Selbst et al. (2019) offer a list of traps that one should keep in mind when framing a problem for analysis. These traps are sometimes easily fallen into, and it is important to call them out explicitly when analyzing any system. These can serve as checks on our own thinking when we go about the process of framing and abstracting a real-world system.

We illustrate these concepts with concrete examples of technological systems that illustrate what it looks like when we fall into one of these traps:

• Formalism trap: Failure to account for the full range of meanings of social concepts such as fairness, which can be procedural, contextual, and contestable, and cannot be resolved through mathematical formalisms.
  • Example: Although fitness-tracking apps claim to give users power over their own health through objective data, critics have argued that fitness-tracking enthusiasts tend to reduce the holistic concept of health to a combination of proxy measures like calorie count (Sharon and Zandbergen 2017).
• Framing trap: Failure to model the entire system over which a social criterion, such as fairness, will be enforced.
  • Example: A machine learning system, used to help decide which pneumonia patients to admit to the hospital, learned a dangerously misleading rule that people with asthma were less likely to die of pneumonia; this was true in the data set, but only because of the unmodeled factor that doctors had assigned more intensive treatment to the data set’s asthmatic patients (Caruana et al. 2015).
• Portability trap: Failure to understand how repurposing technological solutions designed for one social context may be misleading, inaccurate, or otherwise do harm when applied to a different context.
  • Example: Publicly available computer vision systems trained with affluent Western households in mind were far less accurate when identifying objects in lower-income or non-Western households (DeVries et al. 2019).
• Ripple effect trap: Failure to understand how the insertion of technology into an existing social system changes the behaviors and embedded values of the pre-existing system.
  • Example: A tool that recommends criminal sentences based on likelihood of reoffending may warp the justice system to more heavily emphasize one value (incapacitation) and diminish others (deterrence, rehabilitation, etc.) that originally shaped a sentencing decision (Selbst et al. 2019).
• Solutionism trap: Failure to recognize the possibility that the best solution to a problem may not involve the particular technology you are currently working with or working to develop.
  • Example: A designer, working on a high-tech project that used location tracking technology to help people with sight loss to navigate train stations, pointed out that painting high-contrast lines on the ground would avert the severe accuracy limitations of the location tracking technology while also being helpful to the majority of blind users who have some residual vision (Yates 2018).

In this section we offer four questions that one should keep in mind when analyzing or discussing technology in a social context. Although these are not the only questions you will ever need, they are important starting points. We will use each of these questions when we turn to examining and analyzing specific social spheres in section 5.5.

One pitfall to avoid when engaging with the design, development, or deployment of technology is believing that if we think really hard, we can anticipate all issues that may arise from a technological intervention all by ourselves. However, we all have our own biases and preconceptions, and these will necessarily limit the issues that we are able to conceive of, no matter how open-minded or smart we think we are (McIntosh 1988).

As we saw with the tomato harvester example in section 5.3, it is critical to attend to the questions of inclusion. Who has access to discussions about the development and deployment of any technological intervention? And how are these concerns understood and incorporated into the technology development process? When discussing our spheres for analysis in section 5.5.1, we use the term pluralities for this collection of concerns around who has access to the conversations about the development, design, deployment, and oversight of a technological system, and whose input is valued in that process.

There are many ways to address these issues, and perhaps the most straightforward is to include as many affected people as possible in the entire process of technology conception, design, development, testing, and deployment. One of the most important lessons from the scholarship on participatory design (M. K. Lee et al. 2019; Ricks and Surman 2020) is that we must acknowledge that technological systems are situated in a broad social context, and we must engage all who are affected in a cocreation relationship (Kretzmann and McKnight 1996), whereby as many communities and stakeholders are included as possible.

Embedded in a concern for pluralities is a concern for epistemic asymmetries: that is, attending to situations in which some kinds of knowledge claims, or some people’s claims to knowledge, are systematically undermined or de-legitimized, in ways that are treated as natural but which can be explained by present and historical power imbalances and powerful entities’ desire to maintain power. For a concrete example of epistemic asymmetry, see the discussion of cultures of knowledge in chapter 3, particularly the closing example about how Indigenous knowledge of the environment is rarely taken seriously by Western-trained scientists. You can also look back to the discussion of feminist ethics in chapter 2, an approach that specifically aims to redress the injustices caused by the exclusion of so many people’s knowledge and voices. Epistemic asymmetries can lead to an inequality of access for self advocacy. For instance, in thinking about access to medical interventions and care, who has access to different types of care? Who gets to decide what is appropriate and what is not? Is everyone on equal footing when it comes to deciding what is the best care and who receives it?

Technology has the potential to magnify the asymmetric effects on different groups of people. For example, when cities create open data sets of policing and crime, it can reinforce ideas about which neighborhoods are “good” or “bad.” As we discussed in chapter 3, it is important to remember that any choice of data and what to record represents a specific framing choice and a particular angle on the world. This is not an abstract concern. Consider the large, linked databases on services for the poor, including those about who is seeking care (e.g., addiction therapy). A record of such care may disqualify someone from public benefits, and therefore many people in need seek off-the-record solutions (Eubanks 2018). It is important to consider. how technological systems create or reinforce norms both explicitly and implicitly.

As we discussed in the first part of this chapter, it is often impossible to analyze a technological system as some separate entity from the context in which it is used. This idea of interconnectedness, that everything affects everything else, is an important notion to keep in mind when discussing or analyzing technological systems. This concern reinforces the ripple effect trap that we discuss in section 5.4.3: there are risks associated with our always-limited understanding of the interconnections between a technological system and the broader environment of where it is implemented or deployed. Whereas this effect can be mitigated, it can never be averted entirely—there will always be more layers than we can anticipate.

Examining a decision in a vacuum can tell us only so much; we consider the interconnections between parts of a system to go beyond those parts that are changed deliberately and discover those other parts that might also change in response. This includes the interplay of different granularities (e.g., how the collective level affects the individual level and how the individual level affects the collective level) and different kinds of elements of the system. One approach that can be useful for capturing the complexity of various interconnections is known as actor-network theory.

Actor-network theory (ANT) is a methodology developed in the 1980s by scholars (most notably Bruno Latour, Michel Callon, and John Law) working at the intersection of science, sociology, and technology studies. ANT is known especially for the novel way in which it specifies that actors—both human and nonhuman, individuals as well as collectives—influence and interact within a complex sociotechnical system (Latour 2005). Ecosystems, ideas, organizations, and more are viewed as a dynamic web of actors alongside people when considering the effects of a technological intervention (Law 2009).

One of the traps we listed above is the solutionism trap. This trap can also lead to a mind-set that is sometimes referred to as technological solutionism. Technological solutionism is a mind-set that limits creativity and limits critical thinking. It presupposes, first, that all problems can and should be fixed; and second, that the best way to fix a problem is to invent new technology or further devleop the technologies we have. Operating from this stance, one can find it hard or even impossible to recognize that sometimes the best solution may not be a new piece of technology but rather some other intervention, or perhaps no intervention at all (Jasanoff 2016; Selbst et al. 2019).

