Models in the circular economy: envisioning waste’s potential

Editors’ note: This commentary is part of a series initially presented at the Waste/Economies/ Ecology hybrid international symposium at the Institute for Culture and Society at Western Sydney University, Australia in February 2023. The symposium was part of the research project ‘Investigating Innovative Waste Economies: Redrawing the Circular Economy’ funded by the Australian Research Council. The symposium brought together academics and artists from around the world who are thinking with waste to enable novel response to its ethical and political challenges. We thank all our contributors to this special commentary section for their participation and thinking.

enables us to arrive at a definition. Yet the variety it alludes to helps us recognize the range of models operating in the circular economy and heed their telling differences and contradictions.
In part, digital models in the circular economy hail back to the pre-industrial contexts of candle-, plate-, and horseshoe-making, when the terms 'mold' and 'model' were used interchangeably (Wharton 2021, 11). With industrialization, this notion of the model as a type or prototype evolved to be used widely in the service of standardized mass production alongside the claim to particularity of design. In the context of the circular economy, this notion of the model informs a retrospective classificatory lens onto waste. For instance, in the case of e-waste, it carries on in classification schemes involving item, brand, production year, etc. Such classification undergirds the menus of take-back vendor machines or platforms that ask users to identify and indicate qualities of the waste items that they are about to submit. Algorithmic applications can automate this process. This happens by means of visual recognition through image-based machine learning, whereby the challenge is to correctly recognize and distinguish the large variety of items and models. Another challenge is to recognize and classify those items that are damaged and deformed in wide-ranging or even unique ways. Distinguishing brand products from counterfeited ones in automated ways is considered a further achievement. In these cases, algorithmic models that incorporate and encode industrial models have a restorative function: waste can be redeemed and become again the product it once was. At least, components can be salvaged and reused.
Yet modeling technologies of circular economies do more than adhering to a fixed scheme and essentialist ontology of what items are or were, in origin. A second type of modeling technology seeks to exploit the potential inherent in waste matter and speculate and forecast what items can become. Waste as indeterminate potential draws on the Deleuzian-Simondonian notion of the virtual. Process-relation or immanence-based philosophies such as Deleuze's (Ivakhiv 2014) highlight material potential that 'is incorporeal and yet immanent to matter, forming a latent creative force' (Hoyng 2023, 3). In Deleuze and Guattari's (2005) example of metal, this potential encompasses the ability of alloys to change and recombine. Hence, under the right conditions, metal is capable of continuous transformation not just in terms of form but also in terms of matter. This transformability belies its appearance as a sturdy and solid substance. The potential value of metal seems readily recognizable, and indeed, Deleuze and Guattari theorize the pre-industrial historical practices of blacksmiths and metallurgy. But for those in the waste industry it is often exactly the unlikely possibility to repurpose a certain kind of waste matter that is most exciting and profitable.
