several ways of grouping organisms into species. Both models guide this project of artifact concept pluralism. Drawing on these models and taking the output of relevant practices seriously, artifact concept pluralism proposes that there are multiple correct ways of grouping entities into artifact kinds.

1 Artifact Essentialism

John Locke famously distinguished the real essence of things from their nominal essences (Locke, Essay, Book III, chap. III, §15, cf. Reydon 2014, 127). The former is generally construed as the mind-independent nature of things, whereas the latter depends on how the relevant minds conceive of entities (Reydon 2014, 127). Although Locke was pessimistic about finding real essences of things, in the case of natural kinds, those authors who prefer the semantics put forward by Kripke (1980) and Putnam (1975b) seek out kinds whose nature is constituted by mind-independent essences (Thomasson 2007, 54). For instance, in the case of a natural-kind term like gold, all gold atoms share the same atomic structure, and this structure is discoverable by the relevant scientific practices. This mind-independent essence of gold, in turn, fixes our reference to the term ‘gold’ and enables us to distinguish genuine gold from fool’s gold (Reydon 2014, 127).

Some suggest that a similar strategy applies to artifact-kind terms and claim that functions can serve a reference-fixing role for artifact-kind terms (Putnam 1975b; Kornblith 1980). Others argue further that some artifact kinds have a mind-independent nature akin to natural kinds (Elder 2007; Franssen and Kroes 2014). However, it is not at all clear that the traditional distinction between mind-dependent and mind-independent essences and its bearing on reality is uncontroversial (Reydon 2014, 130). Not all natural kinds neatly follow this distinction. For instance, it is now commonly taken for granted that biology failed to provide genetic essences unique to species simply because species are found to be subjected to constant evolutionary change (Reydon 2014, 131). A new form of essentialism is on the rise in the philosophy of science (Boyd 1999; cf. Reydon 2014).

According to the new essentialism, the essences need not be non-relational properties. The paradigmatic cases are biological kinds. Historical and relational properties are now considered part of biological kinds’ essences (Reydon 2014, 130–131). The new form of essentialism is also suitable to accommodate artifact kinds. After all, possible candidates for artifact essences refer to how artifacts are being used, why they are reproduced, etc. Having briefly elucidated both forms of essentialism, I formulate essentialism about artifact kinds broadly as follows:

I will consider Artifact Essentialism as a condition about kind essences as opposed to individual essences. There are at least two distinct construals of individual essences. First, it might mean a particular instance \(i\) of the kind \(\mathrm{K}\) essentially belongs to \(\mathrm{K}\). On this understanding, \(i\) cannot exist without being a \(\mathrm{K}\) (Bird and Tobin 2022). According to the other construal, individual entities might have essential properties besides the essential properties shared with the other instances of the kinds they belong to. For instance, if we agree with Kripke (1980) on origin essentialism, then being a child of my parents is an essential property of mine, while it is not an essential property of the kind human. Having made the distinction between individual and kind essences, we can state that throughout this paper, a kind essence \(\mathrm{E}\) indicates a non-trivial essential property or a set of properties that are shared by the members of an artifact kind. For instance, if artifact essentialism is best understood in terms of functions, one would expect individual chairs to have the function of seating a single individual, and by this functional property, one could assess whether a given chair is a proper chair, a malfunctioning chair, or a non-chair (e.g., a chair beyond repair).

To make my discussion more exhaustive, I take artifact essentialism to be neutral on traditional and new forms of essentialism. The most commonly discussed artifact kind essences (\(\mathrm{E}\)) are the following (Grandy 2007; Vega-Encabo and Lawler 2014; Koslicki 2018): i) functions, ii) maker’s intentions. I will not provide a detailed explanation of any individual account. Having provided the general essentialist outline, I raise two problems against artifact essentialism, namely, the extensional problem and the definitional complexity problem. Both problems are raised by Mag Uidhir and Magnus (2011) in their attack against art concept monism. I follow a similar argument.

1.1 Function Essentialism

A quick survey both of the literature and pre-theoretical intuitions shows that functions are the most favored artifact essences.¹ Even many familiar artifacts around us are named after their functions (Baker 2008). To list a few: screwdriver, corkscrew, pencil sharpener. Kornblith (1980, 112) writes, “At least, for the most part, it seems that what makes two artifacts members of the same kind is that they perform the same function.” Kornblith’s statement provides us with the basic intuition behind function essentialism.

According to Tim Juvshik, function essentialism favors function as the best candidate for artifact essences. To elaborate by way of an example, a triangle screwdriver and a magnetic screwdriver have distinct designs and perform their functions differently. The former’s design is more safety-oriented, whereas the latter, with the help of magnetic force, performs a better job with smaller screws. Yet they both drive screws. Given the significant multiplicity of form and design, according to function essentialists, functions provide a prima facie suitable artifact essence that can bind various artifacts under a single artifact kind (Preston 2013).

However intuitive the functional characterization of artifacts and artifact kinds is, there is no consensus on how to characterize functions. The first attempt to characterize functions may be taking functions as answers to “what is it there for” questions, which in turn explains “how the thing got there” (Wright 1973, 146–156; Vega-Encabo and Lawler 2014; Juvshik 2021b). For instance, I can use a towel as a cover for my favorite snacks, yet a towel is for drying hands, just as the heart is there for pumping blood, not for producing a unique sound. Wright (1973) calls the former function of my towel function as and the latter the function.² The main difference between these two senses of functions is that the latter has the explanatory force that accounts for the historically successful reproduction of, say, towels, which the former lacks.

