1. Introduction
There is increasing interest in artifacts among philosophers. The leading edge is the metaphysics of artifacts and artifact kinds (Baker 2007; Dipert 1993; Elder 2004; Franssen, et al. 2014; Hilpinen 1992, 2011; Koslicki 2018; Margolis & Laurence 2007; Preston 2013; Thomasson 2007). However, an important question has been neglected. What is the status of the category ‘artifact’ itself? Philosophers have taken it pretty much for granted. It was left to Dan Sperber (2007), an anthropologist by training, to mount an ontologically-based argument against the theoretical usefulness of this category for the purposes of naturalistic social science. Moreover, there is some implicit support for Sperber’s argument in the philosophical literature that should prompt concerns about its usefulness for philosophical purposes. But Sperber’s argument has a fatal weakness. Thus, even if his conclusion is substantially correct—as I believe it is—his argument is not cogent. An alternative argument is required. The purpose of this paper is to supply one.
In Section 2, I lay out Sperber’s argument, which is based on the observed continuum between natural objects and artifacts, and review the implicit support for it in the philosophical literature on artifacts. In Section 3, I diagnose a fatal weakness of continuum-based arguments. They fail to take into account recent thinking about classification in philosophy of science, which has been forced to accommodate continua in biology and elsewhere. In Section 4, I focus specifically on the species problem in philosophy of biology, and on pluralism about species concepts as a response to it. I then identify an analogous artifact problem, and argue that pluralism about artifact concepts is an appropriate response. In Section 5, I show that just as species pluralism leads to skepticism about the species category, so artifact pluralism leads to skepticism about the artifact category. Finally, in Section 6, I return to Sperber’s conclusion that the artifact category is not theoretically useful and argue that it is substantially justified by an argument based on artifact pluralism and skepticism about the reality of the artifact category.
2. The Continuum Argument Against the Artifact Category
According to The Oxford English Dictionary (OED Online, March 2022), an artifact is something made or modified by human beings, as opposed to occurring naturally, without our intervention. Thus ‘artifact’ and affiliated terms mark a fundamental categorization of the world into natural objects and artifacts. The burden of Sperber’s argument is to show that this categorization fails because nature and culture are in reality continuous. Consequently, ‘artifact’ as a theoretical term cannot be usefully defined, since any attempt to do so will be frustrated by the continuum. We will call this the continuum argument.
Sperber begins with Risto Hilpinen’s (2011) frequently cited definition from The Stanford Encyclopedia of Philosophy: ‘An artifact may be defined as an object that has been intentionally made or produced for a certain purpose.’ This definition yields three dimensions along which artifacts are supposed to differ from natural objects.
Artifacts are the products of intentional action.
Their production involves the modification of objects and materials.
They are produced for a specific purpose.
Sperber shows that along each of these dimensions we encounter a continuous series of cases. He focuses on domesticated plants and animals, for here, if anywhere, nature and culture are thoroughly confused. On the one hand, Sperber says, many domesticates are ‘biological artifacts’ fitting Hilpinen’s definition. They are produced by intentional, selective breeding resulting in modifications of their morphology and behavior that suit them to serve specific functions useful to us. On the other hand, he argues, the realm of domesticates is rife with cases that do not fit the definition.
First, the modified traits of domesticated plants and animals often come about by unintentional selection (Sperber 2007: 125). For example, wild wheat was gathered and eaten by paleolithic hunter-gatherers. A trait of wild wheat is that the seed heads shatter when touched, which helps disperse the seeds. But a gene for non-shattering heads occurs with some regularity, so the gatherer was likely to come home with proportionally more seeds from non-shattering heads. Consequently, when she intentionally planted—or unintentionally dropped—some of her seeds around her campsite, she unintentionally started a breeding program to produce plants with the non-shattering heads characteristic of fully domesticated wheat (Smith 1995: 72–74).1 Domestication thus illustrates a continuously varying role for intention in the production of artifacts.
Similarly, there are numerous cases of unmodified plants or animals regularly used by human beings. Sperber (2007: 126) mentions leeches, which are typically gathered from the wild to be used for medical purposes. No modification occurs in this case, because breeding still occurs in the wild, not under human control. There are many similar examples, including Asian elephants (traditionally captured from the wild and tamed) and shade trees such as sweetgum or sugar maple (traditionally transplanted from the wild). There are also many plants, fungi, and animals that are bred under human control, but nevertheless remain identical with their wild counterparts. Examples are rooibos, shiitake mushrooms, and ostriches. What we see here is a continuous series of cases bridging from completely wild (no part of the organism’s life cycle under human control) to completely domesticated (living only under human control).
