1. Introduction
What would the ‘sharing economy’ look like if platform providers optimised for racial and other forms of diversity? In this article, I consider that question. In so doing, I contribute to an enquiry that is both deeply interdisciplinary in nature and that has become nothing short of an absolute imperative in the US, where I live and work. My use of the term ‘optimisation’ signals the interdisciplinary nature of this enquiry. Having participated in – and benefited from – academic discussions with engineers, economists and social scientists on the topic of how to ‘re-engineer’ the next generation of sharing economy platforms, I believe that it will be crucial for such platforms to be conceptualised, designed and regulated in such a way as to achieve goals beyond efficiency and profit.Footnote 1 ‘Optimisation’ is the term my engineering colleagues use to describe the effort of achieving the best design to meet the specified goals. As to the question of how I define ‘diversity’ as a goal that must be prioritised, my primary focus is on racial and ethnic diversity, which again became the subject of a national reckoning in the US after the murder of George Floyd in 2020. However, it is important to note that the bulk of my analysis applies also to the connection between diversity and platform access on the basis of disability, gender, gender orientation, income and economic status, and many other forms of marginalisation. In meaningful respects, then, I define the pursuit of diversity in the sharing economy as an effort to optimise breadth of access.
My analysis is built on several assumptions. First, platform proprietors do not currently optimise for diversity, instead emphasising their own profit over all other priorities. Second, because of this single-minded mission and perspective, the challenge of optimising for diversity in platform design and operation is primarily a regulatory challenge. It requires regulatory innovation in forcing platforms to optimise for a broader range of values, the most important of which is to ensure fair and equitable rights of use and access by diverse consumers and workers involved with platforms. Third, while civil rights and discrimination laws are an important source of regulatory innovation in reforming platforms, other fields of law can and must be reformed in support of optimising diversity.
Property law is one such field. Indeed, it is one of the most important fields in which to reform laws for the purpose of optimising diversity because such reforms could be the most impactful in benefiting both consumers and workers involved with the platform economy. In this article, I explain why. Part 2 of this article reviews the widespread nature of race and other forms of discrimination in platform technologies. Part 3 uses core strands of property theory to analyse the ways in which racial privilege translates into property entitlements. Part 4 discusses a range of reforms within property law that can contribute to eliminating the value – and ultimately the fact – of whiteness as a property entitlement in the platform economy. While this article focuses on the US context, it raises issues that demand research attention in many other countries and contexts as well.
2. Whiteness bias in platforms
It is now a widely acknowledged fact that sharing economy platforms, and many other tech platforms, do not optimise for diversity. To the contrary, a rich interdisciplinary literature has captured the proliferation of racial and other forms of discrimination across such platforms. Presumably these platforms were not developed to cater specifically to the needs of those who are White or who have been privileged historically on the basis of status. Yet there is compelling evidence that many platforms fail to eliminate discrimination once it is discovered. Worse, some platforms appear to adapt in order to take advantage of the perceived benefits of discriminatory behaviour. In other words, some platforms choose to develop towards further discrimination. From among the sad overabundance of examples supporting these observations, this Part briefly describes two. I have chosen these two examples because they help me stake out my claims in Part 2 about the translation of whiteness bias in platform-based industries into a form of property entitlement.
An iconic example of the proliferation of racial discrimination via platform technology was provided by Nancy Leong and Aaron Belzer when they described the differing experiences of White and Black Uber customers: the former were able to obtain Uber rides quickly and easily, with negligible wait times or other inconveniences; the latter had more difficulty obtaining Uber rides quickly, easily – and sometimes at all. Leong and Belzer traced the differing experiences partly to discrimination by Uber drivers, who regularly appeared to give Black passengers lower ratings than those given to White passengers (Reference Belzer and Leong2017, 1271).Footnote 2 As recently as 2020, multiple studies also traced discrimination to the very algorithms used by Uber (Lu Reference Lu2020). In the US, these algorithms incorporated geographical and other data that reflected residential racial segregation resulting from redlining and other hallmarks of structural racism. Thus, differing valuations of the safety (or more opaquely, the ‘desirability’) of picking up passengers in certain neighbourhoods were made on the basis of race just as surely as algorithms used to determine creditworthiness have been shown to rely partly on race (Citron and Pasquale Reference Citron and Pasquale2014, 13).