Wright’s distinction is more or less retained in the subsequent theories of function. Benefiting from the literature on functions, philosophers recently put forward elaborate theories on artifacts. The attempts can be largely divided into two camps: etiological functionalism and intended functionalism. Emphasizing the etiological aspect of functions while eschewing the intentional properties, Elder, one of the champions of etiological functionalism, suggests that many artifact kinds share a similar nature with natural kinds; these kinds essentially instantiate a cluster of properties that are copied among the members (Elder 2007, 37). The cluster of properties for artifact kinds includes three main elements: particular shape, proper function, and historical placement (Elder 2007, 43). The kinds of objects that satisfy all these elements, in Elder’s view, are copied kinds. Copied kinds include both natural and artifact kinds without having any ontologically significant difference between them.³

However, etiological functionalism leaves us with conclusions that are at odds with our ordinary linguistic practices (Thomasson 2007; Juvshik 2021b). In Elder’s view, for instance, a familiar artifact kind such as corkscrew turns out not to be a copied kind since its nature is not specific enough because the shape shows high variations among corkscrews. Thus, this view admits only specifiable artifact kinds like winged corkscrew, which has a certain shape (e.g., winged), proper function (e.g., to remove corks), and historically proper placement (e.g., H.S. Heely’s 1888 patent) (Thomasson 2007). This result is controversial for those who try to account for intuitive artifact kinds such as corkscrew and chair (Thomasson 2007; Juvshik 2021b).

Many philosophers, on the other hand, emphasize the intentional aspect of functions rather than the etiological aspect. Artifacts, after all, for intended functionalists, are, in a significant sense, dependent on the activities of conscious agents. Given the importance of intentions of the relevant agents, intended functionalists claim that artifacts have functions that make necessary reference to our “needs, desires, and plans” (Thomasson 2009, 205). Thus, according to intended functionalists, artifacts have the functions because their makers bestow them those very functions.

However, this quickly leads to the following problematic cases: Some corkscrews are only produced or used for aesthetic purposes and are not intended to remove any cork. Similarly, some ships and chairs are produced as exhibition ships and chairs (Bloom 1996, 5). We can add motors, cars, guitars, and many other artifacts to the list. Bloom presents these cases as a threat to intended functionalism. Because in such cases, either one should admit that artifact kinds are not united by a shared intended function or that those particular entities are not members of the relevant artifact kinds.

One can defend intended functionalism by underlying the feature of reproductive success that is associated with the functions. Chairs, after all, are reproduced throughout history because they were highly useful in seating people, not because they are good decorative pieces in exhibitions. This is, for instance, the route Evnine takes in his distinction between kind-associated functions and idiosyncratic functions (2016, 119–124). For Evnine, the kind-associated function of chair is to be sat upon, while if someone produces a chair for exhibition purposes, then that chair has an idiosyncratic function (being an exhibition piece) in addition to its kind-associated function (seating a single individual). Thus, for Evnine, artifact functions are still present even when they are not performed or not intended to be performed (2016, 121–124).

Although Evnine’s distinction seems to secure kind-associated functions for Bloom’s cases, it still suffers from a more serious case: artworks. Artworks are considered the epitome of artifacts. However, if artifacts are grouped under an artifact kind by their kind-associated functions, then many highly esteemed artworks (especially the modern works after Marcel Duchamp’s The Fountain) of the twentieth and twenty-first centuries turn out not to be artifacts simply because they lack functions (Koslicki 2018, 218; Juvshik 2021b). Furthermore, even if specific paintings have functional properties such as invoking religious feelings (e.g., religious paintings), painting kind does not seem to have unifying functional properties (Juvshik 2021b). Thus, functional theories can only account for specific art kinds that are produced to fulfill certain functions.

To sum up, etiological function essentialism faces the extension problem because the view is extensionally inadequate—it can only provide an arbitrary fineness of grain at best and, thus, leaves out many familiar artifact kinds. In contrast, intended functionalism is better at dealing with intuitive artifact cases; nonetheless, the view suffers from the extension problem as it cannot easily explain Bloom’s cases (e.g., exhibition ships). Even if there is a possibility to parry Bloom’s cases, many non-functional artworks still constitute a deep extensional worry.

Given the heterogeneity of the artifactual world, some proponents of intended function restricted their domain of inquiry only to cover “technical artifacts” (Baker 2007, 49). This, however, leads to a further problem, namely, the definitional complexity problem (Mag Uidhir and Magnus 2011, 85). Mag Uidhir and Magnus (2011, 85) write, “In order to capture art’s plurality and thereby avoid extensional worries, definitions often become dangerously complex, borderline arbitrary, or circular.” Similarly, in the case of artifacts, delineating a distinction between technical artifacts and non-technical artifacts is not principled (Koslicki 2018, 235; Juvshik 2021b, 19). Because appealing to the “technical artifact” restriction cannot be profitably defined to exclude “technical” artworks (Juvshik 2021b). For instance, the cases of computer art discussed in Lopes (2009) show that there are technically complex artifacts that have no obvious function (Juvshik 2021b). Therefore, given the definitional complexities and extensional problems, it seems that both etiological and intentional theories of functions fail to serve as an overarching artifact concept. Acknowledging this problem, Evnine (2016, 129) also admits a kind of pluralism by considering artworks as sui generis artifact kinds.