A limited mind-set can also be at work in the way we anticpate responses to technological change. New interventions into current technological systems can stimulate responses that are unexpected and surprising. For example, widespread adoption of teleworking technology has meant that individuals are no longer tethered to living near their working location, which has resulted in many individuals taking actions like moving to more remote locations or leaving cities. The response effect is closely bound to the idea of interconnections, in that one technological intervention that may lead to ripple effects, but also to feedback: the use of more robots on a factory floor may make job retraining more available in a given location, or it may just cause increased unemployment.

An individual’s or group’s reactions to technology can be used as a social signifier: in other words, a way of telling others something about their social role or status, whether through their embrace of a technology or their resistance to it. For instance, when Google Glass was released, many lower-income people saw it as a symbol of class elitism, and some differentiated themselves by displaying hostility toward wearers (Noble and Roberts 2016). This kind of response can be understood if we think about how the use or nonuse of particular technologies gets translated into social capital by signaling virtue, class, or status; some people will go out of their way to reject “improvements” that a technology creates, even if the improvements themselves are something they want, because rejecting that technology helps them claim a particular social identity (Bourdieu 2018).

Medical care is perhaps the most obvious system of care that most, if not all, of us will have encountered in our lives. In this section we discuss some of the ways technology is changing the conversation around care, and also the conversations about the technological developments that may be important in the future. Advances in technology have massive effects not only on the cost of care but also on what kind of care is actually available. Modern technology can replace limbs and restore sight in ways that were the provenance of science fiction only decades ago. It is also useful to look at less dramatic kinds of medical care, such as eldercare and care for those with physical and cognitive impairments. Even the integration of different parts of the medical system can itself be considered a structure of care. Looking closely as these common, widespread forms of care can help us attend to fundamental questions of how we care for the most vulnerable among us and provide some relief and freedom for those people to whom the caregiving falls. In this section we first focus on eldercare and then zoom out to some issues of medical care writ large.

Technologies for older and debilitated populations focus on a number of areas: dependent living, fall risk, chronic disease, dementia, social isolation, depression, poor well-being, and poor medication management. In terms of concrete technological solutions for these issues we have advances in general internet and communication technology (ICT), robotics, telemedicine, sensor technology, medication management applications, and even video games (which can help maintain balance and cognitive function, as well as being fun); all of these can be used to help with both physical and mental aspects of aging (Vallor 2011). One of the most visible areas in which we see advances in technology and robotics is in mental health care for the elderly and those with cognitive impairments, which is likely to become an ever-growing segment of our global society (Economist 2020). Mental health care is sometimes overlooked, and there is a huge gap between those who cannot access the services they need and those who can. The area of socially assistive robotics engages with both eldercare and mental health care more generally, giving rise to many successes over the years (Rabbitt et al. 2014). While you may have heard of the PARO robot, a seal that makes happy-sounding noises when petted or interacted with, there are other advanced robots available for both companionship and physically assisting individuals. These helper technologies meet with mixed responses from the people they are designed to help: many older patients are happy with assistive robots but may find conversing with them tedious (Abdi et al. 2018; Broadbent et al. 2009). Yet, although there are large differences in care outcomes, there is broad acceptance of robots and assistive technology across these areas and many examples of positive outcomes (Góngora et al. 2019; Khosravi and Ghapanchi 2016).

It is important to think of some of the aspects of eldercare through the lens of our analytic concepts. Recent advances in robotic technology allows us to imagine possible futures in which everyone who needs physical assistance—moving or lifting objects, or even bathing—can be attended to by a robotic presence. But although this imagined future provides solutions to many of the challenges in our immediate present, it also raises important and difficult questions. Who decides who gets such a robot or not? How does the robot decide between what may be beneficial for its charge in the current moment versus in the future? There is also an interesting question of pluralities: what if someone does not desire the help of a robot? When will their desires be overridden by considerations for their children or other caretakers? Vallor (2011) points out that when considering care interventions based on robots, it is important to consider not only those receiving care but also those providing care, including considering who may no longer be able to provide care because of the technological intervention; displaced caregivers may lose their sense of agency, purpose, or value (moral or otherwise) as those who provide care to family and friends. Finally, there is a question of accessibility: older patients may not be able to read screens, or use buttons or switches that require hand strength or fine motor control, or otherwise be interactive in the ways that technology designers imagine. Likewise, many products designed for constant internet connectivity may not be useful in rural areas or to underserved populations who do not have reliable internet access.

Turning to medical care more generally, we find a number of radical changes being realized through advances in general ICT and its application broadly across the medical domain. Perhaps the most important of these is the widespread adoption and use of electronic health records (EHRs), digitized medical charts. An EHR allows care providers to pool and collate patient information, charts, X-rays, labs, and other important information between providers and through time (Atasoy et al. 2019). A fully envisioned EHR is passed seamlessly through the various providers, be they institutions or people, and is continually updated, providing caretakers with a kind of complete digital picture of the patient. EHRs have advanced rapidly between 2000 and 2020 with the advent of the Patient Protection and Affordable Care Act (ACA) of 2009 and its incentives for uptake of EHRs across the United States. EHRs were slow to catch on in the first decade of the twenty-first century (Jha et al. 2009), even though a handful of large hospital networks invested in EHRs in the early 2000s with very positive results (Buntin et al. 2011). But by the 2020s, over 90% of US hospitals were reported to have taken on EHRs, and a meta review shows that this shift has significantly increased the quality of patient outcomes while decreasing the cost of care (Atasoy et al. 2019). As with any other technological intervention, there are still issues related to standardization, interoperability, and access, but there is strong evidence that widespread adoption of EHRs has had significant positive impacts.

And yet for all the advancements that have been made in medical ICT and EHRs, there are a number of risks and concerns. We must think about the increased information flow between health care providers: who controls that data and when can it be taken back? When can it be used for other, nonmedical purposes? And if the state has full control over your care history, is that a problem? Is it a problem if that control lies with a private business? While there have been some recent laws passed in the EU and the United States, there is still a dearth of regulation around the volume, variety, and velocity of information that is collected about you while care is being provided (Anderson and Agarwal 2011; Electronic Frontier Foundation 2021). Although there are a number of important tools that expand access to more populations, there is growing concern over information exchange and privacy. Even more, if the psychologist at the other end of the call is a chat bot, who controls and regulates the bot’s behavior? (Fiske et al. 2019). Finally, along with the increasing connections brought about by advances in ICT, there are significant increases in the entire system’s ability to perform testing and other diagnostic tasks that can be automated through the creation of new machines (Improving Diagnosis in Healthcare 2015). For instance, the level of testing that the United States and other countries achieved during the COVID-19 pandemic wouldn’t be possible without massive automation of test processing.