I have analysed a similar manifestation of differential treatment on the basis of race and/or ethnicity by autocorrect systems, a form of technology that I describe as part of the basic technological infrastructure that we all need in order to participate in civic and cultural life today, but that has a different utility for those who are Anglo or White as compared to those who are not (Dyal-Chand Reference Dyal-Chand2021, 191). In particular, I have catalogued the ‘correction’ of non-Anglo names by a range of autocorrect technologies to the closest Anglo approximations of those names (e.g. Aziza to Alicia), or to words that somehow recognise ethnicity without recognising the value of the names as such (such as Rashmi to sashimi), or to words that are not proper names at all (e.g. DaShawn to dash away).
The proprietors of platform technologies have argued (often convincingly) that the product features that have contributed most straightforwardly to discrimination on their platforms produce other claimed benefits. For example, Uber’s ratings system is intended to increase transparency for both drivers and passengers, which Uber has claimed makes its ridesharing service safer for all involved (Sullivan Reference Sullivan2015). Airbnb and other home-sharing services have made similar claims about their ratings systems (Airbnb n.d.a). Meanwhile, technologies like autocorrect reduce many kinds of eccentricities (not just those that could be categorised as racial or ethnic in origin or motivation) to the most popular norms in order to increase the convenience of typing seamlessly – and in the case of small touch screens, with one’s thumbs. Whatever the claimed purpose of such features, however, these examples capture the disturbing reality that platforms – like human beings – cannot eschew race-conscious decision-making. To the contrary: like other forms of artificial intelligence, platforms clearly discriminate between consumers, providing greater benefits to whiteness and other privileged statuses.
Discrimination by platforms is not just limited to the treatment of their consumers; it also extends to platform workers. There is now a substantial and compelling literature on the labour and employment injustices generated by sharing economy platforms. Recent scholarship has also begun to capture the racial inequalities among workers that are perpetuated by such platforms. In her analysis of one survey of platform workers, Roithmayr (Reference Roithmayr2018) stated:
‘Approximately 14.5 million workers (about a third of the workers in this economy) are motivated workers, meaning that they make more than 40% of their income through their gig work, describe their work as their primary source of income, or say they can’t find traditional employment. A whopping 68% of these workers are people of color. […] Because workers of color have fewer options than their white counterparts, they are less free to refuse precarious work, and are more likely to form the core component of motivated workers on which the on-demand economy relies.’
Such studies have extensively documented the reasons why the sharing economy has become known by labour and employment specialists as the ‘gig economy’. While proponents of gig work herald its flexibility and accessibility to non-traditional workers (Rosen Reference Rosen2021), the reality is that gig workers get virtually none of the benefits and stability of traditional employment, including occupational safety protections, union representation, a minimum wage and much more (Fairwork 2023).
On this front too, it appears that platform proprietors were motivated to develop away from the traditional model of employment for reasons other than race discrimination. Thus, an early refrain of companies like Uber and Airbnb was that they were software companies whose purpose was to provide a community marketplace for peer-to-peer exchanges (Airbnb n.d.b). They did not employ the workers who provided goods or services through their platforms; they merely provided an efficient mechanism for such workers to find customers. When successful as a means of avoiding regulation, this refrain allowed platform proprietors to reduce radically the costs and liabilities associated with traditional employment models. But as analyses such as Roithmayr’s have shown, the negative impact is felt disproportionately by workers of colour (Dubal Reference Dubal2023, 1939; Roithmayr Reference Roithmayr2018). Thus, racial differentiation, and discrimination, is evident both in the way platforms have privileged certain racial and ethnic (and often gender) categories of consumers and in the way they have treated their workers.
