1. Introduction
The world we live and work in today is one dominated by large technological companies. Work, too, is mediated by new technology, which in reality means that it is mediated by Big Tech. Even when one doesn’t work directly for a Big Tech company, one is likely to be part of the increasing numbers of industries that are now fulfilling the role of subcontractors to Big Tech. Such Big Tech-mediated work includes piece rate workers who manually categorize and annotate images for artificial intelligence (AI) algorithms. It also includes workers in small businesses that primarily operate on Big Tech platforms, such as those who work for brands that sell on Amazon.
Gig workers are a large and expanding category whose work is controlled by big capital and technology. Even workers in industries that do not come directly under the ambit of Big Tech interface with it. Office workers who use enterprise software or street hawkers who use digital payment tools are part of this category. Its vast capital reserves also enable Big Tech to influence government policy related to labor in most parts of the world. In short, whether directly or indirectly, workers are increasingly affected by the existence and growth of Big Tech, and are compelled to respond to its encroachment on their lives.
On a global scale, the shifting dynamics between labor and capital under digitalization can be understood as a change in the relations between countries, and between classes in countries, where technological development and global economic relations shape each other. Such an analysis is mainly macroeconomic in nature — it seeks to note the big shifts and contradictions. This essay takes a microeconomic perspective, analyzing instead how work in the digital economy functions for the individual worker, a type of worker, the individual firm, and a given trade union. Both types of analyses are crucial not only to arrive at a complete picture of work in the digital economy, but also to understand how to resolve contradictions in the digital economy.
The rise of Big Tech has contributed to automation, which has implications for the number and quality of jobs available. This essay does not make the automation of jobs its primary focus, although it does consider it. This essay is more concerned about the set of choices available to a worker or a group of workers while negotiating with capital. The interests of labor and capital are for the most part divergent; in a manner of speaking, both negotiate with each other to arrive at a given wage rate, benefit level, and working hours. In this context, negotiation also involves strikes, lock-outs, the use of violence and the state’s coercive power, and so on. Extraneous factors such as the rate of unemployment, the orientation of the state, the prevalence of unionization, and the level of technological development affect the relative positions of labor and capital. This essay explores how labor’s ability to negotiate with capital has changed, by examining how specific actions have become more difficult with the rise of Big Tech in particular and with digitalization in general. The following two sections deal with this question, as well as the question of what should ideally be easier to negotiate for labor with current and emerging technology. By the end of the essay, conclusions from both sections are used to suggest a substantive agenda for labor.
2. How Does Big Tech Diminish the Negotiating Power of Labor over Capital?
This section explores some of the ways in which work for Big Tech and work mediated by Big Tech has led to a change in the balance of power between labor and capital in favor of the latter. Each category of negotiation builds from an example of a labor struggle under digitalization.
2.1. Algorithm and Data-Related Issues
In August and September 2020, a group of Swiggy food delivery workers in Delhi held a “log-out strike” to protest the company’s reduction in delivery fees as well as other labor issues. Among these other issues, one revolved around Swiggy’s claim that its algorithms provided incentives to riders for completing a certain number of orders, even as riders alleged that the app intentionally manipulated order flows such that no one would be able to complete the required number of orders. The incentive was nevertheless advertised as part of potential pay for riders. Additionally, workers complained that the customer rating system was not transparent and thus open to manipulation. Some algorithmic choices made riders’ lives difficult. For instance, at one point, the app arbitrarily made it mandatory for the rider to send a video of rain in order to earn an incentive for delivering food in inclement weather, rather than relying on publicly available weather information. 1This case is drawn from the author’s personal experience covering the strike.The protesting workers were unable to contact any executive from Swiggy to collectively register their grievances and demands. As algorithms and automated processes increasingly determine the nature and conditions of work, the ends to which these are consciously and unconsciously put play a part in determining labor choices. Algorithms often work opaquely, especially if they involve machine learning or other similar techniques. This means that while a machine learning algorithm can recognize patterns, it cannot always explain how it determined that there was a pattern. For instance, a machine learning algorithm at a ride-hailing company can assign rides to optimize for cost and time. However, a worker who feels aggrieved by this algorithm, perceiving it to be discriminatory, may have no way of knowing how the algorithm makes its decisions, and thus of ascertaining whether its working is indeed discriminatory. The potential regulatory workarounds to this issue are explored in the final section.