1.2 Intention Essentialism

The basic motivation behind intention essentialism is rooted in Hilpinen (1992) and Bloom (1996). Bloom (1996, 10) writes, “Although someone can create a chair without intending anybody to sit on it, it is difficult to see how someone can create a chair without intending it to be a chair.” The upshot of Bloom’s insights is that function and shape do not provide a stable ground for artifact groupings, but the maker’s intention does.

Based on Bloom’s insights, Amie Thomasson further defends the essentiality of intentions (Thomasson 2003, 2007, 2009, 2014). According to her, what lies at the core of artifacts is the maker’s intentions:

Necessarily, for all \(\mathrm{x}\) and all artifactual kinds \(\mathrm{K}\), \(\mathrm{x}\) is a \(\mathrm{K}\) only if \(\mathrm{x}\) is the product of a largely successful intention that \((\mathrm{Kx})\), where one intends \((\mathrm{Kx})\) only if one has a substantive concept of the nature of \(\mathrm{K}\)s that largely matches that of some group of prior makers of \(\mathrm{K}\)s (if there are any) and intends to realize that concept by imposing \(\mathrm{K}\)-relevant features on the object. (Thomasson 2003, 600)

Unlike functionalist essentialist accounts, Thomasson’s intentionalist account does not imply any strict necessary and sufficient condition. Even if intention essentialism does not impose strict necessary and sufficient conditions, nonetheless, as the above quote shows, Thomasson claims that the maker’s intentions are necessary for all artifacts. Assuming that Thomasson’s intentionalist account constitutes some form of essentialism, it faces several problems. As I focus on the cases that seem to be artifact cases but fail to be one given the definitional restrictions of essentialist accounts, I will leave the discussion of other problems aside.⁴ Intention essentialism leaves out what I will call twilight kinds.⁵ Twilight kinds include kinds such as path, village, trail, footprint, doodle, etc. Members of these kinds are not exhaustively products of intentions. For instance, a path can unintentionally come into existence as a result of many agents’ repeated movements from one place to another via the same way (Koslicki 2018, 219). Similarly, people might decide to build shelters in a close range without any intention to create a member of the village kind, yet might end up unintentionally creating a village. Although some members of twilight kinds come into existence unintentionally, still, as a kind path or village, we seem to agree on their status as artifact kinds. If some members of these artifact kinds are not intentionally created, then this means those artifact kinds do not share the necessary condition of “intending to create a kind \(\mathrm{K}\)” Thomasson (2003) puts forward.⁶

Acknowledging the twilight kinds, Thomasson (2007, 58, n.5) slightly restricts her account by limiting her account to cover only “the essentially artifact kinds,” members of which are exhaustively produced with the right sorts of intentions. To my knowledge, this exclusion is not defended thoroughly, except in Juvshik (2021a), to some extent.⁷ According to Juvshik (2021a), there are two lines of argument against the intention-dependent nature of artifact kinds:

(1) Artifacts aren’t necessarily mind-dependent, but most of the artifacts around us happen to be.

(2) Artifacts are necessarily mind-dependent, but don’t need to be intention-dependent. (Juvshik 2021a, 9316)

To defend intention essentialism, Juvshik (2021a) considers five cases: Regarding (1), swamp and modal cases. Regarding (2), accidental creation, mass production, and automated production. Not all of these cases are relevant to my purposes. Leaving out mass production and automated production, I will discuss swamp and modal cases later. For now, I will focus on accidental creation. My ultimate critique of intention essentialism will take the form of (2).

The closest case discussed by Juvshik to the twilight cases is the case of accidental creation. Accidental creation is distinct from proper creation because, in the former, the intention to create that item is lacking. His discussion of accidental creation mostly revolves around the cases of failed-attempts-turned-into-new-artifacts. For instance, the piece of bread I forgot in the toaster turns out to be pretty good charcoal for my new drawing. So, I accidentally create a new piece of drawing charcoal. However, Juvshik aims to show that there is neither a toast nor a piece of charcoal, unless they are appropriated in the right sort of way. The moment of my appropriation of the failed toast as a piece of drawing charcoal marks the moment of the new artifact’s coming into existence. Appropriation also requires me to have, at least, a basic awareness of the relevant success conditions of making a piece of drawing charcoal.

However, twilight cases do not result from failed attempts. Instead, their coming into existence does not involve attempting to create an artifact. Yet, Juvshik might respond that, even if some members of twilight kinds are not failed-attempts-turned-into-new-artifacts, they are still non-artifacts, unless they are correctly appropriated. If that is the case, then the path formed as a result of my repeated commuting from the barn to the house is not actually a member of the path kind. Unlike Thomasson, Juvshik rules out not the kind itself but the unintentional cases. However, this will end up admitting that a large number of twilight cases, even though they share a similar morphological structure with their intentionally created counterparts, are ultimately waiting for an appropriator to confer them a status of artifactuality. I do not think that an archeologist or an anthropologist would accept the result that the unintended path was not created, say, one thousand years ago, but at the moment they approve it as a path. Archeologists and anthropologists discuss the significance of the path for that culture regardless of it being a product of specific intentions. Thus, contrary to Juvshik, I think that the twilight cases amount to genuine artifact cases without requiring a strict intention dependence. Twilight cases can be considered mind-dependent without being intention-dependent since their coming into existence requires the presence of agents with cognitive capabilities.