3. Sharing whiteness
But why is unequal racial treatment by platforms a property issue? How does race discrimination create property interests and rights? Why is the value to White consumers using these platforms a property value? And why is the value to businesses that perpetuate such discrimination a property value? Answering these questions requires a two-step analysis of the ways in which the value of whiteness accrues to both the more privileged consumers of goods and services on platforms and the platform proprietors.
The first step builds on Cheryl Harris’s authoritative analysis of whiteness as property:
‘When the law recognizes, either implicitly or explicitly, the settled expectations of whites built on the privileges and benefits produced by white supremacy, it acknowledges and reinforces a property interest in whiteness that reproduces Black subordination.’ (Reference Harris1993, 1731)
Some scholars have concluded that, in making this claim, Harris was arguing that whiteness was a special form of ‘status property’ (Belzer and Leong Reference Belzer and Leong2017). However, there is no reason to confine her conclusions about the value of whiteness in such a manner. Whiteness is not just a special or limited form of property. To the contrary, it has the core attributes that the law recognises in a wide range of property forms and types.
Understanding this point requires an understanding of the core function, and indeed relevance, of property rights and rules in contemporary society. As Singer (Reference Singer2000) has observed, property rules support ‘a social system composed of entitlements which shape the contours of social relationships’. While we allocate property rights by giving ‘title’ to resources that we consider valuable to individual ‘owners’, the primary function of title, as Laura Underkuffler has described, is to govern our rights to such resources vis-à-vis others:
‘Property rights are, by nature, social rights; they embody how we, as a society, have chosen to reward the claims of some people to finite and critical goods, and to deny the claims to the same goods by others. Try as we might to separate this right from choice, conflict, and vexing social questions, it cannot be done.’ (Reference Underkuffler-Freund1996, 1046)
Moreover, our rights to resources encompass not just outright ownership (or title), but also a range of rights that Singer calls ‘entitlements’. As Singer discusses, entitlements to property regularly conflict, and regularly also, courts and other authorities allocate access to property on bases other than superiority of title (Reference Singer2000). One example of such entitlements about which Singer has written extensively is rights of access to public accommodations (Reference Singer2015, Reference Singer1996). By expanding our understanding of property rights as conflicting entitlements, Singer and Underkuffler have provided a foundation on which contemporary property scholars have rendered more visible the structures that create and maintain privileges over resources. Entitlements can include rights over ‘things’, broadly defined, that are and only become valuable property because they support access to and control over key resources.
This is where Harris’s insights become particularly relevant. As Harris demonstrated by means of a sweeping historical analysis, from the very origins of the US, White individuals and groups developed and nurtured entitlements to their own whiteness. In doing so, they protected their access to a broad range of critical resources while also excluding those who were non-White from those resources. Whiteness itself thus became a valuable resource (Reference Harris1993, 1721). Harris captured the value of whiteness partly by reference to surveys such as one conducted by Hacker (Reference Hacker1992), in which he described the responses of White students when they were asked how much financial ‘recompense’ they would request in return for having their racial identities changed from White to Black: ‘Most seemed to feel that it would not be out of place to ask for $50 million, or $1 million for each coming black year.’ (Reference Harris1993, 32) Examples such as these compellingly demonstrate the extent to which whiteness is a privilege that is commodified in social and economic relations. Thus, while Harris relied on Margaret Jane Radin’s theory of property as personhood to claim that whiteness is a form of ‘personal’ property that is essential to ‘self-realization’ (Reference Harris1993, 1730, 1761), it is evident that whiteness as property is perceived to have significant financial value. In this view, to harness it is to harness both a potent economic benefit and a powerful means of wealth creation.
Harris also articulated the contemporary value of whiteness as property by applying the ‘bundle of sticks’ view of property to this particular property form (Reference Harris1993, 1734). Thus, as she discussed, the owners of whiteness benefit from enhanced rights of access, use and enjoyment both of their racial status and of the benefits they can access on the basis of that status. They also benefit from a strong exclusionary right, which is supported by both formal and informal rules and norms (Harris Reference Harris1993, 1736–37). Of course, contemporary analyses of property rights have established that rights of exclusion are also the bluntest legal tool for exercising such privileges. While information cost theorists justify clearly demarcated ownership rights on the ground that they ‘save[…] on the transaction costs of delineating and processing information about rights in terms of uses and users’, the primary instrument for delineation of ownership is the right of exclusion (Smith Reference Smith2005, 79).