Even when machine learning techniques are not involved and the algorithms are simpler, their parameters are hidden from the view of workers. When customers rate a delivery worker’s performance, the rating algorithm likely performs a simple aggregation function to arrive at the worker’s overall rating. However, the worker has no way of knowing whether the algorithm works as intended, whether it has included all ratings, and how each rating is weighted. Regulations on algorithmic transparency can go a long way to correct this information asymmetry.
Algorithms can impede workers’ autonomy to perform tasks by using their own best judgment. Workers in call centers see their flexibility decreasing with granular monitoring and algorithmically-determined targets for each customer service request. Some text-based customer service workers are forced to respond mechanically to two chats at once, decreasing their ability to solve either customer’s problem. Until recently, drivers working for the ride-hailing company Ola were unable to see the drop location of customers before the ride started, leaving them without crucial information as they decided whether to accept or reject a ride. This led to a surge in the number of ride cancellations by drivers. Ultimately, customer irateness forced the company to make this information known, albeit approximately. Truck drivers in AI-enabled ports are unable to determine their own routes and have to follow routes that have been algorithmically determined to the smallest detail. The autonomy they lose as a result of this enforced dependence on algorithms is also accompanied by a loss of skills (such as the knowledge of routes around a city). A limitation of skilling also occurs in the information technology (IT) industry, where some employees work for years exclusively on a proprietary software and consequently have specialized but limited skills, thus restricting their job mobility. Algorithm-driven management can also restrict workers’ ability to communicate with company management or even an immediate superior. It is not uncommon for all worker interactions, from onboarding to termination, to be undertaken via an app with no human interface. This means that workers have no room to negotiate around their problems at all. This is true for both small problems, such as an incorrect payment, to larger issues, such as a group of workers unhappy with a rate change. The pandemic saw some platforms laying off even the few interfacing professionals that existed, leaving gig workers with no recourse in the form of communication. Strikes, like the ones at Swiggy, have seen “bouncers” and other enforcers of the platform showing up, but no communication from the management. The algorithm, in such cases, functions as a stand-in for management, one that cannot be reasoned or bargained with.
The algorithms used to assess workers are at times developed externally. These cannot be tweaked even by company management, meaning that responsibility is shifted to the external provider in case of errors or disagreement. This includes assessment tools such as those bundled with Microsoft Office software. Company management may not be aware of, and may not consider it worthwhile to be aware of, the exact mechanisms of how these assessment systems work.
Algorithm-driven management can restrict workers’ ability to communicate with company management or even an immediate superior. It is not uncommon for all worker interactions, from onboarding to termination, to be undertaken via an app with no human interface. This means that workers have no room to negotiate around their problems at all.
2.2. Disconnecting from Work
In March 2021, a group of first-year analysts at Goldman Sachs threatened to quit the organization due to long working hours stretching up to 105 hours a week. The workers said that remote working during the pandemic had made the situation worse. The Goldman Sachs CEO admitted that the pressure to be “connected 24/7” contributed to the grievances. While investment banking work at junior levels has always seen inordinately long hours, this outcry in 2021 mirrored what other remote workers were going through during the pandemic as work stretched into home.
The eight-hour workday, won after much bloodshed and struggle, is quietly retreating in the era of Big Tech. Particularly for service professionals, digital technology has enabled work to extend into the home and therefore into non-work hours. Work has also extended into the commute to work. In addition, the rapid uptake of educational technology during the pandemic has had a deleterious effect on working parents, whose burdens increased with children learning from home. It has also dramatically increased work hours for teachers, who now need to prepare longer for classes and undertake more teaching activities than before.
To be sure, remote working, in general, has allowed for flexibility in work timings, reduced commute times, and reduced monitoring (in some cases). The flexibility in work hours has been particularly beneficial to working parents and caregivers. However, the inability to disconnect from work, greatly exacerbated during the pandemic, is also linked to serious and even terminal health effects for workers. In fact, Penn sees the right to disconnect as an example of “algorithmic silence” —– a much-needed conscious relief from digital decision-making that keeps the system in check, much like silences make music operative.
Even prior to the pandemic, European and South American countries had been at the forefront of enacting laws that prevent employers from contacting employees outside work hours. By making remote work far more ubiquitous and further increasing working hours, the pandemic has simply reiterated the importance of the right to disconnect from work.