These kinds of transformations are taking place across the medical field as a whole. And though these most recent changes seem very striking, it is important to remember that technological revolution in medicine is the norm and not the exception. Safe and widely available vaccines were never heard of 300 years ago (although the practice of variolation dates back much farther) (Silverstein 2001). The idea of looking inside a patient without cutting them open was, until recently, impossible. Even in more modern times we have moved from not having X-rays, to physical X-rays on film, to digital X-rays, to imaging tools such as MRI scans that can see soft tissue. This increasing reliance on elaborate medical machinery has led some people to fear that we are transforming our doctors from professionals to entities that blindly follow orders from a computer—one example of a phenomenon called deskilling, to which we return in section 5.5.6, “Work and Labor.” However, the evidence is clear that doctors and practitioners who work with well-designed decision support tools outperform other doctors. As our tools advance, so will the type, cost, and ability of medicine to care for us all (Wachter 2015).

Another key form of care that society provides and that technology affects in myriad ways is welfare and subsidies, two key forms of economic care (Myers 2001; Schulz-Forberg 2012). A subsidy, in the broad view, provides support to a particular group or sector by shifting some of the full economic cost from the particular group or sector onto society as a whole. One example of this kind of subsidy is public education: the cost of public education is not paid fully by those that receive it but is subsidized in various ways by society, because having an educated workforce is a public good. The same can be said for many other things that we enjoy as a society, from roads to basic research. The three main types of subsidy are as follows:

1. Preferential treatment for particular sectors or classes, such as providing housing payments, free medical care, and education payments.
2. Payments or encouragement for certain activities such as renewable energy credits and recycling.
3. Insurance subsidies, which include price floors for food production and manufacturing (Myers 2001).

Each of these types of subsidy can be direct or indirect. For instance, we can pay for food for individuals directly or can subsidize farming of crops, indirectly lowering food costs. If a town pays for garbage collection, it may be indirectly subsidizing the manufacture and distribution of one-time-use packaging. All these forms of subsidy comprise what we typically think of as economic care: we are choosing, as a society, to relieve some of the economic cost to a particular group or sector.

Subsidies also include welfare, the welfare state, and the social safety net. We can think of these as subsidies to individuals rather than firms or sectors; however, subsidies to both groups are forms of economic care. The concept of the welfare state, or the idea that the social state has a role to play in ensuring certain benefits to its citizens, is widely credited with starting in the German Empire in the 1880s with guarantees for workers against sickness and accidents (Schulz-Forberg 2012). However, as discussed in chapter 4, older religious states had forms of social safety nets going back millennia. One can view the welfare state as a form of subsidy to individuals (i.e., assistance to improve the lives of vulnerable individuals and families) (Ivaschenko et al. 2018). Examples include social pensions, food transfers, cash transfers, and fee waivers. There is a long history of discussions about the role of welfare in a society and questions around when it is appropriate, whether it distorts the operation of the markets, and whether it is still appropriate to provide for citizens in this way (Schulz-Forberg 2012).

Automation is changing the form and delivery of subsidies to poor families through the welfare state. A powerful examination of this technologized shift can be found in Virginia Eubanks’s book Automating Inequality (2018). In her book, Eubanks discusses three specific instances: the delivery of Medicaid in Indiana and its changeover from human caseworkers to an automated system; the coordinated entry system for homelessness in Los Angeles and its centralization through a large, standardized database; and child protective services in Pennsylvania and its automated scoring system to rate child risk. The case studies provide a comprehensive set of examples of how new automated decision-making tools are alleviating some problems but also creating new ones. Foremost when analyzing these systems, we must take into account the entire interpretative frame and include the people affected, the technology itself, and the history of the development of that technology. For instance, in the Medicaid example, efforts to automate the monitoring and dispersion of medical payments focused on a goal of clearing claims quickly, hence the contractor responsible started closing or declining claims at a rate many times higher than before, leading to a massive loss of care, numerous lawsuits, and often catastrophic impacts on those relying on Medicaid. This focus on system efficiency, clearing claims quickly, rather than health-care results, ensuring adequate and appropriate care, led to harmful effects on the most vulnerable.

A compelling argument from Eubanks is that modern technological tools allow us to “manage” the poor in plain sight instead of shipping them to work farms and poorhouses, as was done in the 1800s and 1900s. There are a number of concerns raised in this account, from the amount of information collected about the poor through these service providers and who it is shared with and how that information is used, to concerns of pluralities and who gets to decide what is the best form of care for an individual. Eubanks’ accounts demand that we look at poverty through a new lens. Too often, we view administering care to the poor as a systems management issue, wherein the problems would solve themselves if we only had better systems to deliver the scarce aid faster, in the right place and at the right time. This assumption limits our ability to consider the ways in which existing strategies are fundamentally mismatched to the problems, and therefore to consider alternatives seriously. For instance, the cost of building housing for all unhoused people would, in many places, cost less than maintaining the infrastructure required for providing more narrow and focused payments, such as food stamps, bus passes, or rent stabilization. Indeed, our focus on the systems instead of the outcomes can limit us to thinking more in terms of surveillance or disciplinary powers: if we pay for only certain things we can revoke payments when those receiving them do not act how we think they should, instead of helping to promote and ensuring the capabilities of those receiving economic care. Eubanks argues that it is important for everyone to pay attention to how these systems are designed and deployed, as one day they could be used for everyone. For instance, the monitoring systems currently used on the poor may be put to use, at some later time, to determine whether or not you are a safe enough driver to deserve car insurance. These examples are an interesting study in what we saw in section 5.4.3 as the portability trap, that is, using a system automated for one purpose (poverty risk) for some other context (car insurance) leading to misleading, inaccurate, or undesirable results (Selbst et al. 2019).

Another key form of subsidy in the United States is related to home ownership. In the post–World War II period in the United States there was an aggressive expansion via government programs to expand home ownership. However, these loans, insurances, mortgage guarantees, and other financial incentives were largely extended only to white individuals, while people of color were actively disengaged through the use of redlining (the practice of undervaluing homes in “bad” neighborhoods) and other tactics (McGhee 2022). Beginning in the late 1960s, due to the saturation of homeownership by whites and the need for more financialization of the home market, lenders began to increasingly lend to Black and minority families. This strategy is dubbed predatory inclusion by Keeanga-Yamahtta Taylor in her book Race for Profit. In this scheme, African American home buyers were granted access to conventional real estate practices and mortgage financing, but on more expensive and comparatively unequal terms (Taylor 2019). Today we see potential for many of the same problems in the expanded use of technology to measure credit and determine access to financial institutions. There are numerous companies focused on increasing access to financial markets for those historically marginalized, often at worse rates or at the cost of their privacy—often using access to “nontraditional risk models” to evaluate applicants’ creditworthiness through, for example, their Facebook friends (Bruckner 2019). This area is ripe for more forms of predatory inclusion that affect not only Black families but young people, old people, and many other populations, and it is important to view this through a holistic frame to understand how and why new technologies enable and/or cause harm.