The centerpiece of Sperber’s argument is his discussion of function. He begins with the observation that biological artifacts have both biological and cultural teleofunctions.2 If there were a joint in reality between nature and culture, you would expect their biological and cultural functions to be distinct. However, Sperber (2007: 130–131) argues, they actually coincide. Domesticated organisms carry out their cultural functions in virtue of carrying out their biological functions and vice versa. For example, wheat carries out its cultural function as food for human beings by carrying out its biological function of producing seeds as reproductive devices. But counterintuitively, the reverse also holds. Wheat carries out its biological function of reproducing by providing food for human beings. Although we eat some of its seeds, we more than compensate the loss by protecting and planting the rest. Natural selection is quick to take advantage of this opportunity by modifying the wheat seeds to make them even more attractive to us. This is even true of the seedless grapes of Sperber’s title. Seedlessness might seem to be a purely artifactual trait, since the reproductive function of the fruit appears to be lost along with the seeds. But in fact, Sperber argues, the fruit retains the biological function of attracting us to eat the fruit and then spread the plant, not by dispersing seeds, in this case, but by propagating the grapevines vegetatively. This coincidence of biological and cultural functions in domesticates shows that far from being the locus of a divide between nature and culture, the realm of domestication is the locus of their imperceptible merger.
Sperber concludes:
Here I have tried to cast doubt on the idea that a theoretically useful notion of artifact can be built around its usual prototypes: bracelets, jars, hammers and other inert objects, or that it can be defined in a more systematic way. There is a continuum of cases between public productions that are well characterized by a specific purpose and others where purpose is unclear. There is also a continuum of cases between public productions that are wholly designed by humans, and others where humans exploit, with little or no modification, a pre-existing structure…. There is no good reason why a naturalistic social science should treat separately, or even give pride of place to, cultural productions that are both more clearly intended for a purpose and more thoroughly designed by humans, that is, to prototypical artifacts. (Sperber 2007: 137)
Sperber’s concern here is ultimately epistemic, but grounded in ontological concerns about the continuities between natural objects and artifactual ones, which provide no epistemically privileged place to draw the desired line between nature and culture. The artifact category is thus rendered theoretically unsound, and we are misguided in ‘taking artifacts as a proper category for scientific and philosophical theorizing’ (Sperber 2007: 136).
The continuum identified by Sperber has also been noted as problematic by a number of philosophers, but they do not take the further step that Sperber takes of declaring the theoretical bankruptcy of the artifact category. For example, Randall Dipert begins his essay on ‘artifact theory’ with definitions the stated purpose of which is ‘to demarcate artifacts from objects that are utterly natural objects’ (1993: 14). The result is a taxonomy of use objects that, counter to Dipert’s stated purpose, actually demonstrates the continuity between artifacts and natural objects.
Instruments are unmodified, naturally occurring objects intentionally used for some purpose, for example, a stone used to pound in a tent stake.
Tools are instruments that have been intentionally modified for an intended use, for example, a flint pebble knapped for use as a knife.
Artifacts proper are tools whose toolhood and intended purpose are intended to be recognized, for example, a screwdriver or a painting.3
This taxonomy encounters some difficulties. First, it turns out there is a no man’s land Dipert (1993: 33–37) calls the ‘artificial’ lying in between objects falling under his taxonomic categories and natural objects. It consists largely of objects that are unintended side effects of the modifications required to make tools or artifacts proper, such as smog or sawdust. We might add that this difficulty is further complicated by the fact that such unintended side effects are often intentionally used, as when you fertilize your vegetable garden with wood ashes. And what are we to say about something like a fingerprint, which may well start as a seemingly natural residue left unintentionally by a burglar, but which is quite a different animal after it has been dusted, black lighted, photographed, and digitally enhanced by the forensics unit? However, this evidence of continuity between nature and culture does not induce Dipert to question the integrity of ‘artifact’ as a category.
Similarly, an important strand in the recent philosophical literature demonstrates the continuity with regard to function. Ruth Millikan (1984, 1993) started this strand with an account of proper function she intended to be applicable to biological traits, artifacts, words, and mental states alike. Underlying this account was the idea that all of these are the products of a copying process that is selective, on the model of natural selection in biology. Crawford Elder (2004, 2007) adopted this copy-based notion of proper function as a central element in his own account of copied kinds. Members of a copied kind share a specific proper function, a literal or figurative ‘shape,’ and ‘historically proper placement’ in a context of other copied kinds (2007: 38–39). Elder, too, intends his account of copied kinds to apply pari passu in nature and culture. He draws the consequence succinctly:
There is no joint in nature between the artifacts copied by intelligent and conscious artisans, such as desk chairs, and the devices produced by naturally selected behaviors, such as beaver dams…. What is written in the book of nature is copied kinds. (Elder 2007: 51)
Elder thus makes explicit what was already implicit in Millikan’s program, namely, that there is no clear demarcation between natural objects and artifacts. But neither of them draws any further conclusions about the integrity of the artifact category.
Finally, Kathrin Koslicki (2018, Chapter 8), lists cases she considers problematic for the ontological project of distinguishing natural objects from artifacts and natural kinds from artifactual kinds. These concern exactly the sorts of continuities that exercise Sperber, including domesticates, found objects used for human purposes such as a piece of driftwood adopted as a coffee table, unintended outcomes such as paths or villages, residues such as sawdust, and possibly functionless objects such as art works. Koslicki concludes that there is no uncontroversial way, given current metaphysical resources, of splitting artifacts off neatly from natural objects. But rather than expressing skepticism about the artifact category on this basis, she vows to continue the investigation.