2.3. Unionization
In 2016, UberEats delivery workers in London held a protest against a unilateral reduction in pay. Workers were responding to the food delivery company’s decision to reduce rates two months after the launch of its services in the U.K., just as many workers had switched to UberEats for its higher rates. Lacking a workplace or common meeting spaces, workers used an innovative method to unionize: they posed as customers by ordering food on the app, and spread information about the union when another delivery worker came to deliver their order.
The move from factory floors and offices to decentralized or distributed workplaces has been detrimental to trade union organizing in general. Under the constant surveillance of the employer, workers are unable to talk to one another and organize. The use of digital tools for union-related communication are often restricted. For instance, in the United States, companies are allowed to ban the use of work email addresses for union organizing. In the gig economy, company surveillance extends to digital tools, such as WhatsApp groups, used by workers to organize or even converse about work. In 2020, Amazon was found to be monitoring closed Facebook groups meant for contractor activism.
Even when workers are present together in one location, they are under constant digital surveillance. Big Tech companies use worker profiling to predict the possibility of union organizing, for instance, by tracking ID badges to see which workers meet their colleagues and for how long. Amazon uses software to analyze large events occurring outside the gates of its facilities, including employee gatherings. Warehouses are closely surveilled for any unaccounted break time to the last second. Store locations are heat mapped according to likelihood of union-related conversations.
A consequence of such surveillance is that employers are able to delineate productive time from non-productive time for workers, although both are necessary for a job to be performed. Some analysts believe that such delineation can lead to further gig-ification of the economy, as employers can decide to pay workers only for productive time — for instance, paying retail workers only when customers are inside a shop — much like Uber drivers are paid only per ride, and not per day of work. This will result in a decrease in pay for workers without a corresponding decrease in work.
In some contexts, however, digitalization has improved unionization, primarily through the prevalence of online conferences and digital membership drives. It is also sometimes aided by the use of social media tools. Workers’ rights centers in China, which provide free legal and other aid to migrant workers in the country, are a good example. The centers have gone from providing aid through telephone hotlines to extensively using social media. Social media tools are used by the centers for both direct consultations and mass dissemination of information, and this use is contributing to the organizing of labor as well.
Inherent centralizing tendencies still exist in the supposedly decentralized gig economy, also sometimes aiding unionization. For instance, there has been a consolidation of restaurants into cloud kitchens that are optimized only for delivery. These cloud kitchens, which are spaces where delivery workers often gather and have conversations, have also become sites of recent workers’ strikes and demonstrations.
2.4. Life Outside Work
In February 2022, the German state-owned broadcaster Deutsche Welle terminated the employment of at least seven journalists for writing pro-Palestinian posts on their personal social media profiles. The journalists were not given an opportunity to explain their posts or otherwise make their case. Similarly, in May 2021, a journalist was fired from the Associated Press for making pro-Palestinian comments on her personal social media profile. These journalists joined other workers who lost their jobs because they posted their opinions on Palestine and Israel on social media. Quite separate from geopolitical events, these cases demonstrate the technology-enabled tendency and ability of employers to monitor employees’ personal lives and impose costs on them for the same.
Employers have an increasing array of tools to carry out such monitoring. Besides the forms of social media monitoring discussed earlier, corporate surveillance products now include software to “help employees stay focused” by monitoring their social media usage. One survey revealed that 46% of employers in 2018 used non-traditional methods to monitor employees, even as another pointed out that 72% of employees objected to such forms of monitoring. There have been instances in the U.S. where employees have been fired after declining to share their social media accounts with employers.
Health technology linked to employer-provided health insurance can also impact workers’ life outside work. Health insurance premiums are, in many cases, linked to metrics generated through technologies such as activity-tracking wearables, including smartwatches, fitness apps, and blood sugar monitors. Many health technology companies offer corporate programs that are based on employee monitoring. 2 Smitha Krishna Prasad. Health Tracking Technologies: Privacy and Public Health. Paper presented at a workshop on Public Health and Privacy in New Delhi, organized by Smriti Parsheera and Thakur Foundation. October 2021. Advertised as voluntary, these programs are in reality often forced on employees. They intrude into workers’ personal lives and compel them to live in a way that reduces insurance premiums for employers. There are also possibilities of health data being used by employers for unrelated and illegitimate purposes. In general, the involvement of Big Tech in worker monitoring has tied workers’ lives outside work more firmly to employers’ wishes. This has reduced the space for autonomy and diluted the idea that non-work time is a worker’s right.