These property rights of access, use and exclusion are evident in the examples of status-based privilege that have proliferated across platform industries. Those with privileged racial and other statuses regularly enjoy products and services that perform in ways that cater to those statuses, whereas those with less privileged statuses regularly access lesser products and services. To examine this point, let us return to the example of Uber passengers. Studies have found that Uber customers perceive convenience and price to be among the greatest benefits of using Uber (Grinberg Reference Grinberg2019). It should be evident, therefore, that Uber is a more valuable service for those who have shorter wait times before cars arrive to pick them up, who have more Uber cars available to them at any given time, and who pay lower prices for their rides. Yet these values have been diminished for Uber consumers who do not have privileged racial or other statuses (Belzer and Leong Reference Belzer and Leong2017; Ge et al. Reference Ge, Knittel, MacKenzie and Zoepf2020). In these respects, whiteness and other privileged statuses have become a resource that entitles some consumers to enhanced access and use of Uber’s service and that contributes to the exclusion of other consumers from some of Uber’s benefits.
Moving now to the second step of the analysis required to understand the value of whiteness as property in the platform economy, let us focus on how the value of whiteness to platform consumers translates into property value for the platform proprietors. The short answer is that those platforms that appear to develop towards discrimination have business models that appear to take advantage of the property value of whiteness and other privileged statuses. One way in which they do this is simply by developing products and services that provide, and even promote, a feeling of exclusivity in a manner quite analogous to the exclusivity of Whites-only restaurants, clubs or other accommodations (Edelman et al. Reference Edelman, Luca and Svirsky2017). In so doing, they promote the sense of self-realisation or status accompanying the use of certain products or services (Harris Reference Harris1993, 1759–63). Another way in which platforms’ business models capture the value of whiteness is by developing their products or services to be more useful to those who are White or Anglo. Uber’s provision of convenience partly on racially responsive grounds is an example here, as is the tendency of autocorrect technologies to change non-Anglo names to Anglo ones. Another way for platform businesses to capture the value of whiteness is by having their products contribute to the negative and racist stereotyping of people of colour, which can obviously have a range of market consequences. Google’s search results for ‘Black girls’, which predominantly returned links to pornography sites, provided a vivid example of the marketisation of racial and gender identities for the presumed consumption of those with privileged status (Noble Reference Noble2018, 147–49).
The primary property form in which platform proprietors capture the value of whiteness in these varying manifestations is by means of their intellectual property. Unlike White consumers, platform businesses do not themselves benefit from being White. Rather, the algorithms of such platforms, which are owned as intellectual property, capture and expand on the value of whiteness and other privileged statuses. Intellectual property thus is another means of translating race and other forms of discrimination into property value. This particular property form is extraordinarily valuable for many reasons. Perhaps most obviously, intellectual property is the primary asset of many platform companies. Since their operations do not require many physical assets, their financial worth is largely a product of their intellectual property. And the value of that intellectual property is often astoundingly high (Brachmann and Quinn Reference Brachmann and Quinn2015; Ogier Reference Ogier2016). My point here is not that the intellectual property of platforms is more valuable because they incorporate discrimination into their business models. Rather, it is that their most valuable assets have at times included technological processes and methods that were discriminatory. In this respect, their discriminatory behaviour contributed to their wealth creation just as surely as the operation of Whites-only businesses contributed to their wealth creation.