This leaves us with an important question. Should we follow Sperber’s lead and reject the artifact category? Or should we maintain it in the face of the continuum as other philosophers have done, albeit without specific justification? A closer examination of Sperber’s continuum argument in the context of recent thinking about such continua in philosophy of biology will show that Sperber’s conclusion is premature, but will also point the way to a different argument for a slightly modified version of that conclusion.
3. Who’s Afraid of the Continuum?
It is true, an objector might say, that there is a continuum between nature and culture. That is to be expected, since culture is necessarily built out of natural materials using our biological cognitive and physical abilities. We may well need a more fine-grained taxonomy with resources for capturing borderline phenomena. However, the existence of a continuum is not by itself a sufficiently strong reason for giving up the artifact and natural object categories, provided there are substantial phenomena that fit them. Even Sperber concedes that there are paradigmatic objects that do fit standard definitions of ‘artifact.’ Lots of them, in fact. Even some domesticates measure up. In short, Sperber has some salutary points and some really cool examples, but his rejection of the artifact category is not warranted.
This objection is substantiated by recent developments in the philosophy of natural kinds. A natural kind, as originally understood, is a group of things that belong together independently of any human interest or purpose in grouping them that way. Biological species were traditionally cited as the paradigm case. But changes in the understanding of biological species have cast doubt on this construal of natural kinds. Starting with Aristotle, biological species were defined by fixed and immutable essences, shared properties that were necessary and sufficient for something being a member of a particular species. For example, the bobcat (Lynx rufus) has properties common to felids, such as retractable claws, a muscular body, and obligate carnivorous behavior. In addition, it has properties distinctive of its species, such as a short tail and black bars of fur on its legs. For essentialism, species are distinct; there can be no continuum of cases between them. Darwin posed an enormous problem for this view. If new species evolve by incremental variation out of existing species, then there are no fixed biological essences and no sharp dividing lines between species. So either species are not natural kinds after all, or we need a revamped understanding of natural kinds that would allow for kinds with continua between them. Both options have been tried. Proponents of essentialism sometimes deny that species are natural kinds and focus instead on the kinds of physics and chemistry as paradigmatic (Ellis 2001). Biological organisms are often similar enough to each other that for practical purposes they may be treated as natural kinds, but that is as far as it goes. Another currently popular view is that species are not natural kinds, but rather individuals with organisms as parts rather than members (Ghiselin 1974; Hull 1976). This, too, solves the problem about the status of species as natural kinds, but like essentialism, gives us no insight into the continuum problem.
The other option—revamping our understanding of natural kinds—is a better bet in this regard. A leading example is homeostatic property cluster (HPC) theory (Boyd 1991, 1999).4 On this view, natural kinds are defined not by necessary and sufficient properties, but by property clusters which are relatively stable in virtue of underlying causal mechanisms that hold them together. Bobcats share properties in virtue of mechanisms such as sexual reproduction and subjection to common selection pressures. But these mechanisms do not operate completely uniformly, so property clusters can vary over time and across space. For example, bobcats occasionally interbreed with other lynx species. The result is known as a blynx, and at least some of them are fertile. Boyd (1999: 144) says that in cases like this it may not be possible to make a rational decision about which species an organism belongs to, thus rendering the extensions of some species indeterminate and the boundaries between some species vague. In cases where a decision is possible, it is made a posteriori, in light of locally relevant theoretical considerations. This is because the operation of the underlying homeostatic mechanisms may vary with local conditions in time and/or space, resulting in variations in the property cluster. HPC kinds are thus individuated not in terms of their changing, possibly indeterminate extensions, but as historical processes or entities.
What makes HPC kinds real in spite of their historicity and the indeterminacy of their extensions? On the traditional view, kinds are real if they reflect divisions in nature independent of any human interest or practice. But this view has been steadily eroded by recognition of the theory-ladenness of observation and the social construction of reality. On Boyd’s view, the reality of kinds cannot be divorced from the practices in which those kinds figure. Specifically, our inferential and explanatory practices seek to accommodate themselves to the causal structures of the world. Boyd’s ‘accommodation thesis’ (1999: 147) states that we can achieve this accommodation only in virtue of reference to kinds. It is their success in this role that constitutes the reality and naturalness of the kinds to which we refer. HPC kinds thus reflect the causal structure of the world, but not independently of human interests and practices in the process of accommodation to that structure.
HPC theory illustrates the crucial point that on some recent and well-regarded understandings of natural kinds, a continuum of cases between kinds is perfectly normal in many domains, and yet these kinds have the same claim to reality as kinds in any other domain. Furthermore, HPC kinds normally change over time, and may have vague boundaries. But HPC kinds are nevertheless theoretically sound and useful. Indeed, on Boyd’s view they are essential to the reliability and success of our inferential and explanatory practices. This means that Sperber’s argument does not go through. The continuum of cases he identifies, with the consequent problems in assigning objects to either the artifact or the natural object category, is not by itself sufficient to make the artifact category theoretically unsound. His argument would provide sufficient justification for this conclusion only for the classical essentialist, who denies that true natural kinds can ever be connected by such a continuum of cases. The HPC theorist, in contrast, would remain unmoved. The point here is not to claim that artifact kinds or the general artifact category are HPC kinds, but only that HPC theory shows that continuum-based arguments are not sufficient for rejecting kinds or categories. So, the dispute is provisionally settled in favor of the objector who wishes to retain the artifact category in spite of the continuum. However, pursuing our investigation, we will find that the artifact category is indeed theoretically unsound—just not for the reasons Sperber gives.