The involvement of Big Tech in worker monitoring has tied workers’ lives outside work more firmly to employers’ wishes. This has reduced the space for autonomy and diluted the idea that non-work time is a worker’s right.
3. How Technology can Increase the Negotiating Power of Labor over Capital
All economic and social developments that impede the autonomy and rights of people also contain seeds that allow new forms of resistance to flourish. Technological progress is contextual, and a change in social relations can change the ends to which technology is put. Even a small increase in the bargaining power of labor can open opportunities for alternative uses of technology within existing social relations
3.1. Access to Data
Workers subjected to close surveillance can use surveillance data to further their own goals. This data can be used by workers individually and collectively to demonstrate, for instance, that they are being overworked, that their work targets are unrealistic, or that some workers are being discriminated against.
Currently, workers receive little support on this front from labor laws which do not seem to be in tandem with the reality of increased data generation at the workplace. A case in point is Indian labor law, which is regressing when it comes to data rights — new provisions make data less, not more, accessible to workers. For example, workers have effectively lost the right to check the allocable surplus of their employer to determine the amount of bonus they are entitled to. What was a right in the analog world is no longer a right in the digital world.
For gig workers, access to their own data can not only help seek redress in case of exploitation or discrimination by employers, but can also determine future work opportunities. With the ability to port their data in a prescribed format from one employer to another, gig workers would be able to demonstrate their experience and expertise. There is a good case to be made that the non-portability of this work data is anti-competitive in addition to being detrimental to workers’ interests.
Gig workers can also use their work data to make the case for formalization of gig work. Data from platforms can show that many gig workers work full time as platform policies leave them with few other alternatives. Such facts can go a long way in proving in court that gig workers should have all the rights and benefits accruing to permanent workers, given that they de facto perform permanent work. Currently, only survey data points to these realities. For instance, a survey from Indonesia shows that most app-based delivery workers work between 12 to 18 hours a day, without overtime pay.
Today, data collection and use is a one-way street – employers collect data about employees and use it for their own interests, and the opposite rarely occurs. There is almost never any transparency on the types of data collected, the period of retention, and the purpose of collection. Challenging this status quo can help labor negotiate from a more advantageous position due to increased knowledge and proof about their own collective working conditions.
Today, data collection and use is a one-way street – employers collect data about employees and use it for their own interests, and the opposite rarely occurs. There is almost never any transparency on the types of data collected, the period of retention, and the purpose of collection. Challenging this status quo can help labor negotiate from a more advantageous position due to increased knowledge and proof about their own collective working conditions.
3.2. Real-time Finance
The digitalization of finance has made real-time transactions possible and increased the overall efficiency of financial transactions. This should mean that workers (especially contract workers) are able to receive payment in real time, and access other financial products that benefit them. In fact, the opposite is true. Gig workers receive payments with a significant delay even as platforms discourage customers from making cash payments. Ride-hailing companies are particularly notorious for delaying drivers’ payments. This arbitrage using digital payments is profitable for digital platforms — even a week-long delay in payments of millions of drivers means credit-related earnings for the platform. In effect, drivers subsidize platforms by receiving their payments with a delay, a practice that did not exist prior to the platform’s mediation. It is no surprise that gig workers have some of the fewest financial options in the U.S.
Other financial products, including loans and insurance, ought to be more easily available to workers in a way that improves their safety net outside work and thereby increases their autonomy. Public financial infrastructure, including data aggregators, bank accounts, and a framework for digital payments, can go a long way in encouraging the development of these products. The dangers of financialization include over-lending and debt traps, but with regulation to minimize these, more financial options will likely be beneficial for workers.
The above benefits have so far not accrued to labor due to a variety of factors, including the nascency of these innovations, the lack of organized labor in many industries, and the scant demonstrations of benefits of these interventions. However, if labor organizations and others prioritize these agendas in their technology-related programs, there is a good chance of these innovations benefiting labor.