An intention essentialist might also respond by weakening their account only to require mind-dependence. However, the weakened account would not be helpful in distinguishing many other mind-dependent entities from artifacts. For instance, since the existence of many kinds of plants (e.g., seedless grapes) requires human activity, these plants and animals would be wrongly included in the domain of the overarching artifact concept, for which the only necessary condition is being mind-dependent. This strategy, therefore, would not be desirable for an intention essentialist who works in a monistic framework.

Even if one agrees that the twilight cases pose a legitimate worry against intention essentialism, a proponent of intention essentialism can still point out that those cases are a burden for everyone and thereby suggest that those cases are best left out until our most promising theory can account for them (Juvshik 2021a). However, we should not opt for the inference to the best explanation without examining other alternatives in depth. There is a neglected alternative. I will outline artifact concept pluralism as an alternative to the artifact concept monism after I challenge artifact anti-essentialism in the next section.

2 Artifact Anti-essentialism

The preceding discussion indicates that there seems to be a plethora of essentialist accounts. In contrast, unfortunately, there is not any fully developed anti-essentialist account. This is the reason why Koslicki (2018, 237–240) discusses general anti-essentialist frameworks that might apply to the case of artifacts. Here, I will focus on the artifact literature in order to extract some anti-essentialist views.⁸

One anti-essentialist strategy takes artifact groupings as context-relative. Reydon (2014, 133) defers the task of grouping artifacts to particular relevant epistemic contexts. For Reydon (2014, 137), “These epistemic contexts include academic disciplines such as archeology, art history, cultural anthropology, museum studies as well as engineering and design practices.” As explicated in the previous section, etiological functions, intended functions, and maker’s intentions fail to provide an overarching account. Given the problems they face, each requires some form of domain restriction, and, thus, for Reydon, to avoid counter-intuitive or arbitrary restrictions, we should settle down the ontological questions only after determining the epistemic context (2014, 141). Thus, the main task of a metaphysician (or, in this case, an anti-metaphysician) is to track how the different artifact concepts are used in the relevant epistemic contexts.

According to Koslicki (2018), pure context-relative solutions of artifact anti-essentialists are not plausible in the case of artifacts. Koslicki (2018, 239) writes, “Empirical questions only arise once we have taken as fixed that screwdrivers are primarily intended to be used by agents who wish to engage in certain kinds of actions, viz., to tighten and loosen screws.” This implies that we engage with artifacts not on an explanatory basis but on practical grounds (Koslicki 2018, 239–240). For Koslicki, while we engage with the members of natural kinds to discover their shared properties, what it means for an entity to be an artifact is something we decide before we engage with the candidate entities.

Reydon agrees with Koslicki that the metaphysics of artifacts primarily aims at specifying the general nature of artifacts before we engage with them. However, Reydon (2014, 141) argues that metaphysical approaches, so far, have failed to agree on how to specify the general nature of artifacts; that’s why it is better left “open.” One implication of leaving the nature of artifacts open is that if metaphysical approaches are far from settling on the general nature of artifacts, we should better track how epistemic contexts fare with artifacts; only then can it be decided whether “an overarching metaphysics of artifact kinds is feasible or a pluralist metaphysics is required” (Reydon 2014, 142). Agreeing with Reydon, I believe context relativity can help us solve the definitional and extensional problems artifact essentialism faces. However, I do not believe that the solution is purely epistemological. In the remainder of this paper, I will argue for an epistemically informed pluralist metaphysics for which Reydon’s discussion paves the way. Once I explicate the form of artifact pluralism I have in mind, I will qualify this claim in section 3. For now, note the following points by Mag Uidhir and Magnus that make pluralism suitable for both the species concept and the art concept. I adapt the following points for artifacts.

Multiple concepts are profitably used by practitioners [1]… Even without a settled [artifact] concept, we are able to agree on the rough boundaries of many [artifact kinds] [2]… [N]o [overarching] concept can profitably apply to all instances [3]… Some of the concepts involve an arbitrary fineness of grain [4]… (Mag Uidhir and Magnus 2011, 90)

Artifact anti-essentialists seem to endorse [1] and [3]. They use [2] to argue that the nature of the artifact concept is better left open. However, they miss the fact that not only do we agree on the rough boundaries of many artifact kinds but also on the ways individual artifacts can be grouped under those artifact kinds. The pluralism I motivate in section 3 is also similar to the anti-essentialist proposals in spirit. I take that there is no single way of dividing the artifactual world. Recent theories concentrate on at least four productive artifact concepts: morphological artifact concept, purely intentional artifact concept, and intentionalist functional artifact concept. I argue that even though none of these concepts are extensionally or definitionally unproblematic, they still play distinct yet significant roles both in ordinary talk and other disciplines. Instead of completely withdrawing from classificatory aims or leaving the nature of artifacts unspecified, I suggest that by adopting artifact concept pluralism, we can rather focus on the merits of artifact concepts individually. For now, I will turn to another anti-essentialist account that might be based on Richard Boyd’s Homeostatic Property Cluster (HPC) view, which aims to account for the extensional and definitional problems artifact essentialism suffers from.