Perhaps more significantly, by capturing whiteness and other privileged statuses in their intellectual property, platform proprietors are able to exercise rights of exclusion that can be extremely broad depending on the form of the intellectual property. This is particularly true of platforms that make use of algorithms to analyse ‘big data’:
‘[T]rade secrecy is preferable to patenting when an invention can easily be kept secret for a period of time longer than it would take other inventors to come up with the idea on their own. Many big data practices fall squarely into this category. Like Google’s Pagerank and the algorithms used by high-speed trading companies, big data practices yield commercially valuable products and services while remaining entirely out of view. A legal expert on big data at Microsoft supported this conclusion, stating that “lf [big data practices] are going to be used almost entirely internally, behind a firewall, then the company may not need or want patent protection and the disclosure it requires.” This would seem to make trade secret protection, or mere casual nondisclosure, even more attractive to big data producers than it has long been to software producers. Unlike software object code, most big data products cannot be reverse-engineered to reveal the processes that went into their creation.’ (Mattioli Reference Mattioli2014, 572–73)
Unlike patented inventions, technology that is protected by trade secrets in the US does not have to be disclosed – even to regulators.Footnote 3 Thus, the fence that trade secret protection allows platform proprietors to erect around their algorithms and other technologies is analogous to high walls and ‘no trespass’ signs around country clubs and other places of public accommodation. These are such complete barriers to access that they preclude others from even divining the ways in which discrimination is perpetrated by and through the businesses operating behind those walls (Foss-Solbrekk Reference Foss-Solbrekk2023). In these respects, part of the value to platform proprietors of whiteness as intellectual property is the value of exclusion. By capturing this particular form of property interest in privileged status, platform proprietors have the power to prevent consumers and regulators alike from scrutinising their operations for the purpose of vindicating rights, changing behaviours, or even determining the extent to which discrimination has occurred.
One final and significant way in which some platform proprietors have business models that take advantage of racial discrimination is by the extent of their reliance on workers of colour to whom they avoid providing many of the benefits of employment (Dubal Reference Dubal2023, 1939; Fairwork 2023). This divestment of responsibility to their workforce allows for a straightforward calculation whereby platform businesses are able to save on the costs of providing traditional employment-based benefits because workers of colour do not have as many options for finding more stable employment. Such a calculation is only possible because of the systemic racism in employment markets such as those pervasive in the US (Dubal Reference Dubal2023, 1939; Roithmayr Reference Roithmayr2018). It is an unrecognised property value that is possible only because of differential access on the basis of race.
While this discussion has focused on forms of property entitlements that accrue to large platform businesses such as Uber and Google, it is crucial at this juncture in the development of platform industries to consider the extent to which property interests in whiteness could exist and proliferate in peer-to-peer sharing contexts as well. Platforms that genuinely promote peer-to-peer transactions may argue that they are not employers. Such arguments ought to fail for platforms such as TaskRabbit. But they might have more legitimate traction for platforms such as Airbnb (Dyal-Chand Reference Dyal-Chand2015, 241). However, in other crucial respects, such platforms should still be regulated, and regulate themselves, to ensure that their participants are not capitalising on the values of whiteness and other privileged statuses. Here, Airbnb’s experience with race discrimination is both telling and sobering (Clemence Reference Clemence2022). It reminds us that the world from which artificial intelligence learns is a racist world (Katyal Reference Katyal2019, 62). It is therefore incumbent on such platforms to consciously design and maintain systems that avoid creating property entitlements in whiteness. Thus, one important question to ask moving forward is to what extent peer-to-peer platforms incorporate some of the same features that have contributed to discrimination, and how it might be possible to design around these features.
4. De-propertising whiteness
The foundation for optimising sharing platforms for diversity is for policy-makers explicitly to recognise the property value of whiteness and other privileged statuses in the platform economy. This is not just a rhetorical benefit. Rather, it is the basis for developing a more detailed understanding of the operation of platform businesses as a precursor to regulating them.
The recent history concerning the advocacy around, and to some extent the regulation of, the labour and employment practices of platform businesses provides a useful model here. As I have discussed, platform proprietors have thrown significant resources into efforts to skirt labour and employment regulations, claiming that they did not employ the workers who provided their goods and services. Rather, they claimed that they simply provided online spaces (platforms) for such workers to connect with customers. Despite these efforts, the legal tide has begun to shift so that now, even in the US, some lawmakers are calling platforms such as Uber: employers (Dubal Reference Dubal2023; People v. Uber Technologies, Inc. 2020; O’Connor v. Uber Technologies, Inc. 2018).Footnote 4 Labour and employment activists achieved a powerful victory when they forced some level of legal recognition of such platforms as employers. This naming function alone was an important foundation for the development of necessary regulations that recognise the employment rights of (at least some) workers in the platform economy.