The above benefits have so far not accrued to labor due to a variety of factors, including the nascency of these innovations, the lack of organized labor in many industries, and the scant demonstrations of benefits of these interventions. However, if labor organizations and others prioritize these agendas in their technology-related programs, there is a good chance of these innovations benefiting labor.
4. A Substantive Agenda for Labor
Given the shifts in the negotiating power of labor and the emancipatory possibilities of technological progress, what should labor demand in the near and long term? This section outlines a few key demands at the regulatory level.
- Regulation of algorithms: Algorithmic governance of work reduces labor’s bargaining capacity in relation to capital. Accordingly, an important agenda ought to be enlisting the state to regulate algorithms. The ways of doing so include but are not limited to:
a). Mandating that algorithms have a minimum level of explainability. This involves clarifying not only the internal workings of the model, but also the extraneous factors such as the choice of data, the methods used, and so on. Different degrees of explainability can be incorporated into models during the design or even testing stages. In general, this requires more engineering resources on the part of the employer.
b). Adopting algorithmic standards that set minimum levels of performance for an algorithm. Such standards can include the stipulation that the error rate be lower than a specified level. They can also be related to algorithmic transparency, the disclosure of possible risks, etc. An example of a standard is the U.K. government’s Algorithmic Transparency Standard, which covers government agencies, and prescribes, inter alia, how data about external suppliers needs to be collected and coded. Workplace safety norms should also take algorithmic standards into account, especially when errors affect safety. For example, a facial recognition software that is used to allow workers entry into and exit from a sensitive laboratory must specify a maximum error rate.
c). Mandating human involvement in decision-making. This can ensure that some needs do not fall through the cracks due to technical errors, and that sensitive decisions are taken with a human understanding. The European Union’s General Data Protection Regulation (GDPR) requires humans to be in the loop of decision-making in some cases. - Increased data sharing: A general response to datafication has been to insist on privacy measures, including data minimization measures. Data minimization is a principle according to which only that data is collected as is necessary to fulfill a specific purpose. It can limit privacy harms at the point of collection.
However, the mere minimization of data collection cannot solve problems that are not related to individual privacy. Data can also be incredibly useful to workers and society in general, as explained above. Currently, however, this data is captured in private enclosures, preventing the realization of its full value. Such enclosures also perpetuate economic inequality between large technology companies and others. Education International, a global teachers’ union, has pointed out that educational technology offerings turn schools into conduits for making large companies wealthier on the backs of students’ data. This is because valuable student learning data is not shared with students, parents, or even schools, but is instead retained only by the educational technology firm which uses it to improve its product offerings.
Platform cooperatives also face the distinct disadvantage of not possessing the data that their privately-run competitors possess. This disadvantage would not exist if mandatory data sharing agreements with workers were enforced. Workers could then decide to switch employers or start their own platform using this data.
To bridge these data gaps, workers and their representatives must demand data sharing for workers. Going a step further, they must demand knowledge generation on the terms of labor. This means that rather than minimizing the collection of data, it would be more appropriate to reorient it towards goals that are worker- and society-friendly. Privacy laws, too, must be in tandem with these demands for collective rights to data. When occupational safety laws mandate that private enterprises must collect data and file reports about their safety measures, this is data collection on the terms of labor. Similarly, companies can also be mandated to collect, analyze, and share big data to protect workers’ rights in other spheres. - Remote work rights: There are indications that remote work might outlast the pandemic. With more workers being given either the option or the mandate to work from home, new kinds of workers’ rights take primacy. These must include the right to disconnect from work as well as the right to be paid for work equipment, including the cost of internet, laptops, and chairs. Remote work has led to significant cost savings for some employers, which means that at least some of these saved costs are being borne by workers. Workers’ demands must be updated to reflect these hidden costs.
- Financial rights: Workers must make financial demands of both the state and employers. The state should make financial interoperability a reality through regulation and public frameworks. An account aggregator system, like the one set up by the Reserve Bank of India, or an open banking system for data sharing, such as the one mandated by the UK, are examples of such mandates and infrastructure that increase financial choices. The account aggregator framework facilitates data transfer from financial providers (such as a bank with whom a person has an account) and financial information users (such as a loan company that is assessing the person’s creditworthiness). The facilitation of this information transfer minimizes the information asymmetry in favor of people for whom information is collateral, which is to say, for a majority of the working population. Workers must demand that employers use the benefits of real-time finance in favor of labor, with instantaneous, cash-like payments through digital means. Transparency of financial transactions is also made easier through digitalization. It is not a significant cost addition for a gig work platform to display the commission it charges, the incentives it applies, the change in rates over time and geography, how price changes are determined, and so on. Labor must demand financial transparency from employers. Governments and/or courts must retain the right to demand access to source codes to understand and verify this information when required.