4. Pluralism
4.1 The species problem
In addition to the question of whether to construe species as natural kinds, biologists and philosophers of biology also confront a proliferation of species concepts, that is, definitions of the term ‘species.’ Jody Hey (2001) lists 24 of them. These concepts often cross-classify organisms, or leave whole swathes of them out. The influential biological species concept, for instance, defines species in terms of the ability to interbreed and produce fertile offspring, leaving out the many organisms which reproduce asexually. Moreover, the plethora of species concepts is not a trivial problem, since most of them have a significant degree of acceptance.
Responses to the species problem fall into two broad camps—pluralism and monism. Pluralists hold that more than one species concept is correct. Pluralism comes in more and less liberal forms. For example, some pluralists admit species concepts based on structural similarity criteria in addition to the historical criteria more standard in contemporary biology (Dupré 1993; Kitcher 1984). Others argue for accepting only species concepts based on historical criteria in order to maintain the connection to evolution fundamental to the biological sciences (Ereshefsky 2001; Mishler & Donoghue 1982). Monists reject pluralism on a number of different grounds. Writing in response to Kitcher’s arguments for pluralism, Elliott Sober (1984) asserts that it is much too soon to give up on a unitary species concept. Biologists may well reach a consensus, even if it takes several more centuries. More recently, some monists have argued that a unitary species concept has already been found. All species are separately evolving metapopulation lineages, a criterion which provides a unitary species concept at a theoretical level. Multiple secondary species concepts are still required at an operational level, though, to provide criteria for delineating species for specific purposes (De Queiroz 2007; Richards 2010). On this view, pluralism is, in effect, relegated to the arena of scientific practice, while monism is maintained for theoretical purposes. These monistic holdouts notwithstanding, pluralism remains a popular option.
However, pluralism raises important questions about realism, the idea that scientists are in the business of discovering the objective features of a reality that exists independently of our theories and practices. The proliferation of species concepts reflects, in large part, the legitimately different projects and explanatory aims biologists have. This makes it sound like the epistemology leads the metaphysics here, which, in turn, leads straight to anti-realism. If it is ultimately the interests of biologists that determine where the lines between species are drawn, then those lines do not reflect real, mind-independent divisions in nature. Consequently, species do not exist as objective entities, independent of the scientific practices and projects employed to delineate them. Some pluralists have supported anti-realism about species on precisely these grounds (Stanford 1995). But most pluralists have denied that their position forces them into anti-realism about species (Boyd 1999; Dupré 1993; Ereshefsky 1998; Kitcher 1984). The proliferation of species concepts is, as Marc Ereshefsky puts it, ‘the result of a fecundity of biological forces rather than a paucity of scientific information’ (2001: 140). The problem is not that biologists are failing to get it right, pluralists hold, but that there are multiple ways of getting it right. Consequently, multiple, real divisions in nature may be correctly identified as the basis for classification. It is, then, no wonder that different classification systems may cross-classify or fail to classify some organisms. But it is also no wonder that under such conditions the explanatory interests of biologists necessarily play a significant role in determining what counts as the basis for any given classification system. This is nevertheless compatible with realism about what these same biologists discover about how the world is divided up in reality.
4.2 The artifact problem
Is there an artifact problem on an analogy with the species problem? When we discussed Sperber’s continuum argument, we noted that definitions of ‘artifact’ aim to identify a boundary between natural objects and artifacts. But precisely because of the continuum, this is a fraught project destined to generate multiple, competing definitions. We mentioned two philosophical definitions: Risto Hilpinen’s (2011) influential proposal that artifacts are objects that have been intentionally made for a specific purpose, and Randall Dipert’s (1993) more restrictive proposal to reserve ‘artifact’ for objects which, in addition to meeting Hilpinen’s criteria, are intended to be recognized as meeting them. We must now look in more detail at a few representative definitions, with an eye to characterizing this incipient artifact problem more precisely. We can then ask whether pluralism is a reasonable response in the artifact case.
Dipert’s restrictive view reserves ‘artifact’ for objects that have been intentionally modified for a specific purpose, and are, in addition, intended to be recognized as such. He has a particular interest in art works as an example domain, but standardized utensils and household items qualify as well. What does not qualify are what Dipert calls ‘tools,’ that is, objects intentionally modified for a particular purpose, such as a nice, straight stick a gardener cuts and sharpens for use as a plant stake. Also not qualifying are ‘instruments,’ that is, natural objects intentionally selected for a purpose but not modified, such as a handy stone used to hammer in the stick. In acknowledgement of the continuum between natural objects and artifacts, Dipert (1995: 121) refers to tools and instruments as ‘near-artifacts.’ His definition is not as restrictive as it might first appear, though, for almost everything we would ordinarily call an artifact does qualify because at this point in history, the built environment we interact with day in and day out is pervasively standardized. In addition, Dipert claims that his definition covers artistic performances such as music or dance concerts. However, he rules out non-performance actions on the grounds that they are not normally intended to be recognized as intentionally made for a purpose.5 Dipert’s definition highlights the social and communicative aspects of artifacts. He points out that tools would not be effective if we could not recognize them and identify their purposes (Dipert 1995: 130). This makes it clear that for him the divide between natural objects and artifacts is between standardized use objects that communicate their intended purpose and everything else.