If we apply this lesson in the realm of property law, it ought to be a powerful step forward for advocates and policy-makers to recognise explicitly that part of what platform proprietors own is whiteness in the multiple forms that were described in Part 2. As I have discussed, the benefits of whiteness and other privileged statuses accrue not just to platform providers, but also to more privileged consumers in the platform economy, a point to which I will return at the end of this analysis. However, from a remedial perspective, the ‘low-hanging fruit’ for purposes of advocacy and regulatory intervention concerns the benefits accruing to platform proprietors from owning whiteness. Thus, for remedial purposes, the most obvious point of intervention concerns the ownership of intellectual property in the systems that protect and perpetuate racial and other forms of discrimination, thereby allowing some platforms to capture the value of whiteness and other privileged statuses.
Continuing with the analogy with labour and employment law, then, what space might the naming of whiteness as intellectual property open up for regulation? First, such a recognition allows policy-makers to use property tools to redress the harms accruing to people of colour from the value gained by platform proprietors who leverage racial privilege. Second, such a recognition provides a path ultimately to eliminate the value of whiteness as property.
Turning first to the possibilities for using property tools to redress racial harms, an efficacious starting point would be to shift our thinking from envisioning the platform as something that is owned by the developer of the platform technology to thinking of the platform as the subject of conflicting entitlements. From a property theory perspective, we could describe this as a shift from the ‘ownership model’ wherein the platform is analogised to a finite resource the control of which is allocated to a single private party (in this case, the ‘inventor’ of the particular platform technology) to an ‘entitlement model’ wherein we recognise the platform as a resource with respect to which a range of individuals and groups (other than the ‘owner’ or ‘inventor’) have entitlements. Most importantly, such a move would require policy-makers to recognise the compelling rights of access that consumers of colour should have to equal participation in the platform economy. Currently, many platforms still operate within an opaque ‘black box’, a concept that Pasquale (Reference Pasquale2015) coined in his forceful critique of the privacy invasions regularly made by companies that collect, distribute, and use big data (Foss-Solbrekk Reference Foss-Solbrekk2023). Regulations that recognised rights of access would force such companies to change their operations in such a way as to ensure equal access to goods and services, regardless of racial and other privileged statuses, as well as to share information about the accessibility of their platforms (Foss-Solbrekk Reference Foss-Solbrekk2023).
In short, such a shift would justify – and perhaps require – a great deal more regulation, in particular to ensure access by diverse consumers. And it would also require a recognition that intellectual property rights – and particularly trade secrets – should not have the almost sacred status and level of protection that they currently have in the US. Let me briefly review a range of regulatory reform possibilities.
4.1. Thick versions of transparency
The wide-ranging calls for transparency by legal scholars and advocates in the tech space (Citron Reference Citron2008; Foss-Solbrekk Reference Foss-Solbrekk2023; Pasquale Reference Pasquale2015; Van Loo Reference Van Loo2019) are strongly supported by a property framing of differential racial treatment in the platform economy. Thus, for example, when evidence becomes available that Uber’s rides appear systemically to be priced differently for passengers of different races, it ought to be a regulatory requirement that such companies share information about their technological systems and methods of operation. As a precursor to that, these companies must also open up their operations to regulatory oversight in such a way as to allow the collection of data by regulators. Both forms of regulation are justified by a property framing that recognises equal rights of access by consumers (and regulators) rather than just protecting the exclusionary rights of platform proprietors. Under this view, title to intellectual property should not be paramount over entitlements by others to access the platforms created and maintained by such intellectual property. Because platform proprietors have thus far largely been able to hide their business systems and operations from public view, the corrective move required to protect conflicting rights of access is to require a much higher level of transparency.