2.1 The Homeostatic Property Cluster View

Reydon (2014) outlines the second anti-essentialist strategy by considering the possibility of artifact kinds being homeostatic property clusters. But he does not expand on it. I think it would be informative to explicate the HPC view briefly and contrast a possible anti-essentialist view based on the HPC view with the pluralistic metaphysics I have in mind.

Boyd (1999) develops the HPC view for natural kinds. According to the HPC view, members of a certain kind are not united by virtue of necessarily instantiated essences but by virtue of similarities. The similarities among the members of a kind are stable enough to sustain our taxonomical practices. Furthermore, these similarities are not clustered arbitrarily; as Boyd (1999) argues, they result from some “underlying homeostatic mechanisms.” One advantage of the HPC view over essentialist proposals might be that it accounts for the flexibility and change in both natural kinds and artifact kinds. The reason is that the HPC view takes the nature of species as open. This means that the HPC view takes the nature of species, contrary to traditional species concepts, is not fixed by some essential properties (Reydon 2014, 134).

However, a quick concern regarding the kind membership conditions arises against the HPC view: How do we assess whether a given organism or an artifact belongs to a certain kind? The answer is not straightforward. The HPC view suggests that there is a property cluster associated with a kind. The properties are not necessary or essential to a given cluster because it can lose some of the associated properties or gain others over time (Reydon 2014, 134). Furthermore, Boyd (1999, 143) claims that not all members of a kind need to instantiate all the properties of a given. For instance, assuming that the kind chair has the functional property of seating a single individual necessarily, a functional essentialist would expect all individual chairs to have that functional property. However, since the HPC view takes properties as neither essential nor necessary when adapted to artifact kinds, the view admits the possibility of non-functional chairs. Thus, an exhibition chair or a malfunctioning chair (or a chair beyond repair) can be considered as a member of the chair kind. The reason is that the HPC view still seems to work if artifacts instantiate only some properties associated with an artifact kind. Adapting the HPC view to artifacts, one can leave which conditions are minimally necessary and sufficient for an artifact to be a member of an artifact kind as unspecified. Although there are no minimally fixed necessary and sufficient conditions that entities need to satisfy, this still does not mean that the nature of artifact kinds is determined arbitrarily. Similar to the case with species, according to the HPC view, the properties associated with a certain artifact kind might result from certain causal-historical relations. These causal-historical relations might, for instance, include the reproductive history of an artifact kind being selected for a certain intended function over a certain period, which, in turn, might not result in associated properties as stable as in the natural kinds. However, this might be the price an anti-essentialist who argues in the line of the HPC view might be willing to pay to account first for the extensional problem artifact essentialism faces and second for the evolutive nature of artifact kinds. One benefit, or for some philosophers an additional cost, of the HPC view is that this form of anti-essentialist account, in turn, might admit accidental creations as well as byproducts that lack intentional properties. Simply because, in this view, artifact kinds do not have their associated properties necessarily or essentially.

Although an anti-essentialist view advanced in these lines seems to account for the extensional problems, the cost is worrying. Eliminating the necessary and essential features from artifact kinds leaves us with vague boundaries, as Reydon (2014, 140) acknowledges: “the HPC view fails to provide membership criteria for kinds.” I believe this cost stems partly from assuming the monistic framework as the backdrop because, according to anti-essentialists, if it is not possible to come up with an extensionally adequate overarching artifact concept, then the nature of the overarching artifact concept should be left open.

Consider the following case with anti-essentialism about art concepts. To account for the revolutionary artworks of the twentieth century that defied the limits attributed to the preceding artworks and art traditions, Weitz (1956) argues that we should regard art as an open concept. This does not mean that the nature of art is lacking; rather, it means that there is not any property such that it is necessary for something to be an artwork (Mag Uidhir and Magnus 2011). Similarly, an anti-essentialist view based on the HPC view proposes to account for the flexibility that artifact kinds show at the cost of denying necessary properties. However, just as being an artwork seems to require something to be an artifact, being an artifact seems to require, at least, one necessary property: being mind-dependent.

If artifact kinds are not necessarily mind-dependent, in other words, if artifact kinds do not require the presence of agents with cognitive capabilities, then there seems to be no basis for discarding the swamp and modal cases from our artifact ontology (Juvshik 2021a). Swamp artifact cases are cases in which an entity structurally similar to paradigm cases of artifacts comes into existence by sheer luck. Modal artifact cases are artifact cases occurring in a possible world that lacks agents with cognitive capabilities (Juvshik 2021a). A proponent of the HPC view might respond to modal and swamp cases by claiming that those cases lack the causal and historical mechanism required for the existence of the members of the HPC clusters. However, this answer is in tension with the principle claim of the HPC view. Consider the following case: Due to a strange accident of nature, a swamp village comes into existence at time \(t\). Then, what would preclude one from arguing that the nature of the village kind is changed in a way that, after \(t\), the village kind does not have being mind-dependent among its associated properties? I can imagine that the proponent of the HPC view might deny that a single case suffices by itself to change the nature of an artifact kind. However, it is not hard to twist the example so that many modal and swamp villages come into existence over a certain period of time. The point is that I do not see a reason why sufficient frequency of modal and swamp cases would not participate in determining the associated properties of a given artifact kind. As a response, one can insist on the necessity of causal links between human activity and the artifact kinds, but this undermines the HPC view’s main thrust as a form of anti-essentialism.