Definitions of ‘artifact’ also show up in formal ontologies, which are typically upper-level, domain-independent ontologies expressed in a formal language and designed to provide a basis for more specific ontologies and classification systems in domains such as engineering or natural language processing. A review by Stefano Borgo and Laure Vieu showed that only two out of five leading formal ontologies included an artifact concept, and in both cases, they maintain, the characterization is ‘quite shallow’ (2009: 274). In SUMO, for instance, artifacts are characterized as created physical objects whose parts are contiguous in space and time. This would rule out objects such as disassembled vacuum cleaners under repair and not yet assembled Ikea purchases. Borgo and Vieu propose to improve on these accounts by providing an artifact concept within another ontology, DOLCE. Their focus is a concept appropriate to a formal ontology that is also foundational, by which they mean a system of the most basic and general categories (Borgo & Vieu 2009: 273–274). The ideal is to formulate the concepts of such a system independently of any application domain, and yet in such a way as to promote their potential usefulness as the basis for a wide range of domain-specific formal ontologies. It is, then, no surprise to see that they have whittled the artifact concept down to intentional selection for a purpose (Borgo & Vieu 2009: 292). The object does not have to be modified in any way, nor does it have to be used for the intended purpose. It just has to be intentionally selected. So on this view, an interesting stone picked up by a hiker with the intention of taking it home and using it for a paperweight is an artifact. Indeed, it would appear that if the hiker loses the stone ten minutes later, it remains an artifact. This artifact definition constitutes a liberal counter to Dipert’s restrictive concept, incorporating into the artifact concept not only Dipert’s near-artifacts, but even erstwhile natural objects that are merely contemplated by a human being for a purpose. For Borgo and Vieu, then, the divide between artifacts and natural objects lies between objects that have been contemplated for some purpose, and those that have never been so contemplated.
Artifacts are also central in anthropology and archaeology. Encyclopedias of anthropology rarely include an artifact entry. But in one that does, the author, Daniel Miller, declines to provide a formal definition precisely because of concerns about the continuum. Asserting that ‘there is little point in attempting to distinguish systematically between a natural world and an artefactual one’ (1993: 398), he illustrates his misgivings with some useful examples. The most telling of these is seemingly natural forests on islands in the South Pacific, which are in fact the unintended products of generations of selection and cultivation of trees useful to the human inhabitants. This downplaying of intention also characterizes the use of ‘artifact’ in contemporary archaeology. In her preface to the Dictionary of Artifacts, for instance, Barbara Ann Kipfer defines artifacts as ‘anything made and/or used by humans, including tools, containers, manufacturing debris, and food remains’ (2007: vii). So unintentionally produced items such as bread crumbs from your morning toast or the shavings from a wooden wheel spoke are artifacts for archaeological purposes.
In contrast with Borgo and Vieu, modification is here taken to be the defining feature of artifacts, and intention is regarded as inessential. For the student of stone tool industries, for example, the flakes that are struck off a stone core, known as debitage, are important for a number of reasons (Andrefsky 2001).6 They can be painstakingly ‘refitted’ to the core, to yield information about how the core tool was made. This reveals knapping techniques, including the use of other tools such as hammers made from antlers. But it also reveals a great deal about the strategies and decisions made by the individual knapper, and about their manual dexterity and cognitive skills. In addition, the presence and characteristics of debitage deposits can tell the archaeologist a lot about where people lived and worked in prehistory, the organization of their toolkits, and their mobility patterns. The more general lesson here is that for an archaeologist the unintended products of human activity are just as important as the intended products when it comes to evidence of daily activity, technological organization, economic structure, and the like. The focus is therefore on making and actual use—which typically modifies the object used, even if only incidentally—not on intention. Thus, the boundary between artifacts and natural objects is drawn between items made or modified, whether intentionally or unintentionally, and everything else.
These definitions focus on very different aspects of artifacts, and consequently draw the boundary between artifacts and natural objects in different places. Moreover, just as multiple species concepts are widely accepted, so the artifact concepts we have reviewed are accepted in their different contexts. They are serious proposals made in light of legitimate discipline-specific motivations. So although it is not as well developed, there is a certifiable artifact problem on analogy with the species problem.7 ‘Artifact,’ we can now see, is a heterogenous category that focuses on different, real features of objects, on the one hand, and on different, legitimate interests of philosophers and scientists on the other. How should we respond? Is pluralism the best option here? To answer these questions, we need to do more than just assert the heterogeneity of the artifact category. We need to explore the options for reducing it to a unique, correct definition, as the artifact monist would require. If those options do not look good, we will have a reasonable case for pluralism.