4.2. Public accommodations
Leong and Belzer make a compelling argument that many platform businesses replace, or at least operate in the same markets as, physical businesses that are regulated as public accommodations. These include hotels, transportation companies, and other businesses that are open to the general public. A property framing supports the important insights and suggestions for law reform proposed by Leong and Belzer. These authors emphasise the need both for self-regulation by platform businesses and for federal legislation that would explicitly extend Title II of the US Civil Rights Act of 1964 to all platform businesses ‘that offer goods and services tantamount to those provided by public accommodations in the traditional economy’ (Belzer and Leong Reference Belzer and Leong2017, 1318).
A property framing would also add to Leong and Belzer’s prescriptions by providing a range of remedial possibilities intended to protect access by all members of the public. Currently, both judicial and scholarly interpretations of public accommodations laws as civil rights laws are regularly limited by the perception of a too narrow imperative of ensuring access only for certain protected classes. By contrast, a property framing creates literal and figurative space for upholding access entitlements of all members of the public who wish to access platforms that market themselves as available for public access and use. As I have described in my analysis of autocorrect technology (Dyal-Chand Reference Dyal-Chand2021), a property framing suggests that one of the most efficacious models for protecting access entitlements is the State of New Jersey’s insistence that all members of the public have equal rights to access public accommodations (New Jersey Annotated Statute 2020). New Jersey ensures this by prohibiting owners of public accommodations from arbitrarily excluding members of the public whether or not they are members of classes protected by federal or state civil rights statutes. Whether such broad rights of access were protected by federal statute, federal administrative rules, or state or federal court decisions, they would have the potential to radically reduce the proliferation of bias in platform technologies by alleviating the need to prove that such discrimination somehow disproportionately affected members of a protected class.
4.3. Regulation as utilities
Another possibility that goes even further down the path of recognising that some platform industries are more akin to public infrastructure than private markets is to regulate them as public utilities. Such regulation could take the form of treating some platforms as ‘essential facilities’, a possibility that Guggenberger (Reference Guggenberger2020) discusses as efficacious as a means of limiting monopoly power:
‘The platforms’ monopoly power mainly stems from network effects – that means the participation of additional users almost exponentially increases the utility of the network and creates enormous market entry barriers for potential competitors. The characteristics of data and algorithms further foreclose the markets.
To define the suitable remedies and to open the digital economy for competition, we can learn from the past. In the early twentieth century, the railroads controlled critical infrastructure and excluded competitors from crucial markets.’Footnote 5
A property framing could build on Guggenberger’s powerful analysis by regulating such platforms as essential utilities, which have been much more extensively regulated precisely in recognition of the public’s dependence on services such as electricity, gas, and water. While it may seem dramatic to argue that some platform industries are achieving the status of utility service providers in the modern economy, it is not such a great stretch. Consider the extraordinary power Zoom held over us all during the Covid pandemic, as many of us depended on this particular platform to do nothing less than facilitate our ability to earn income for a full year. Consider also the extent to which many of us rely on cloud technology to store some of our most precious documents, pictures, and information.
Finally, let us consider the regulatory challenge of completely eliminating the value of whiteness as a property form in the platform economy. Doing so requires consideration of the value of whiteness to sharing economy consumers as well as proprietors. Moreover, it requires consideration of how race discrimination proliferates in genuinely (as distinguished from nominally) peer-to-peer sharing contexts as well as platform contexts that operate as more traditional firms. This is a more difficult challenge than the regulatory challenge of recognising conflicting entitlements of access and use in the platform economy (even though that will be difficult enough as a political matter). It requires nothing less than teaching algorithms how to be non-racist. Yet this has the very same endpoint as Harris’s analysis of whiteness as property (Reference Harris1993, 1789–91). Ultimately, our regulatory aim must be to eliminate, or as Harris puts it ‘delegitimate’, the property interest in whiteness. While length limitations preclude me from achieving depth or detail in developing a regulatory plan for how to do this, I do want to propose two foundational requirements for beginning this important work.