If waste items' futures are not necessarily provided by their pasts, models are tasked with interjecting a prospective lens that relinquishes an ontology of discrete objects and classified items. Rather than object recognition and classification, this approach requires 'smart,' data-driven modeling technologies that engage in speculation about potential futures enabled by possible relations within which waste matter becomes something else. A model here is a mathematical description to optimize recovery, or rather profits from recovery, instead of a prototype, to return to Goodman's terminology. Rather than just modeling waste items only, such prospective models can include many parameters related to material design, consumer behavior, logistics, and global markets, along with sustainability indicators. The emergence of such modeling technologies underscores that the instability and indeterminacy of waste is not an issue of material agency that is intrinsic to waste per se. Rather, matter's proclivity toward transformation becomes actualized through particular material and social relations. Hence, a logistics company such as DHL has emerged as a somewhat surprising advocate of the circular economy, calling for action and embracing legislation such as the right to repair. Specialist logistics companies do not just facilitate operation and transport but distinguish themselves through the science they apply to optimize profits from secondhand devices and recovered materials by rerouting them. They are involved in state-of-the-art market prediction by means of dynamic price modeling through machine learning and real-time data analytics. Their promise is that profits are captured by anticipating shifting consumer needs and desires and delivering the second-hand consumer goods quickly to aftermarkets around the globe, when and where they can generate most income. In other words, the just-in-time mode of production of commodities (Rossiter 2016, 36) gets extended to their afterlife. Big Data, AI (Artificial Intelligence) and other algorithmic modeling technologies are promoted as instruments to navigate global markets and contexts characterized by overwhelming complexity. The Ellen MacArthur Foundation (EMAF) is a globally leading charity in circular economy advocacy and is affiliated with tech giants such as Google, Cisco, Apple, and IBM as well as institutions such as the World Economic Forum and the consultancy firm McKinsey. In their view, 'Ultimately, AI could be applied to the complex task of redesigning whole networks and systems, such as rewiring supply chains and optimizing global reverse logistics infrastructure, in any sector' (2019, 5). Following such a proposition, one might say that exploiting waste requires tuning in to general conditions of emergence and heightened eventfulness brought on by today's social complexity and market instability. And this, supposedly, has become possible with the arrival of modeling technologies.
Such circular economy discourses coalesce with Massumi's philosophical view of the prototerritory. Massumi (2015) argues that current historical conditions bring to the fore the importance of a prototerritory, which ontologically precedes the articulation and consolidation, indeed territorialization, of strata, spheres, and categories. For Massumi (2015, 45), this ontogenetic prototerritory is an 'active "zone of indeterminacy determined-to-be-determined."' It 'does not acknowledge any foundational categories and strata but only a "logic of self-organizing complexity."' Rather than 'a foundational replay of predetermined classifications,' at stake is 'a force of becoming productive of classifiable differences' (Massumi 2015). Building on the assumption of a prototerritory, waste exists as potential and indeterminacy determined-to-be-determined, its value and force consisting in its ability to become distinctive in self-organizing, complex relations. Contrary to the conventional epistemological link between the physical and the visible, the importance of waste matter does not lie in what is actually and physically 'in front of your eyes,' but in its potential as part of the evolving prototerritory that may not be visible to the human eye. The latter appears in the speculative gaze of modeling technologies, providing those in possession of such technologies with unique actionable knowledge. Meanwhile, the actualization of classifiable difference, that is, the emergence of particular, classified products such as marketable metals, rare earths, or plastics, is mediated by such technologies. Halpern's (2020) remarks on data-driven strategies for resource extraction in natural mines also apply to the context of mining waste in the circular economy: the enthusiasm about novel technological possibilities represent a 'fantasy of stretching finite resources to infinite horizons' (248). The stated aspiration is that circularity eventually undermines the need for humans to take anything more from the planet thanks to optimized waste mining and exploitation.
As modeling technologies bring into vision waste's potential and articulate chaos and complexity as circularity, what are their politics? I want to tackle this question by drawing on information theory. In the vocabulary of information theory, waste recovery hinges on a transition from noise, which corresponds to contingency, disorder, and uncertainty, to signal. Yet the identification of noise, and, here, of waste as noise, is a matter of perspective and positioning, wherein noise can be seen as not simply disruptive of signal and order but also constitutive of them. Given the complexities and paradoxes surrounding the relation between signal/noise and order/disorder, how do models and modeling technologies aim to recover waste by making it signify again, and how does the restoration or induction of signification imply a certain kind of politics?