Although we have used most of this section to discuss how we provide care for individuals through direct subsidies or welfare, it is important to realize that society’s economic care is not limited to individuals. There are important normative questions about how and why we subsidize corporations and businesses through loans, farms through price floors, and specific industries through large, guaranteed purchases (Myers 2001). Indeed, even the development of many of the technologies discussed in this book, from the microchip to the internet, were heavily subsidized by the US government (Isaacson 2014). The structure and deployment of these subsidies is constantly affected by technological change: the internet itself has opened up vast areas of new business; it is a piece of public infrastructure that has arguably enabled a completely new way of life. The original goal of the internet was to provide a mechanism for nuclear silos to communicate in the case of a nuclear war, but it has become enmeshed with our everyday life including the delivery of many structures of care from telemedicine to remote learning. This is again an example of our society existing in an ongoing conversation with our technological systems.

Unlike her father, Anna truly feels guilty that she does not visit more often. Her college classes and her two jobs leave her too tired to visit often, but she still wishes she could. So she calls every night, and I monitor the calls. Sometimes when Mildred falls asleep early, Anna talks directly to me. At first she did not understand my emulation abilities, but now she appreciates them. She shares with me thoughts and secrets that she would share with Mildred if she could, and she trusts me not to share them with anyone else.

Mildred is elderly, frail, and suffering from dementia, and she needs constant care. Her family provides her with a highly sophisticated android that is capable of emulating the appearance and personalities of people that it has scanned—and therefore, because it emulates her family members, Mildred believes that they are present and caring for her, rather than an android. There are signs throughout the story that Mildred’s family intended to remain active in her care, but now that the robot is actually there, they have adapted, showing up less to be with Mildred—and even replacing her with the robot in some ways. This story obviously raises questions about who is responsible for care work, but it also offers a rich and textured exploration of all the things that constitute care work: not just meeting a patient’s physical and medical needs, but the various kinds of emotional and psychological support they need, and who they need it from. And if we are willing to look past the empty pieties that often shape conversations about eldercare—as “Today I Am Paul” urges us to do—it equips us to confront uncomfortable but necessary questions about how to balance the needs and well-being of the caretakers against the needs of those under care.

In a democratic context, public discourse and free speech go hand in hand. One example of the regulation of public discourse is the US Constitution’s protection of freedom of speech. In the history of intellectual thought, these ideas are said to be oriented toward the human good—hence there is a presupposition that we all have an interest in preserving free speech as well as a space for the public exchange of ideas. Why is social exchange good? Why is it valuable? These are important questions to ask, especially in contemporary times when social media tends to blur the boundaries between private and public discourse.

In order for public speech to be truly public, there needs to be a forum or a venue where members of the public can speak and be heard. These spaces can be geographic, such as a town square, or they can be metaphorical, such as a newspaper, a radio broadcast, or a Reddit forum. However, these spaces do not simply exist spontaneously. Policy making and the discussion of policy is said to take place in a “public sphere” (Habermas 1989; Mill [1859] 2002). As we stated in chapter 4, the “public sphere” or “public space” in the classical sense is the space, real or imagined, where citizens interact as citizens. The government in many ways establishes the ground rules for how collective decisions are made at the public level. The government also has a role to play in how public discourse takes place, and so does technology (recall our discussion of technological mediation in section 5.2.4).

In chapter 4 we discussed the legal arguments that Warren and Brandeis offered at the end of the nineteenth century regarding the right to privacy. They argued that because recording technologies like the phonograph and camera allowed people to obtain information and distribute it in ways never seen previously, legislation had to be written to protect privacy as a right. These technologies altered the way in which individuals relate to the public as well as the public’s access to the government. We argued in chapter 4 that the guarantee of privacy was important both to preserve the freedom to form opinions and to make decisions about our own lives without interference from others, and also to allow us a space to come together as a citizenry: the public sphere.

The ability to engage freely in public discourse is not only of benefit to the person speaking; it is essential to the common good. John Stuart Mill, the utilitarian thinker discussed in chapter 2, was a strong advocate of free speech and the exchange of opinions and ideas. He thought that they benefited not only the political society of which he was a part, but humanity as a whole:

But the peculiar evil of silencing the expression of an opinion is, that it is robbing the human race; posterity as well as the existing generation; those who dissent from the opinion, still more than those who hold it. If the opinion is right, they are deprived of the opportunity of exchanging error for truth: if wrong, they lose, what is almost as great a benefit, the clearer perception and livelier impression of truth, produced by its collision with error. (Mill and Miller [1859] 2002, 18–19)

In other words, for Mill, even when an opinion is wrong, people should have the liberty to express it so that their view can be corrected. This is not to say that Mill believed there should be no regulation on speech; participation in the public sphere also entailed the use of reason and some regulation around what was acceptable. But Mill and others writing during the same time also insisted that free speech was grounded by the idea that human beings are free and equal and that we could reasonably agree on a set of goods and values, even if we did not always agree on how to achieve them.

The ideals of democratic communication have always been difficult, even before the advent of the internet and social media. Ideally, citizens participate in the discourse in ways that are both inclusive and aimed at shared understanding, if not practical consensus (in other words, as part of the same culture of knowledge). This type of exchange relies on two conditions: a shared base of information and knowledge about what is going on and what is at stake, and a good-faith effort to include everyone in the conversation. And yet, as far back as the founding of the United States, there have been concerns about the information that was being distributed by pamphlets and news articles (Campbell 2019). In other words, “fake news” is not a new problem. But the more quickly it can travel, the bigger a problem fake news will be—and on social media, it can travel faster than ever before.

Inclusiveness has likewise always posed challenges, but social media in the twenty-first century creates a new combination of opportunities and challenges by providing an alternative environment in which citizens can come together, exchange ideas, and participate in political discourse (Vallor 2016). Discussions about social media are an extension of discussions that have been ongoing for years about expression on the internet and other “public” spaces. One of the most pervasive issues in social media is the worry about “echo chambers” and political polarization that we saw in chapter 4 or the ability of the internet and social networking sites to “fragment” the public discussion in a way that promotes extremism (Sunstein 2009). Many of the laws and regulations around free speech were not written with social media as a starting point and do not help us understand current issues with free speech, much less address them.

New technologies for communication invite us to think differently about what it means to deliberate freely about political matters and what it means to participate in that conversation. As we have seen with the Black Lives Matter movement, rallies for independence in Hong Kong, and even the riot at the US Capitol in January 2021 (Baker and Tavernise 2021), social media can be used to rally large actions by potentially disparate geographical actors, often without central coordination. These changes in the scope, scale, and speed of interaction in the public sphere challenge us to consider which values and goods should be prioritized in public discourse, what role the government should play in matters like the distribution of resources and the protection of certain rights, and the role of technology in these contexts. Public discourse cannot exist without a public space, and the government has traditionally played a role in protecting— and, some would argue, in creating—public space. Changes in technology have raised questions not only about how the government should regulate social media but about whether social media companies ought to regulate the government.

Contemporary democracies are faced with all kinds of issues: environmental concerns, sustainability, and climate change; emerging technologies, biotechnologies and nanotechnologies, and issues regarding energy and engineering. As we have seen, citizenship comes with certain expectations including engaging with the collective process of making decisions on these and many other topics. In order to participate in the political process and exercise certain rights, citizens need to be willing, equipped, prepared, and presented with opportunities to participate in the public exchange of ideas and weigh in on matters of public policy.