4.4. Optimise for diversity
The first requirement is obvious, yet essential. Regulators must prioritise diversity in the design and the regulation of sharing platforms. Doing so requires regulators to mandate the optimisation of diversity as a direct goal, rather than an indirect benefit or outcome of platform regulation. It also, of course, requires platform proprietors to prioritise a ‘double bottom line’ of optimising not just for profit (which currently appears to translate into an effort to optimise for efficiency), but also for diversity.
Thus, when a platform proprietor develops a feature such as ratings by Uber drivers of their passengers, or by Airbnb hosts of their guests, such a feature should not be operationalised until the platform proprietor reviews its effect on maintaining equal access and diversity within the platform’s operation. Indeed, given the incentives for platform proprietors and their financing sources to focus first and foremost on profit, it is incumbent upon regulators to demand information about the development and planned use of such features and to undertake such a review themselves.
While there are many regulatory possibilities for prioritising the optimisation of diversity, one recent regulatory innovation in the US has been legislation requiring businesses such as real estate developers to submit racial impact statements describing the anticipated effects on racial minorities of their proposed property development.Footnote 6 Such impact statements are modelled on the well-established environmental impact statements that are required under both federal law and the law of many states (Hing et al. Reference Hing, Kennedy and Sonnad2013). Among their many likely benefits, perhaps the most important is that they require businesses and regulators to think hard about the short- and long-term effects of their planned activities on communities and individuals of colour. In this respect, racial impact statements are a powerful mechanism for thinking hard about structural racism, and they are also easily translatable to the realm of online and platform businesses.
4.5 Democratise the platform
The second requirement is to continue to look for ways to democratise platform governance. This is a more generalised version of the theme that runs throughout the regulatory reforms I have proposed. The calls for transparency, access, and power de-concentration all have at their core the impulse towards democratic governance. Until very recently, thanks partly to their innovations intended to avoid regulation (Pollman and Barry Reference Pollman and Barry2017), platforms have operated under the opposite regime from one of democratic governance. They have taken advantage of regulatory voids, thereby avoiding consumer safety, employment and other crucial regulations intended to protect those parties whom the platforms need in order to profit, but whom they have had no incentive to protect.
By democratising governance of platforms, regulators should consciously seek reforms that ensure a more diverse representation of those who have oversight and the power to dictate operational changes for platforms. Diversity in the workforce within regulatory agencies will help, as will efforts to diversify the workforce of the platform businesses themselves. In describing democratic governance as a foundational requirement, I am also relying on the assumption that the ‘crowd’ will be more adept at protecting and promoting diversity and broad access than the platform proprietors, when left to their own devices, will ever be capable of achieving. Furthermore, and crucially, there is every reason to believe that democratic governance requirements will achieve a great deal more equity for workers and for consumers involved with platforms.
Here again, there is a broad range of regulatory reform possibilities. In my analysis of the whiteness bias in autocorrect technology (Dyal-Chand, Reference Dyal-Chand2021), I discussed the relevance of Yochai Benkler’s analysis of open source and crowdsourced platforms such as Wikipedia (Reference Benkler2016, 27; Benkler et al. Reference Benkler, Hill, Shaw, Malone and Bernstein2015, 189). My argument that the crowd can be expected to protect diversity in autocorrect systems is equally applicable to a very broad range of platforms, including those that are hierarchically managed as well as peer-to-peer platforms. Mechanisms such as relational contracting and advisory boards also could recognise the value of peer-to-peer regulation that is still informed – and monitored – by regulators (Dyal-Chand Reference Dyal-Chand2015). Such mechanisms also seem well suited to the task of ensuring the continued prioritisation of diversity.
The possibilities for regulatory reform are both extensive and highly pragmatic. Neither technology nor regulatory complexity are barriers to most of the regulatory innovations I have described. Instead, what is required is the recognition that a meaningful part of the value of platforms for their proprietors – and some of their more privileged consumers – is the value of whiteness and other privileged statuses. Simply put, achieving profit from this particular form of property is unacceptable. It should be the highest priority of law, and indeed, of our society, to eliminate the sharing of whiteness.
Competing interests
None.