The pluralism I outline in the next section shares the main motivation of an anti-essentialist account based on the HPC view briefly outlined in this paper. That is, to account for the extensional problems without restricting the scope of the term artifact. However, instead of completely eliminating necessary or essential features from the picture, I suggest that we should adopt pluralism without giving up on the mind-dependence condition. Pluralism takes note of the benefits of the artifact concepts individually. Moreover, there are only a limited number of candidate artifact concepts that direct us to fruitful taxonomic practices.

3 Motivating Pluralism

It is not surprising that a single characterization cannot easily capture the nature of all artifacts. This is already implied in many philosophers’ discussions. For instance, Thomasson (2014, 46) writes, “The very term ‘artifact’ is itself used quite loosely, and in many different ways, so there may be no single characterization of what is essential to artifacts that fits best.” Bloom, in a similar vein, states that intentions provide the best source for what is essential to artifacts, but not the one that is exactly correct (1996, 20). However, the background assumption of monism remains unchallenged despite the extensional problems monism leads to.

In this section, by outlining how species and art concept monism leave out other widely used senses of these concepts, I aim to draw a parallel to the artifact concept. I argue that in the case of the artifact concept, too, the multiplication of senses is not a vice but an advantage. However, this does not necessarily lead us to an unrestricted proliferation of the senses. Classifications such as “objects that can be used either as doorstops or as cleaning supplies” do not guide us to a useful concept (Koslicki 2008, 202).

3.1 Pluralism in Other Fields and Artifact Concept Pluralism

Biology provides many different species concepts, such as the ecological species, the phylogenetic species, and the biological species, just to name a few. Ereshefsky (1998) picks out three prominent species concepts that are used by biologists. However, different versions of each concept have pitfalls that leave certain organisms or significantly shared characteristics of those organisms out of the picture.

The phenotypical (i.e., morphological) species concept uses exhibited characteristics of organisms to sort them into species at a given time while ending up disregarding the evolutionary history of species. The biological species concept sorts organisms according to their sexual reproductive capabilities, simply leaving out asexual organisms that reproduce by other means (e.g., vegetative reproduction). The phylogenetic species concept traces the evolutionary ancestry of organisms to situate species in the evolutionary tree of life; however, due to the evolution, the phylogenetic concept does not provide a stable taxonomy (Ereshefsky 1998, 104–106; Mag Uidhir and Magnus 2011, 89).

Similarly, Mag Uidhir and Magnus (2011) argue that there are at least four distinct art concepts that are gainfully used by philosophers of art. These concepts do not overlap while agreeing in many cases. The aesthetic art concept emphasizes the formal properties of artworks and provides a valuable source of information primarily for perception-related cognitive inquiries. The historical art concept emphasizes the historical properties of artworks, useful for historical inquiries. The conventional art concept traces the norms governing the art world’s institutions and practices, providing significant information for sociological and anthropological studies. The communicative art concept focuses on the “representational, semantic, or expressive content” of artworks, serviceable for learning and emotion-related cognitive inquiries (Mag Uidhir and Magnus 2011, 92).

According to Mag Uidhir and Magnus (2011, 92), in both types of pluralism, insisting on monism ends up in a parochial understanding of the relevant domains. Arguing for a single overarching concept disregards the other fruitful senses of both the species concept and the art concept. As explicated above, for instance, in the case of the species concept, the biological species concept does not range over asexual organisms, whereas the phenotypical species concept does. Similarly, in the case of the art concept, the conventional art concept excludes outsider art, whereas the aesthetic art concept can range over those cases (Mag Uidhir and Magnus 2011, 92). However, admitting pluralism does not mean that all senses of art or species are fruitful. The relevant senses that pluralism should include are epistemically informed; in other words, these concepts must already be in use among the practitioners (e.g., biologists, art critics, historians, and philosophers of art). Mag Uidhir and Magnus (2011, 90) name this form of pluralism “responsible pluralism” to distinguish it from “anything goes” approaches. Granted that an epistemically informed responsible pluralism is possible for both species and art concepts, in the remainder of this section, I try to motivate a similar form of pluralism for the artifact concept and defend it against possible objections in section 4.

My aim in this paper is to outline a rough guide for artifact concept pluralism. It is enough for pluralism if I can show at least two different artifact concepts are well-motivated. I state four. These are morphological, purely intentional, intentionalist functional, and residual artifact concepts. I choose to focus on these four concepts as I believe the combination of these four concepts provides the best result extensionally. Before turning to the relevant domains and purposes, let me first briefly state the candidate concepts I have in mind:

Thus, under the morphological artifact concept, we can say that artifacts are grouped into artifact kinds based on their displayed similarities to other members of artifact kinds. These objects need not have functional properties or be intentionally created, but they are mind-dependent. The notion of similarity is vague, and it is left unspecified purposefully as some variations of the morphological concept may require more strict similarity and thus result in a finer-grained classification, whereas others, depending on the inquiry, may involve a coarse-grained classification (Vermaas and Houkes 2013; Franssen and Kroes 2014; Elder 2007).