Since we are rarely able to get the whole of a given society together in one room, some kind of technological mediation is always involved in the communicating and accessing of information, as we have seen both in chapter 3 and section 5.2.4. Various forms of technology, when deployed in this role as information mediator, can blur the distinction between informing citizens and persuading them. Specifically, the notion of “objectively informing,” if it is even possible to do so, implies that there must be some shared expectations about both the type and quality of information that we receive and the veracity and reliability of the sources of that information. But how can we determine whether we are being informed or persuaded? Is there a difference? Is it helpful to make a distinction? Who decides? And how are all these concerns bound up with new and changing technologies?

One unforeseen challenge of both the globalization discussed above and the internet is that we often receive more news at the national and international levels than we do at a local level. In the decade and a half leading to 2018, more than one in five newspapers in the United States shut down (Abernathy 2018). This lack of local reporting can lead to less coverage of important local issues and a fragmenting of the public discourse at a close, local level (PBS 2020). There are now many “news deserts” in the United States, where there are no local papers and no one there to shine a light on important aspects of local interest such as municipal water treatment, garbage collection, building permits, and other details of living in a society (Abernathy 2018).

Both the local and global fracturing of the news scape has also interacted with the rise of social media, as we discuss in the next section. Questions of what types of media count as valid, and what counts as truth, have led some to claim that we are living in a post-truth age. This idea has been described as truth decay and is characterized by four trends: (1) an increasing disagreement about the relevance or reliability of facts and the interpretations of facts and data, which echoes the datafication discussed in chapter 4; (2) an intentional or unintentional blurring of the line between facts and opinions, often on social media; (3) the rise of relying on anecdotes and personal experience over facts; and (4) a declining trust in traditional respected sources of factual information including newspapers, news outlet websites, magazines, and the government (Kavanagh and Rich 2018).

It was understood by all of us involved, I think, without necessarily needing to have it said, because even in this project it seemed crass to have to say it quite so bluntly, that he would memorialize the other, who had suffered at our hands, and apologize for our own, who had committed, that is to say, perpetrated, or at least were implicated where nothing was proven—there’s a reason they sent a poet, not a prosecutor.

Is it possible for a society to make amends for past wrongs? How can a culture or a people—or an industry—alleviate suffering that they themselves have caused? And how can the work of making amends be undertaken in a way that meaningfully benefits those who have been harmed, instead of simply soothing the consciences of those who did the harming? In this story, the dominant culture attempts to rectify its history of violence and destruction by sending a poet back through time to offer apologies at a selection of carefully chosen moments, known to the Poet’s avid audience as “ChoMos.” Even the story’s narrator, the editor in charge of shaping the Poet’s public profile, seems aware that this project is flawed—yet the form of critique he imagines takes the very same form: a sudden visit to his own time from some future apologizer. “Apologia” offers an unsparing look at how the dynamics of public discourse and social media can combine to trivialize important issues. It also challenges us to imagine more productive forms of public discourse, meaningfully centered on the common good.

In many key respects, people’s online lives are similar to or continuous with their offline lives. Concerns about whether online relationships are “real” relationships tend to evaporate under scrutiny (Elder 2014). People’s online behavior tends to correlate strongly with their offline behavior. For example, although internet researchers initially postulated that online dating might be especially appealing to people who found in-person dating stressful, studies have found that people who are more comfortable with in-person dating are also more likely to use online dating sites (Valkenburg and Peter 2007). Nor can sweeping judgments be made about what certain patterns of online engagement indicate about a person’s offline life: in spite of the moral panic about online social gaming, studies have found that those who engage in multiplayer online gaming do not have fewer or weaker offline friendships than nongamers (Domahidi 2018). There is substantial and mounting evidence that people experience their online and offline lives as fully integrated and continuous (Chan et al. 2020).

But it’s still true that interacting with and relating to others online is meaningfully different from in meatspace. Online relationships are largely disembodied and at a physical remove, and the conditions of connection typically make it easier for participants to check in and out more easily. These differences have been helpful for people whose physical disabilities have constricted their interactions with the meatspace world and have also created an opportunity for connection between people whose interests or values are out of step with their geographic neighbors. There is also a well-documented tendency for people to treat one another with greater hostility online. Many people assume that the anonymity of online interactions is the explanation for this high level of hostility, although some research suggests that the lack of eye contact between speakers and the psychological toll of information overload are just as significant as anonymity (Lapidot-Lefler and Barak 2012).

Beyond the general conditions of digital “space,” the kinds of interaction afforded by particular platforms (such as Facebook, TikTok, or Discord) or modes of contact (such as email or text messages) do create meaningful changes in the conditions of relationships for some people. The specific modes of interaction invited and encouraged by a given online platform can re-entrench and reinforce individuals’ existing struggles or concerns. For example, it has been shown that the widely attested phenomenon of social media addiction and the lower self-esteem that correlates with it are far more common among users who are very anxious for social approval, but also that the mechanics of social media exacerbate these existing anxieties and behaviors (Balcerowska et al. 2020; Kanat-Maymon et al. 2018; Shensa et al. 2020). Furthermore, it has been argued that Christian dating sites do not simply match users according to preference and suitability within given parameters but actively enforce the user base’s prevailing theological views of gender and sexuality on all of their users (Hutson et al. 2018; John 2019). Even in the case of platforms or communities that do not proclaim a particular set of norms or values, disciplinary norms often take shape anyhow. But it’s also important to recognize that not everyone will interact with or exploit a particular array of conditions for interaction in the same way. Of the many studies conducted on the social lives of online gamers, one study that took into account the sensitivity and shyness of its subjects found that shyer and more sensitive gamers tended to form more intimate relationships with fellow players than did their less sensitive peers (Kowert et al. 2014).

The patterning of social interactions and expectations that develops online can in turn create new patterns of interaction in the offline world. The advent of nearly instant internet-based communications has significantly reshaped many offices and pressured workers to respond speedily to online messages because the mode of communication permits it, even at the cost of their health, concentration, and leisure time (Barber and Santuzzi 2015). And as social media usage has become more widespread, some people (mostly, but not exclusively, younger people) have begun to rely on social media as a source of information that they can use to learn about potential new meatspace friends. One study of college students found that many considered it essential to “do homework” on their fellow students in order to screen out those whose political views they found abhorrent or who might otherwise be unsafe (Standlee 2019).

Online communities form in all kinds of circumstances, which in turn influence what kinds of communities flourish in any given potential point of contact. For instance, the types of interactions that one will have on email, discussion boards, community-driven websites, multiplayer games, messaging services like WhatsApp, or app-only interfaces such as TikTok will influence what kinds of users are supported and rewarded in their participation and what types of interaction take place. The way these communities take shape is influenced by many factors: what is the reason for the website or community portal’s existence, according to those who built it and maintain it? Who is the ostensible audience, and who else has a reason to be there? What modes of interaction does the website or portal encourage—and what additional modes of interaction does it permit, even if accidentally? What continuities or discontinuities does a given space afford to communities that migrate there from other settings, either offline or online? How much room does a given community make for a diverse range of participants and viewpoints, compared to alternatives? All of these factors have a major impact on the kinds of communities that develop in any given corner of the internet.