The monist has two options. She might argue that it is too early to foreclose the possibility that one of the existing concepts, or one yet to be proposed, is correct. Or she might argue, as some have done with regard to the species problem, that there is a unitary, discipline-neutral artifact concept that can be applied in specific disciplines using secondary, operational concepts. First, might one artifact concept win out? Although it is not possible to rule this out in principle, it seems even less likely than in the species case. The definition of ‘species’ is for the most part an intra-disciplinary concern of biologists. But the definition of ‘artifact’ is of cross-disciplinary concern. In addition to philosophy, archaeology, and formal ontology, artifacts are of concern in other fields, including, but not limited to, engineering, artificial intelligence, art, and ecology. It is vastly unlikely that the denizens of all these disciplines will come to a consensus about how to define ‘artifact.’ The heterogeneity of their disciplinary interests and the complexity of the domains they study bid fair to sustain the heterogeneity of artifact concepts indefinitely. So the case for this first monistic option looks decidedly weak.
Nor does it seem likely that there is a unifying, discipline-neutral artifact concept on an analogy with the evolving population lineage concept proposed in the case of species by De Queiroz (2007) and others. It would have to be based on some common property of artifacts identified in all the proposed definitions, but it is hard to see what that property would be. The three best candidates are intention, modification, and purpose. However, as we have seen, it is precisely these properties that are in play in the generation of different artifact concepts. The formal ontology concept put forward by Borgo and Vieu dispenses with modification, whereas the archaeological concept dispenses with intention. And purpose goes by the board with intention, since unintended byproducts like stone chips and bread crumbs have no purpose as such. You can, of course, collect your bread crumbs for the purpose of casserole topping, but that is a de novo imposition of purpose, not an identification of an existing one. It is impossible to rule out the possibility that someone might come up with a common property involved in all accepted artifact definitions. But as things now stand, that possibility seems remote. The case for monism therefore remains weak.
Artifact pluralism is thus the best option. But we noted above that species pluralists face questions about realism. In the case of species, we want to know whether we have identified real, mind-independent boundaries in nature between groups of organisms. Doubt arises because the boundaries delineated by the different species concepts sometimes cross-classify organisms, or leave some organisms out of account altogether. Nevertheless, most species pluralists claim to be realists, on the grounds that the richness of the biological realm makes possible multiple, legitimate ways of drawing the boundaries between species. In the case of artifacts, we want to know whether we have identified the real, mind-independent boundary between natural objects and artifacts. Doubt arises because this boundary is drawn differently by different artifact concepts. Consequently, an object can be an artifact under one concept, but a natural object—or, in some cases, unclassifiable—under another. Artifact pluralists thus seem to be open to the charge that they are accepting arbitrarily drawn lines conforming to their own interests, not identifying real, objectively existing boundaries in the world. But they can defend themselves and remain realists on the same grounds as species pluralists, for their interests are directed to very real aspects of objects. Attention to the production process, for instance, leads to a focus on modification, while attention to the social lives of things leads to a focus on communication of intent. So realism about the multiple proposed dividing lines between artifacts and natural objects is defensible.
Sperber does not seem to have considered pluralism as an option, but is pursuing a general characterization that would provide a single, clear boundary between artifacts and natural objects. However, even were he to accept a pluralist view and concede the theoretical usefulness of multiple artifact concepts for different epistemic projects, he might contend that his original view is still relevant, if in a modified form. For we can still ask what makes all these concepts artifact concepts. How do we characterize the category that unifies them under that heading? And if we cannot characterize it successfully, why would such a category be theoretically useful? So Sperber may still have a point. Further exploration of the species problem will help us sort this out.
5. Categories
5.1 Does the species category exist?
Pluralism about species raises a further question about realism. If species concepts are heterogenous in the way described, how should we understand the category ‘species’ itself? What, if anything, holds these concepts together and makes them all species concepts? Marc Ereshefsky (1998, 2017) argues that nothing holds them together. Consequently, we should be anti-realists about the species category, while remaining realists about species taxa such as Mus musculus or Fraxinus pennsylvanica. Ereshefsky (1998: 115) points out that this is not a question about whether we should treat ‘species’ as a disjunctive category—either interbreeding species, or ecological species, or phylogenetic species, etc. That is merely a question about how we use the word ‘species.’ Rather, the question is an ontological one. Do all these different ways of delineating species rest on a common feature of the world that might ground our use of the same word? Ereshefsky argues that there is no such common feature. More specifically, there is no common property, such as a process or pattern, undergirding our use of the same word in all these cases. He concludes that as an ontological matter, the species category does not exist.8 Should we accept the same conclusion about the artifact category?
5.2 Does the artifact category exist?
We have already discussed the very different roles of intention, modification, and purpose in the various artifact concepts. But perhaps we can nevertheless identify a common process or pattern recognized by these concepts. Let us start with process. The question is whether there is some common process by which all the things falling under the various artifact concepts are generated that makes them artifacts. The obvious candidate is making: all artifacts are things that are made, usually by humans. One problem is that on some conceptions, living organisms can be artifacts. But the process of ‘making’ cattle is not at all like the process of making a pot. There is usually intentional modification in both cases, but in the case of cattle the modification is carried out by independent biological processes which we merely set in motion, whereas in the case of pots the modification is carried out by us in a hands-on fashion. Moreover, the archaeological artifact concept dispenses with intention as necessary to making, and the formal ontology concept dispenses with modification. So it does not seem that there is a common process here in any case.