Finally, labor must demand that the government create and use digital financial tools to ensure timely delivery of social security and other welfare payments. Digital verification systems for financial accounts can ensure widespread banking services to the hitherto unbanked population while also increasing the capacity of the state to deliver welfare. - Emancipatory automation: Since the advent of industrialization, automation has, more often than not, meant job losses and pay cuts for workers as an immediate consequence. This explains unions’ historical suspicion of automation in work. However, this need not be the case. Automation should mean less drudgery and fewer working hours overall. While these gains do not automatically accrue to workers in the current global economic order, aiming for them can secure at least a partial achievement.
To this end, all automation must be accompanied by corresponding labor benefits, which can be substantial in the era of digitalization. A true assessment of the gains of automation can sometimes only be made at a global scale due to global digital value chains. For instance, automating one part of a food processing business in Brazil can have implications for farm workers in Malaysia and retail workers in Europe. Global labor coordination can thus ensure a fair redistribution of the gains of automation. However, even in the absence of such coordination, assessments about the costs and returns of automation can aid unionized workers in a specific area in demanding benefits in the event of automation.
Examples of attempts at constructing emancipatory automation can help us visualize the concept better. In Salvador Allende’s Chile, Project Cybersyn attempted to automate the planning of economic activities in the growing socialized sector. It relied on data collection and modeling to predict the optimal patterns of production for an economy geared towards the fulfillment of needs and wants rather than the maximization of profit. Automation can assist in making dehumanizing tasks redundant, including the cleaning of sewers. This is an activity that is, to date, relegated to the oppressed castes in India, and used to keep many members of these castes subjugated. Some innovations have been able to automate this task. Their limited use illustrates that social conditions and movements are required for automation to achieve its emancipatory potential.
5. Conclusion
Technological change can have a variety of effects on workers, depending on their existing relationship with capital and their ability to wield these changes for their own interests. In most parts of the world, working directly as part of the digital economy is prompted by disappearing work opportunities outside the digital paradigm. This means that workers entering the digital workforce do so at a disadvantage, that is, with reduced bargaining power due to fewer alternative work opportunities. The rampant exploitation in the gig economy is testament to this unequal bargaining power between workers and employers.
Even for workers who do not directly contribute to the digital economy, the mediation of technology through Big Tech ensures that it molds their work and personal lives. Hard-won rights related to working hours and unionization are once again under threat due to the manner in which digital technology has affected work arrangements in the current era, particularly during the pandemic.
Labor, and people interested in the emancipation of labor, can still be assured that technology does not only arrive with downsides. Social media and other digital communication tools, while being heavily monitored and censored by employers, are still used extensively by workers to organize as well as publicize their issues. Such communications technology allows for easier national, regional, and international solidarity building as well. Nascent efforts, such as the global Make Amazon Pay campaign, demonstrate the potential of global action against multinational behemoths. However, such global action can only be potent when backed by robust local organizing.
Social media and other digital communication tools, while being heavily monitored and censored by employers, are still used extensively by workers to organize as well as publicize their issues.Nascent efforts, such as the global Make Amazon Pay campaign, demonstrate their potential. However, such global action can only be potent when backed by robust local organizing.
Technological development also means that drudgery can be eliminated in the long term, and in the short term, financial and data-related rights can be won to help workers increase their bargaining power with employers.
While labor representatives focus on these struggles, regulators must focus on two aspects. The first is to improve their ability to regulate emerging sectors and new manifestations of old labor issues, such as the need to delineate a fixed number of working hours, specify a minimum wage, and prevent workplace discrimination. The second is to creatively tackle new labor issues created due to digitalization, such as liability for algorithmic decision-making, data ownership disputes, and surveillance in the context of remote work.
With the right confluence of government and labor intervention, digital technology can be used towards the welfare of workers and society at large. In other words, its detrimental effects must not be taken as inevitable; they point us to present openings for future solutions.