But the question is how, rather than whether, such communities develop. Online communities can and do form in any online venue that permits even the barest interaction between individual users. Users have found ways, in the words of Benedict Anderson, to “imagine communities”: to create or identify criteria that enable groups who will never all meet one another to understand themselves as part of a shared, coherent group (Gray et al. 2020; Morimoto and Chin 2017). The internet is not the first time or place in which sociotechnical shift has altered the conditions for imagining communities: technologies as varied as trains and air conditioning have, in different ways, changed the possibilities for community building. In communities as disparate as mailing lists, Twitter, and the online comment sections of newspapers, the same basic strategies appear for distinguishing community insiders from outsiders. Community builders will employ the pronoun “we” to articulate ideas, sentiments, or experiences that, they claim (sometimes explicitly but more often implicitly), are shared by members of their community but not by outsiders. Insider statements are also recognizable as such—in addition to, or even instead of, declarations of shared belief and experience—because of how they are expressed: by means of local catchwords or acronyms, in explicit contrast to designated outsider ideas, or—and this is particularly powerful—by employing the community’s insider humor (Gagnon 2020; Gal 2019; Gray et al. 2020). These “insider” elements of a given community become visible in observing newer community members, who are often at pains to demonstrate their insider status by invoking insider language or ideas more frequently than longtime users (Wakefield 2001).

Beyond these outward-facing markers by which users signal their insider status in a given community, there is significant evidence that long-term participation in online communities can impact a user’s beliefs, perceptions, and overall sense of identity (S. H. Lee et al. 2019; Scrivens et al. 2020). Long-term participants in various kinds of online communities, from X-Files mailing lists to online white supremacist message boards, experience shifts in their own values or interests, “becoming way more obsessed” with the object of a fandom or becoming more invested in some particular “nonwhite” group as the most significant threat to “white” people (Scrivens et al. 2021; Wakefield 2001). These shifts can and do happen under the influence of offline communities as well as online ones, but the availability of nonclosed online groups makes it easy to become immersed quickly and easily, without even necessarily interacting directly with other members. (Scrivens et al. 2020).

It is crucial to note that although this community construction often takes place mostly or entirely online, the divisions that inform the insider/outsider distinctions are drawn from meatspace social organizations and divisions, and often replicate power imbalances among insiders or aspiring insiders from those meatspace organizations (Morimoto and Chin 2017). Online communities create local networks of meaning that make sense only to community insiders, but those locally constructed networks of meaning typically rely on broader prior meanings that originate offline. In particular, the distinctive humor of a given online community is often informed in complex ways by offline dynamics (Gal 2019). Arguments about one topic often play out through explicit conversation/jokes about a different subject, using and sometimes widening existing social divides that exist outside the text. Humor is a powerful tool for community building, because “getting the joke” often requires understanding multiple layers of meaning—the spoken text, the implied real meaning or value, and the gap that exists between them—and there are many different ways in which an outsider can fail to participate successfully in the interplay between these many layers. In this way, offline meaning can be introduced to create a specific online context, because the particular connections between what is said and what is meant often involve importing existing distinctions from the meatspace world. Even if a particular online community’s humor makes sense only to participants in that community, it typically requires a specific set of knowledge or experiences from the meatspace world to underpin it.

Online communities have also begun to influence or integrate with communities that are constituted offline. The most obvious example of such an online community is NextDoor, which exists to connect users within a given neighborhood to one another and uses strict screening policies to ensure that only residents of a given neighborhood (and residents of neighborhoods immediately adjacent) have access to that neighborhood’s shared online space. The premise of NextDoor is that it facilitates communication within an already existing offline community. But as the site’s critics have pointed out (Bloch 2021; Kurwa 2019), NextDoor also engages in active “community imagining” work by creating hard boundaries between neighborhoods, boundaries that do not necessarily match the experiences of all or even most of the people living in the area that has been divided up (Lambright 2019). Furthermore, by adding an additional layer of insider-outsider distinction, NextDoor creates the conditions for a heightened sense of anxiety about “people who don’t belong.” Furthermore, because NextDoor is an imagined community created out of an ostensibly face-to-face one, participants are likely to generalize about their neighborhood as a whole on the basis of those neighbors they know and are more likely to treat as outsiders anyone who does not match that expectation—whether they live in the neighborhood or not. And of course, any neighborhood residents who don’t have ready access to the internet—or who are not adept at using it—can end up excluded from a crucial aspect of their community life. NextDoor is a particularly pointed example of how a tool designed without much reflection on pluralities can end up exacerbating existing tensions between those with more access and epistemic power and those with less.

I’m not sure if A. I.s can believe anything and I’m not supposed to ask her questions about such things, but that’s what the human control is for, right? To ask questions that the other A. I.s would never think to ask, to force this electronic extrapolation of my mother into untested territory and see if the simulated thought matrix holds up or breaks down. “Don’t tell me what you think. Tell me what you believe.”

There is a brief pause. If this were really my mother she’d be smiling by now, relishing the discussion.

Titilope, the narrator of “The Regression Test,” hasn’t spoken to her mother, Olusola, in nearly forty years—because Olusola is dead. But in her lifetime, Olusola was “Africa’s answer to Einstein” as well as Titilope’s beloved and complicated mother, and in the years since her death, a digitally preserved copy of her consciousness has continued to serve as science advisor to LegbaTech. In order to ensure that this digital copy is still meaningfully continuous with the person it represents, LegbaTech brings Titilope in to interview it. During the course of the conversation, Titilope struggles not only with her own grief and longing, but also with the strangeness of this digital preservation, both its similarities and its dissimilarities from her mother. While the mother-daughter relationship in this story undergoes a technological shift that is more dramatic than our current technology makes possible, it nonetheless raises trenchant questions for the present about how we can relate to each other within and across new technological channels and constraints.

In the modern era, interacting with the natural world for pleasure and leisure has often been a privilege available to wealthier classes. But advances in communication technologies, shared maps, geotagged images, and other advances have enabled more people to enjoy the natural world than ever before. In the United States, the National Park Service was established in 1916 to preserve, protect, and promote engagement and education with the natural and cultural resources of the United States (National Park Service 2021c). The National Park Service has made improved access for people with differing abilities a strategic goal, and is using new technology such as adding simple structures such as ramps and walkways and developing new tools for camping for those with different abilities (Charitan 2019; National Park Service 2021a, 2021b). This expanded access to the natural world can be understood as a move toward social equality through technological intervention. Even aside from the pleasure or joy of experiencing beautiful places, immersing oneself in nature has been linked to numerous health benefits, including reducing stress and promoting healing (Robbins 2020).