What about a common pattern? The question here is whether there is some ontologically significant structure shared by all the objects that qualify as artifacts under the different concepts. Unfortunately, this option fares even worse because the heterogenous artifact concepts draw an equally heterogenous range of objects into the erstwhile artifact category. Some are living organisms, while others are non-living. Some are made of organic matter, others of inorganic materials. Some, such as food, cosmetics, and medicine, are consumables that are broken down and destroyed in the process of use. Others are designed to last indefinitely, such as coins and buildings. Some are tangible, concrete objects such as flowerpots and skis. Others are intangible or abstract objects such as languages, actions, or performances. In short, there is no ontological structure that could conceivably be common to all the objects that might count as artifacts under the concepts we have canvassed, let alone others yet to be developed.
At this point it seems the artifact category, like the species category, just does not exist in the traditional sense. But there is one final option we can try. We can ask whether the artifact category is an HPC kind. Rather than a common property, artifact concepts would only have to exhibit a relatively stable cluster of properties held together by an underlying causal mechanism. Given the vastly different artifact concepts that are required to encompass all the things to which we want to apply the term ‘artifact,’ which include even HPC kinds themselves (Boyd 1999: 175), it’s quite possible that no reasonably specific mechanism will be found. But this could be remedied by adopting one of the less demanding variants of HPC theory, such as Matthew Slater’s (2015) ‘cliquish stability’ which eschews underlying mechanisms in favor of stability alone holding property clusters together. What is the property cluster in the case of the artifact concepts? Provisionally, perhaps, intention, modification, and dedication to a specific purpose, since all of the artifact concepts we canvassed involve at least one of these properties. Finally, recall that the reality of an HPC kind depends in the final analysis not on the posited property cluster or underlying mechanisms, but rather on whether the kind is necessary to the process of accommodating our inferential and explanatory practices to the causal structure of the world. Our attention is thus redirected from ontological considerations to epistemic ones.
6. Is the Artifact Category Theoretically Useful?
Sperber claims that the continuum between nature and culture frustrates any attempt to construct a theoretically useful notion of artifact. The failure of the continuum argument means that a theoretically sound artifact concept is possible. Indeed, a plurality of them are not only possible but actual. Furthermore, the artifact concepts we reviewed from philosophy, formal ontology and archaeology all appear to be serving inferential and explanatory practices perfectly well in their respective disciplinary contexts, thus satisfying the HPC requirements for the reality and naturalness of kinds. To this extent, Sperber’s conclusion is incorrect. But as we noted, his real quarry is not this plurality of artifact concepts, which he does not seem to recognize, but a general artifact category which would unify these concepts while dividing artifacts off neatly from natural objects. To this extent, his conclusion may yet be correct and important. We have seen that if this category is a traditional one, then it does not exist because the concepts falling under it do not share a common process or pattern. But it might be an HPC category if it plays a necessary role in accommodating our inferential and explanatory practices to the causal structures of the world: if, in other words, it is theoretically useful. Thus unlike Sperber and Ereshefsky, for whom ontological considerations are decisive for theoretical usefulness, HPC theory urges us to take theoretical usefulness as partially determinative of ontological status.
We cannot just assume that a category whose ontological status is not well established is theoretically useless, for categories may have independent heuristic value. Ingo Brigandt (2003), for instance, claims that the species category should not be eliminated because its real role is epistemic. It is an ‘investigative-kind concept,’ that is, a category that holds open and guides the search for some yet undiscovered mechanism or property that holds together the many species concepts. This idea comports well with Koslicki’s call for renewed attempts to find a general characterization of artifacts. The artifact category might, then, be theoretically useful in preventing premature foreclosure of the accommodation process as we attempt to understand the realm of cultural products. On the other hand, retaining the artifact category would be counterproductive if it kept the search for a general characterization of artifacts open in the face of significant theoretical disadvantages to doing so.
Let us start with a theoretical disadvantage highlighted by Sperber, that is, the distortions introduced when things that are more explicitly designed and intentionally produced are given a privileged position, thus marginalizing equally important cultural productions that do not fit this model. This complaint makes even more sense on a pluralist view with a number of legitimate artifact concepts, each with its own set of paradigmatic objects. The mistake, then, is to take one of these concepts as more representative of the general artifact category, and its paradigmatic artifacts as the standard for all. This privileging would be a mistake, on Sperber’s view, because other kinds of artifacts are actually more important to understanding human life and culture. As he points out later in his article, from the Neolithic transition to agriculture until the 19th century transition to full-fledged industry, the preponderance of artifacts with which people had to deal on a daily basis were biological artifacts, that is, domesticated plants and animals. We might add that these non-paradigmatic biological artifacts are still the basis of the agricultural subsistence strategy on which we are now utterly dependent, regardless of how little most of us may have to do with them. Yet the focus in social science, and even more so in philosophy, remains on the bracelets and jars. In short, our efforts to accommodate our epistemic practices to the causal structure of the world will be held back by this mistake, which we are less likely to make if we eliminate the artifact category as a measure to keep artifact concepts on a more equal footing.