The restorative gaze that I discussed above involves the restoration of signal through digital models that recognize and classify waste, whereby industrial models serve as index. Waste items, as opposed to waste as amorphous mass or 'stream,' are not all noise but contain traces of information. Against the anthropological association between waste and the lack of signification or information (Douglas 1984;Hoyng 2019), Offenhuber (2017) considers waste as the embodiment of traces and hence suggests that 'waste is information' (7). This is important in the context of the circular economy, as data about the composition and provenance of waste is crucial in recapturing value from waste matter (4). This restorative gaze is, among others, claimed by OEMs (Original Equipment Manufacturers), who seek to 'close the loop' on their own production. By means of collection and take-back programs as well as subscription services, this kind of circular economy supposedly endlessly upgrades commodities, whereby discarded items seem 'reborn' as improved versions of themselves, endowed with enhanced use value and sparkling desire again. For instance, Apple's much advertised disassembling robots put about 12.2 million refurbished iPhones out per year (Apple 2022). The company's recovery process encompasses disassembling and classification of items and components. Next steps for Apple's robots include the application of machine learning models that enable sorting e-waste at scale, X-ray imaging with RGB (Red Green Blue color code) imaging to improve the accuracy of automated disassembly and recovery, and high-frequency force feedback and machine learning, so robots can adjust their behavior while coming in touch with an object (Apple 2022).
However, whereas corporate communication foregrounds this idea of closing the loop as a technological aspiration, in practice, Apple appears to undermine current possibilities for a restorative circular economy. Its recyclers are bound by shredding contracts, and rather than restoring devices and salvaging components, Apple is in the business of mining waste for certified recycled materials, including plastics, metals, alloys, and rare earths. A circular economy for Apple, as an industry leader in sustainability, means that just 13% of cobalt in an iPhone is recycled and 30% of the used tin. In 2021, Apple achieved a record in using recycled materials, yet this was still only nearly 20%. 1 Meanwhile, the dominant practice of shredding implies willfully ignoring material traces inscribed in waste objects that offer sensing possibilities and could serve the development of data points; more so, it implies destroying them, and in that sense rendering them into noise. More abstractly, what remains 'noise' to the system are possibilities of reuse and inventive techno-material compositions and socio-material relations between humans and matter that do not abide by intellectual property rights and the 'purity' of untainted brand items. By privileging shredding and recycling over reuse and refurbishment, this type of circular economy exerts a preemptive power residing in stimulating circularities that foreclose these other possibilities associated with tinkering, DIY (Do-It-Yourself), pirate economies, and Shanzhai production. Apple has repeatedly hailed legal power to curtail those who act independently from the company to refurbish phones by means of harvesting and recombining used components. For instance, in 2020, the Washington Post revealed how Apple sued one of its recyclers that due to the doing of a number of roque employees had shipped used devices to China for refurbishment and resale, thereby violating the contractual requirement of shredding. 2 In another case, Apple sued a Norwegian repairer for deriving spare parts from refurbished phones coming from China that combined third-party glass screens with original Liquid Crystal Display (LCDs). Apple considered the repairer's practices an instance of counterfeiting and appealed to trademark law. 3 Though criminalized, what such repairers and informal-sector actors are doing is exactly what in circular economy terminology is called 'cascading,' or reusing discarded items in such a way that the highest use value of a component is retained before seeking lower performance applications and eventually recycling it with the purpose of material recovery.
The kind of circular economy that Apple seeks to develop is corporate-led and revolves around major brand OEMs. Corporate responsibility boils down to corporate agency, as the corporation becomes the owner of, as well as decision-maker over, waste matter (Hoyng 2023). Important from an ecological perspective, the decision-making regarding when a device gets shredded versus refurbished is left to OEMs and criteria for such decisions remain undisclosed as part of blackboxed models.