However, with the increased accessibility and tourism into our natural world, issues have arisen. One drawback to all the photos and videos that we share online is that it has increased tourism to parks and other natural locations. A number of park rangers and other groups tasked with managing visitor numbers to parks have observed surges in formerly unknown locations thanks to popularity on Instagram and other social media sites (Djossa 2019; McHugh 2016). In addition to the surging popularity of some areas, some formerly remote or dangerous hikes and experiences are being sought out by people who do not have the experience, equipment, or training to handle the unpredictable nature of these locations. In many parts of the United States, where parks and preserves can be extremely remote, these sudden surges of interest are leading to an overstretching of local, volunteer rescue groups. Furthermore, advances in cell phone technology are impacting the kinds of risks some travelers are willing to take. Some people enter parks underequipped or underprepared and reliant on the idea that they can be rescued from any dangerous situation they fall into by means of a simple phone call, even if that is not the actual case. Many of these resources are overstretched, underfunded, or unable to reach people in a timely manner, leading to dangerous situations for both the hikers and the rescue workers (Watkins 2021).

In addition to being a source of pleasure or inspiration, the natural world can be treated by humans as a resource, supplying us with food and with the raw materials for various technological productions. In 2015, the United Nations adopted 17 sustainable development goals that outline a pathway toward shared prosperity for both people and the planet. These goals include things like access to clean water, the safety of life on land and below water, climate protection, and food access. Many technology development groups have integrated these goals into long-term plans for development and research. One of these areas is known as computational sustainability, which focuses on projects that encourage computer scientists to play a role in addressing the societal and environmental challenges outlined in the goals (Gomes et al. 2019). One overwhelming problem is monitoring the growth of forests and the use of natural resources all over the world. Here again we see that the use of earth-observing satellite imagery combined with advances in machine learning and image processing to estimate land use, crop growth, and other outcomes have paved the way for more informed policy related to the UN’s sustainable development goals (Burke 2021). These large-scale changes via technological intervention, often called Earth system interventions, present both a risk and an opportunity for sustainability and welfare objectives (Reynolds 2021).

Of particular concern is energy production. The technological artifacts that that we often associate with technology development have used massive amounts of energy for decades. From coal-fired steam engines to gasoline for our cars, new technology often needs a power source that, more often than not, is extracted from the environment. Although there is a move toward more renewables and batteries, this will not solve all our problems. In addition, the change of technology from large, centralized power plants to local solar and wind farms requires a fundamental rethinking of the overall technological system of energy delivery: from extracting from a concentrated source and distributing outward, to collection at distributed points to power concentrated centers (Maly 2015). One aspect of the natural world that is gaining increased attention is how reliant on extraction we are to build and maintain all the electronic devices we use, including batteries, cell phones, cars, and the other conveniences of daily life. Each cell phone or screen you use contains a multitude of precious metals including gold, platinum, and cobalt. Many of these materials come from hard-to-reach places or war-torn countries, or must be mined from virgin sites. The increased turnover of our devices and the need for ever more electronics (e.g., for electric cars) have increased demand for these precious materials (Nogrady 2016). While US recycling is at an all-time high, we are also generating more waste than ever before (Environmental Protection Agency, n.d). Although there is technology that can help with some of these problems, such as new energy capture mechanisms, it’s important not to fall into the solutionism trap—after all, a “the solution must be new machines” mentality is what helped create many of the problems to begin with.

Perhaps the most important way that many of us think about ecology and the environment is as the source of our food. Agriculture and farming is perhaps one of the oldest forms of technology and is often credited with leading to the establishment of early civilizations (Diamond 1999) (although recent scholarship calls into doubt whether farming actually was better for most people [Scott 2017]). Humans have long sought to tame and transform the natural world in both constructive and damaging ways to grow crops, using techniques ranging from the regulation of water supplies with dams to enriching the ground with natural fertilizers such as manure and other chemicals (Harford 2017). In more recent times, agriculture technology, including the Green Revolution, has led to the modern day when we produce more food, with fewer people involved in production, and (debatably) smaller environmental impacts than ever before (Evenson and Gollin 2003). As part of this shift, and in an effort to meet some of the sustainable development goals, inventions such as genetically modified Golden Rice that can produce vitamin A have been used in many places to lower infant mortality and blindness (NYU Langone Health, n.d.). However, the question of genetically modified crops and other technology for food production is a contentious issue, with many questions related to seed ownership, patent lock-in, and possible harm to the natural world (Barlett and Steele 2008). We cannot hope to do the discussion of the impact of farming and technology on the natural world justice in this book, but it is an area ripe for discussion about the impacts of technological systems on the natural world (National Research Council 2010).

“I mean, isn’t that an old story? ” She sips her champagne. “Even America is reducing emissions now. Everyone knows it’s a problem.” She taps her couch’s armrest. “The carbon tax on my limo has tripled, even with the hybrid engine. Everyone agrees it’s a problem. We’re going to fix it. What’s there to write about? ”

Ong is a reporter with a mission: he wants to educate the public about climate change, and call attention to how corporate exploitation and government corruption have hastened the destruction of the natural world. Unfortunately for Ong, the American public is more interested in reading celebrity gossip and reviews of the newest gadgets. Even worse, the media company he works for has embraced the infotainment model and is threatening to fire him if he cannot improve his numbers. When a generous coworker sets Ong up with a flashy celebrity interview, Ong struggles with several moral quandaries that plague many environmentalists: what are the costs of commodifying a thing you consider valuable for its own sake, not just for the pleasure that it can offer others? Is it possible to introduce that valuable thing to others without putting its integrity at risk? And, above all else: how do you convince people not to look away?

The most visible aspect of a typical state security apparatus is its police force. In democratic societies, the understood role for the police is to enforce laws, control crime, and support communities (Bittner 1970). In practice, policing often diverges from this model. In the United States, police violence—particularly against communities of color—has been a site of significant public upheaval and has led to calls for significant police reform or even for the defunding of the police in favor of alternative intervention and support services. Nevertheless, the police remain central to the US social structure and its security practices.

Policing scholar Alex Vitale has argued that police themselves suffer under the burden of being asked to do too much (Vitale 2021). Police work necessarily requires individual officers to exercise discretion about how to fulfill their mandate, and sometimes this includes making split-second decisions to protect themselves or others, sometimes through violence. But as the purview of policing has expanded to incorporate everything from wellness checks to crisis management, even police officers committed to the public well-being often find themselves overwhelmed. Additionally, the history of biased and racist policing has created patterns that are challenging to disrupt even for individual officers who do not wish to perpetuate them.

Many police departments have sought to address both of these problems—the heavy decision-making burden on individual officers and the history of institutionalized bias—by offloading some of this discretion to algorithms, on the grounds that algorithmic decision making will decrease the bias that creeps in through human decision makers. But though the algorithms are trained on data from actual events, this data is flawed, incomplete, or outright misleading in a variety of ways. Consider, for instance, the widely used practice of “predictive policing,” in which algorithms are used to determine which areas of a given jurisdiction should be patrolled most heavily based on past data (Angwin et al. 2016). Other systems scrape social media and use social network analysis to pinpoint individuals and groups that are seen as statistically likely to be victims or perpetrators of crimes (Winston 2018).