A second epistemic disadvantage of retaining the artifact category is the mirror image of the epistemic openness commended by Brigandt. According to him, investigative-kind categories not only hold open the search for a unifying category, but set ‘the standards of what counts as an adequate concrete concept or…definition’ (Brigandt 2003: 1313), thus limiting the options for constructing new concepts or revising existing ones. Pluralism, on the other hand, is inherently open to new concepts. So a unifying category, even if investigative and not substantive, pulls against pluralism, even when its proponents, like Brigandt, accept it in principle (2003: 1306). Eliminating the artifact category, then, would preserve the epistemic openness of artifact pluralism to new conceptions of artifacts and the new projects that call for them. This is even more important in light of the dominant role of biological artifacts in human history and current life noted by Sperber, for none of the artifact concepts we reviewed is particularly well suited to their analysis. So we may well need a few more artifact concepts to progress in our accommodation efforts in this domain.
Finally, retaining the artifact category distracts us from methodological considerations. This is particularly important in philosophy, which has focused on the metaphysics of artifacts to the detriment of both the epistemology of artifacts themselves and the epistemology of the study of artifacts. Because the discussion has revolved almost exclusively around getting the ontology right, objects such as domesticated organisms, mass-produced objects, byproducts, ready-mades, and the like tend to be regarded merely as ‘difficult cases confronting any attempt to offer a general characterization of what it takes to be an artifact’ (Koslicki 2018: 220). Eliminating the artifact category, and with it the drive for a general characterization, would encourage a view of such objects as significant phenomena in their own right, whose roles in human life and cognition are deserving of study. Similarly, it would promote questions about what our purposes are in studying particular groups of objects and what artifact concepts are appropriate given those purposes.
An adequate characterization of a general artifact category has been elusive. The theoretical usefulness of keeping this search open is limited to the negative goal of not abandoning a search that might eventually succeed. On the other hand, eliminating the artifact category would serve the positive goals of improving our methodological approach to the study of artifacts, helping us take advantage of the epistemic openness of artifact pluralism, and promoting a more even-handed approach to important domains of artifacts that have traditionally been marginalized. On balance, then, the usefulness of eliminating reference to the artifact category from our theories outweighs any usefulness accruing from its retention. Thus, not even interpreting the artifact category as an HPC kind can save it, as the accommodation of our epistemic practices to the causal structure of the world is better served by abandoning it.
7. Conclusion
We began with Dan Sperber’s claim that the artifact category cannot be defined in a theoretically useful way. But we concluded that his argument, based on the continuum between artifacts and natural objects, fails to take into account recent thinking about classification in philosophy of science, which has been forced to accommodate continua in nature. If Sperber’s conclusion is to be upheld, then, an alternative argument is needed. The species problem in biology has prompted many biologists and philosophers of biology to adopt pluralism. But pluralism leads to skepticism about the species category, for there is nothing that holds the heterogenous concepts together. We concluded that definitions of ‘artifact’ in philosophy and other disciplines are also heterogenous, that pluralism is a reasonable response, and that this leads to skepticism about the reality of the artifact category. Finally, we concluded that the artifact category cannot be saved by interpreting it as an HPC kind, because the accommodation of our epistemic practices to the causal structure of the world is better served by abandoning it. So Sperber was right. The artifact category cannot be defined in a theoretically useful way.
Notes
[3] Dipert summarizes this taxonomony in the introduction (1993: 17) and then articulates it in detail in Chapter 2.
[4] Ruth Millikan (1999) has a similar view based on a notion of historical kinds whose members share properties in virtue of historical relations they bear to one another.
[5] Simon Evnine has no such qualms. It is, he says, necessary and sufficient for something being an artifact that it be intentionally made (2016: 70). He claims that this definition covers everything from fictional characters and languages, which he calls abstract artifacts, to intentional actions such as turning on the porch light.
[6] For more on this archaeological definition see also Dipert (2014), Hilpinen (2011), Preston (2013), Schick & Toth (1993).
[7] The analogy is not exact because artifact concepts and species concepts are not on the same level. Species concepts tell us how to sort living organisms into an indefinitely large number of groups. Artifact concepts tell us how to sort objects into exactly two groups—artifacts and natural objects. So artifact concepts are on the level of organism concepts, which tell us how to sort objects into living and non-living things. However, a proliferation of concepts can occur at any level. Terms such as ‘naturefact’ or ‘ecofact’ are sometimes used to indicate objects that are neither artifacts nor natural objects. But this terminology is not current in philosophy and does not seem to have penetrated very far into the standard terminology of other disciplines, either. My sense is that such terms really are more an acknowledgement of the existence of the heterogenous definitions of ‘artifact’ than an attempt to carve out a principled third domain of objects. But we shall see. After all, fungi were considered plants for ages before they got their own kingdom.
Acknowledgements
This paper has its roots in a presentation to the Georgia Philosophical Society in 2010. I would like to thank the audience for their helpful comments and questions. Two anonymous reviewers for this journal prompted significant improvements in the current version. My thanks to them as well. Any mistakes that remain are solely my responsibility.
Competing Interests
The author has no competing interests to declare.