It must be noted that the list of artifact concepts briefly elaborated above is not exhaustive; it only aims to cover the widely used senses of artifact concept. As expected, these artifact concepts share many of their extensions. In the case of species and art concepts, people can use “species” and “art” distinctly without specifically stating the concept they use (Mag Uidhir and Magnus 2011, 92). Similarly, in the case of artifact concepts, folk classifications, as well as social sciences and engineering practices, use the artifact concept quite liberally.

By shifting the focus, we do not have to settle down the problem cases as “spoils to the victor” (Juvshik 2021a). The winner-take-all approach flat-out rejects the problematic senses of the artifact concept. However, in a pluralistic framework, we can fruitfully approach specific kinds of problem cases within the boundaries of a specific artifact concept and see to what extent that concept manages to account for such cases (Mag Uidhir and Magnus 2011, 92–95). Many consider artworks as artifacts (Dickie 1984; Levinson 2007; Mag Uidhir 2013). If some artworks are not functional, then we can better approach the philosophy of art with a purely intentional artifact concept as the backdrop.

The substantive necessity of intention-dependence should be seen as posing a philosophical constraint not just for any theory of art but also for the philosophy of art itself. That is, we ought to expect any and all philosophical enquiry into art and its associated relata (i.e., the nature of art, artworks, art forms, art practices, art ontology, art interpretation and evaluation, etc.) to yield conclusions at least minimally consistent with, if not directly informed by, the basic background assumption that intention-dependence is a substantive necessary condition for being art. (Mag Uidhir 2013, 5–6, emphasis in the original)

According to Mag Uidhir, the intention to create an artwork provides significant information regarding the nature of that artwork. Thus, even though a certain snowy hill may have more exciting aesthetic properties than Pieter Bruegel’s Hunters in the Snow, with the purely intentional artifact concept in mind, we can rule out such cases since they are not artifacts, hence not artworks.

This means that depending on the inquiry, we may need distinct concepts to classify certain artifacts. For instance, in the historical inquiries conducted by archeologists, shape may play a crucial role in evaluating the cultural significance of the found object. Archeologist Steven Mithen (2007, 290) notes that “Polly Wiessner (1983), for instance, studied the arrowheads of the !Kung bushmen of Southern Africa and documented how their specific shapes are not only effective at killing game but define individual and social identity.” !Kung bushmen’s arrowheads thus belong to different artifact kinds under the morphological artifact concept. In this case, it is not the function, but the shape plays a more important role in determining the membership conditions. One may object that it is not the shape itself, but the intention to create an arrowhead that has a certain shape is what plays this role. However, we can imagine a scenario in which a !Kung bushman can find an arrowhead-shaped stone in the forest; still, that arrowhead would provide a valuable source of information for archeologists. Furthermore, archeologists may not only classify found objects as artifacts but also accidental or unintentional creations, such as woodchips that result from making wooden spears, are considered to be artifacts (Fullagar and Matheson 2014).

Three things should be noted. First, the variations of the morphological concept result in arbitrary fineness of grain. For instance, depending on the inquiry and context, artifacts can be partitioned into fine-grained artifact kinds, such as Pasha Seatimer grand modèle automatique Cartier watch (Franssen and Kroes 2014, 78), or a coarse-grained classification, such as flat-bodied-with-protruding-legs (Kelly and Thomas 2014, 100). Counter-intuitively, as the !Kung Bushmen case exemplifies, the morphological concept might admit accidentally created or unmodified objects as artifacts, granted that they share a similar morphological structure to members of a certain artifact kind and show cultural significance. The intentionalist functional concept provides a stable taxonomy used both in folk classification and engineering practices; however, it leaves out artifacts that lack function (e.g., artworks). The purely intentional concept performs better in the case of artworks compared to the other two concepts. Given that none of the concepts can single-handedly capture the plurality of artifacts, then this can give us a reason to challenge the monistic framework itself.

Second, even though the pluralism I formulated suggests four concepts, these are not the only viable concepts. Depending on the context or inquiry, a more refined concept might be needed. So, even though I strongly suggest adopting pluralism in the case of artifacts, my wish is not to leave it static. There is no reason to reject that we might require more concepts in the future as taxonomic practices change. Consequently, a pluralistic framework that methodologically privileges actual practices should be flexible enough to capture the dynamicity of the taxonomic practices.

Lastly, all viable artifact concepts share a necessary condition: being mind-dependent. Given the methodology, this condition is needed to account for the current taxonomical practices. As our artifact practices dictate, the items that the concepts pick out should be those that have causal links to the human culture. That is why pluralism cannot afford to admit swamp and modal cases to the artifact ontology. To rule out such cases, therefore, pluralism needs to adopt mind-dependence as a necessary condition.

4 Objections

Pluralism seems to avoid the problems monism faces with relative ease. As we see in the previous section, pluralism shifts the focus from providing the best possible overarching artifact concept to retaining the merits of four individual artifact concepts. By shifting focus, pluralism offers a greater scope. Furthermore, pluralism does not need to appeal to the definitional restrictions to which essentialist accounts commit. However, the general worries regarding the nature of pluralistic approaches make pluralism undesirable. Here, I defend pluralism against three objections one can raise to make it more desirable.