However, another set of actors, namely logistics and specialist recycling companies who operate in more diverse networks and markets, turn toward somewhat different sociotechnical infrastructures including prospective modeling technologies. Prospective models focus on the 'noise' emitted by potential and prototerritory to discover patterns in the data that indicate ways to optimize the exploitation of waste. Recovering waste is a matter of discovering signification and the latest machine learning technologies do so by making good use of noise, proving its constitutive function (Hui 2019, 26). In other words, these technologies aim to integrate contingency recursively and negate disorder by doing so (27). For these actors, modeling forms a response to the chaotic reality of disposal, global supply chains, and markets. They extend data-driven surveillance and control to the realm of waste matter (Hoyng 2023) by building on what Offenhuber (2017) describes as the disembodied understanding of information, namely 'the "it from bit" hypothesis' (6) that presumes that all physical phenomena can be translated into data. In line with cybernetics since Wiener, the abstraction of life and matter as data is geared toward command and control (Mills 2015;Simondon 1980)in this case, control over waste flows and markets that previously seemed overwhelmingly complex, opaque, and often even illicit. However, the prospective gaze relies on messy, heterogeneous datasets and speculative proxies and weightings in algorithmic models. Often lacking proper data and making use of experimental models overall renders these emerging practices in line with analyses of 'smart' phenomena that emphasize the role of uncertainty, speculation, doubt, failure, and contingency in data-driven algorithmic computation as well as the tactical and provisionary nature of applications drawing on such intelligence (Amoore 2020;Halpern 2020;Parisi 2013). Models of this kind may bring potential and prototerritory into (speculative) vision, thereby enabling tactical intervention, but they hardly can be said to afford order and control, as imagined in theories of surveillance societies and control (Parisi 2013).
It is striking then that, concurrent with the novel logistical possibilities, one version of discourse about the circular economy argues that it thrives most successfully if business players can act freely in the face of today's chaotic environments. Conceived in this manner, the circular economy is best served by a global free market that allows companies to innovate and discover new ways to turn waste into profit. The process would suffer from strict legislation, affixing matter to an ontology of product and pollution: once a matter such as e-waste becomes legislated as pollution and indexed as toxic or hazardous, its mobility and transformation would be undermined. For instance, in contradistinction to initiatives from environmental groups that fear violations of environmental and labor regulations in the developing world, e-waste recycling industry associations have pled for export to such regions rather than territorial containment (Hoyng 2018). In this view, waste matter can become either pollution/excess or resource, but it is neither in essence. In other words, waste is naturalized as suspension of difference. In cultural theory from Douglas's (1984) to Morton's (2013), waste has a liminal status vis-à-vis knowledge and information, and its amorphous appearance resists signification. Yet, in the context of the circular economy, such liminality does not elicit connotations of 'dirt,' 'pollution,' and 'hazardousness' or the sense of horror provoked by massive waste as amorphous 'hyperobject.' Contrary to broader cultural understandings, the business view seems rather comfortable with liminality and remains agnostic. And to the extent that waste is perceived as a risk and liability given its potential to turn into pollution if not hazard, this is offset by specters of opportunity and 'sustainable' exploitation of waste as resource in optimization strategies that seek maximal profits at minimal risk (but that do not exclude risk).
In the light of the dual problem of corporate control as well as neglect of possible risks, the question remains: what other circularities are possible? And what could the models or modeling technologies be like that mediate them? There are actual practices and current imaginaries that point to other possibilities. For instance, circular economies could build on histories of Shanzhai production and DIY tinkering but revise their small-scale, local nature. Such ideas emerge more often from China, which as a country has a different technomaterial culture from the West, shaped by its history of the local informal sector handling the world's e-waste and repurposing it as Shanzhai technologies. Moreover, newly proposed circular economy models consider consumers' interests, rather than corporate ones. There are proposals, for instance, for calculating carbon emission based on the real-time data of the flow and recovery rate of recycled items. Against leaving our understanding of possible circularities to select, exclusive machinic gazes that are trained to serve vested corporate interests, can we bring to bear a plurality of sensibilities on waste and sustainability questions? Would it be possible to expand ways of sensing, feeling, and tracing traces by combining heterogeneous human and more-than-human capabilities (Gabrys 2016, 66)? How would this allow us to more fully gauge the potentiality of waste? This includes not just currently foreclosed possibilities for reuse and recovery but also neglected risks of pollution and hazards, and, even, the uncertainty stemming from the overhasty adaptation of 'innovative' materials and chemical formulations in the tech industry.