First, one may object by arguing that adopting pluralism or any disjunctive supplementation brings its own complexities, and, thus, instead of clarifying the concepts, pluralism might end up adopting the “disadvantages of those concepts” (Vermaas and Houkes 2003, 275). Furthermore, Ockham’s Razor dictates we eliminate the murkier senses of a notion, not propagate them—the simpler, the better. However, the artifactual world is not less divergent than the biological world and the art world. Considering the heterogeneity of the artifactual world, I think a unified account is possible only in the case of ad hoc domain restrictions. Even in the case of domain restrictions (e.g., technical artifacts), there is a considerable amount of evidence from psychological research and engineering practices that led Vermaas and Houkes (2013) to argue for pluralism in the categorization of technical artifacts.

Vermaas and Houkes (2013) argue that certain classificatory practices in engineering coincide with psychological findings presented in Malt and Sloman (2007). Malt and Sloman’s experiment shows that there are, roughly, three major features that play significant roles in artifact classification: form (i.e., shape), functions, and intended category membership. Correspondingly, from their experience in the philosophy of technology, Vermaas and Houkes (2013) formulate three types of categorization principles for technical artifacts: id-made-product categorization, functional and goal categorization, and use-plan and make-plan categorization. Even though there are certain similarities worth mentioning, I will not get into details of Houkes and Vermaas’s account since here I attempted to motivate pluralism not only for technical artifacts but artifacts in general and across different disciplines. Each artifact concept I briefly pointed out provides partial partitioning; in other words, the success of a concept is not constrained by its scope, as each concept can only range over a certain portion of artifacts depending on the inquiry.

Second, one may point out that pluralism only amounts to a verbal dispute and claim that it is only a linguistic fact that we use distinct artifact concepts. So, according to this objection, pluralism only tracks people’s different usage of the term artifact rather than metaphysically important features, and there might be a metaphysically salient use of “artifact.” For instance, to account for the metaphysically salient features of artifacts, (dipert:1993?) suggests a tripartite distinction between tools, instruments, and artifacts proper. Leaving out the details, according to this distinction, artifacts are items that are made to be recognized as functional objects; as (dipert:1993?) puts it, they are “distinctively social.” However, his conceptual distinction results in an even more restrictive artifact concept than the restrictions we have seen so far (“technical artifact” and “essentially artifact kind”). Given that the aim is to account for taxonomical practices, the same extensional worries that apply to the previous accounts mutatis mutandis apply in Dipert’s case. So, Dipert’s distinction is not helpful. Going back to verbal dispute objection. Since pluralism tracks important metaphysical distinctions, I think this objection does not pose a threat to artifact concept pluralism. For instance, residual artifacts are not produced with the intention to create those items; they also do not have a specific morphological structure, so they are metaphysically different from intentionally created functional objects such as computers and airplanes. So, we need at least two different concepts to account for the metaphysical differences of these cases.

Lastly, one may doubt the accuracy of the analogy between species/art concept pluralism and artifact concept pluralism along the following lines: Our aim with artifact classifications is not primarily inferential or explanatory, whereas taxonomy for species and art concept is provided by the relevant specialists (Koslicki 2018, 239). Thus, our artifact classifications need not be based on specialists’ vocabulary. I agree that in the case of artifacts, folk classifications are not ultimately determined by the relevant disciplines and practices. For instance, I would not wait for archeologists’ validation for calling my favorite sitting device a “chair,” nor do I think I would be in error if that device turns out not to be a chair in some engineers’ classifications. However, the pluralism explored in this paper aims not only to describe folk classifications but give a more encompassing picture across different domains in which the term artifact plays an important role. Pluralism aims to provide distinct concepts for different inquiries and hence be an alternative to the arbitrary domain restrictions that stem from artifact concept monism. By changing the question from “What concept of artifact can best capture all cases?” to “What specific artifact concept can best capture the specific problem cases?” we need not approach a urinal, Duchamp’s Fountain, a toast, archeological woodchips, and nuclear reactors under an overarching artifact concept (Mag Uidhir and Magnus 2011, 92). Otherwise, as Preston (2014) points out, the gap between metaphysicians’ and other disciplines’ classificatory practices will continue to widen. This, in turn, may result in the philosophical term of artifact having no informative use outside of philosophy.

5 Conclusion

Artifact essentialists focused on finding an artifact essence. Artifact anti-essentialists claimed that there is none. In this paper, I challenged the monistic assumption that pervades the debate. I argued against artifact concept monism first by showing that the prominent essentialist proposals currently at play suffer from major extensional and definitional problems. Second, I aimed to show that current anti-essentialist accounts suffer from eliminating all necessary properties, which results in the proliferation of cases, as shown by the modal and swamp cases. Metaphysical literature on artifacts is a productive field. There are both compelling essentialist and anti-essentialist proposals yet to come. Adopting a pluralistic framework motivates a new focus on the neglected aspects of the artifactual world. I pointed out some of those aspects. Obviously, artifact concept pluralism invites many questions that I could not touch upon or give a detailed answer to. It requires a greater elaboration to properly flesh out the details; however, considering the significantly diverse roles artifacts play in our lives, I believe such effort is both needed and fascinating.

References