This reliance on data may seem impartial, but the training data for these algorithms necessarily reflects entrenched historical biases and unjust patterns of policing, because the only training data available consists of reported crimes. Any bias that has driven police to arrest some people and not others (for instance, on the basis of race, or living in a “bad” neighborhood) will be replicated in the algorithm (see, e.g., O’Neil 2016 and our discussion in chapter 3). Furthermore, training on reports of crimes could leave immigrant neighborhoods largely unpoliced, because immigrants are among the populations that report crime at a lower rate (Shapiro 2017). This problem is larger than crime data; some bail-setting systems, trained on bail decisions from a racially biased court system, have been shown to judge individuals as risky primarily on the basis of race (Angwin et al. 2016).

The social theorist and philosopher Michel Foucault (2008) describes two settings with massive surveillance, with very different goals: to eradicate a pandemic and to maintain control of an imprisoned population. The first setting was European cities in the seventeenth century, during the bubonic plague, when people were locked into their homes and checked (from the windows) for plague symptoms. Foucault contrasts this temporary state of discipline with that of what he calls the carceral society, which is centered (theoretically, if not geographically) around the Panopticon, as imagined by utilitarian thinker Jeremy Bentham: a vision or plan for a prison in which prisoners are each isolated in a cell from which they can see and monitor each other, and such that a guard (or visitor) in a central tower can see all of them, though they cannot see into the guard tower—much like modern prisons, in which guards sit in a protected space, watching video from cameras throughout the institution. The visibility of prisoners and their uncertainty (in Bentham’s model) about when they are being watched causes prisoners to self-police. For Foucault, the prison building itself, the actual architecture, was construed as a technological mechanism. Although prisons seem less technological than iron working, both change society in wide-ranging ways. Foucault points out that constant, anonymous surveillance can be used in other institutions, including military barracks, orphanages, and schools. More importantly, as we saw in our examination of prisons in chapter 4, the knowledge that the Panopticon exists, its threat, causes individuals to self-police, to behave in ways that will not offend those with the power to incarcerate. That same uncertainty about surveillance has become part of our society, where cell phones, drones, and web cams (including baby monitoring and doorbell cams) function as the Panopticon.

In the United States, we have many forms of surveillance aimed at regulating particular behaviors, such as security cameras in shops, banks, and so on, to prevent theft; traffic cams, to prevent speeding, running red lights, and other violations; and software for proctoring online exams. One interesting finding is that red light cameras, which sit above traffic lights, decrease the number of drivers that run red lights, but increase rear-end collisions (Cohn et al. 2020)—an interesting example of a ripple effect! Benjamin (2019) argues that all these technologies, as currently used, reinforce the power structures already in place. For instance, cameras in shops are justified to stop shoplifting and employee misbehavior, but end up creating a more hostile and micromanaged workplace. This is very much part of digital Taylorism (which is discussed in the sidebar in section 5.4.3). Video doorbells are marketed as a means of tracking home intruders and porch pirates, but they can also be used to monitor an individual resident’s comings and goings, which can reinforce control over the resident. Various forms of tracking, through cell phones, webcams, social media “check ins” or images, and GPS devices can be used by stalkers (Messing et al. 2020).

Recent research has demonstrated that the modern carceral system serves several purposes beyond incarceration itself, including elimination of populations. Cisneros (2016) describes the effects of immigrant detention centers in France, which, contrary to international law on the treatment of refugees, serve to break the spirits, cultural traditions, and health of the detained. These centers, and various governments’ policies of breaking up families and sending children to live with current citizens’ families, echo different governments’ deliberate attempts to wipe out Indigenous cultures in Australia (Palmiste 2008), Hawai‘i (Nee-Benham 1998), and the land that is now the continental United States (Adams 1995; Piccard 2013). Looking at the effect of policing on Black and Latino boys, Rios (2011) argues that the boys were not reformed by the experience of the panoptic prison system. Rather, “it stripped them of their dignity and humanity by systematically marking them and denying them the ability to function in school, in the labor market, and as law-abiding citizens” (Rios 2011, 88). Malkia Amala Cyril (2015) describes the continuing systemic and systematic denial of Black people’s dignity by the American government, and the ways, such as gang databases and drone surveillance, in which technology has increased surveillance and intrusion into the lives of people of color. As we discussed in chapter 4, prison systems can function either as sites of reform or of punishment, denying privacy as a means of reducing personhood.

The surveillance of the Panopticon is also supplemented by what researchers call “sousveillance.” The word surveillance, as Poster (2019) points out, literally means “watching from above.” In the Panopticon, prisoners are watched by everyone, including each other. The term sousveillance, by contrast, refers to watching from below, as a means of reflecting back the surveilling gaze. “Sousveillance focuses on enhancing the ability of people to access and collect data about their surveillance and to neutralize surveillance” (Mann et al. 2003). One can see the push for police to wear body cams as a form of sousveillance. Poster uses the term “multi-veillance” to describe the particularly technologically mediated ways in which people monitor one another via technologies like social media and web cams, and to police one another’s behavior and expression through comments sections, and apps that provide ratings opportunities. The pressure that multi-veillance creates on individuals to live up to social norms has effects that include depression, anxiety, and extremism (Bright 2018).

Another major dimension of state power can be seen when states act against other governments—through espionage, propaganda, economic attacks, and sometimes acts of war—on the grounds that such actions are necessary to protect their own citizens. It is important to note that the apparatus of these conflicts has significant internal effects on a society as well. Sometimes, these internal effects are caused when technologies are developed for military purposes and are later repurposed for non-military use. But it is also the case that governments sometimes quietly deploy these techniques against their own citizens.

One technique of warfare that is frequently deployed at home is information warfare. Information warfare can take the form of both propaganda operations (the pushing of explicit messages) and chaos-producing operations (which sow confusion through a flood of contradictory messaging) (Lin and Kerr 2019). These techniques can affect individuals, for instance, by convincing them to vote against their best interests (Woolley 2016), or by deflecting their attention from something that will affect them later. These can be carried out by governmental agencies, by individuals or hacktivist collectives operating in deniable concert with State authorities, or independently. Recently, we’ve also seen such ransomware attacks on infrastructure by groups, some of which are rumored to be associated with foreign governments (e.g., see Sanger et al. 2021). Though these efforts rely on recent forms of technology, propaganda as such is not new. Governments have long used incentives like food or social prestige to incentivize their audience to sign on to particular political positions. In addition, they have used available communications technologies to spread false, slanted, or incomplete information, and when dissenters have circulated conflicting information, they have attempted to disrupt that circulation in whatever way that communication technology requires, such as shutting down radio broadcasts, collecting and destroying paper pamphlets, or arresting public protestors. Yet the speed and scale of ICT makes it an especially effective vehicle for propaganda. The internet allows for an unprecedented volume of bad information to be shared anonymously or pseudonymously. We have seen discussion of the role of social media companies in distinguishing the truthfulness of posts and claims that algorithms can take on that task (see, e.g., Kahn 2021). Algorithms can be useful if they are updated regularly to keep up with the evolving tactics of malicious posters, but we should question whom those algorithms will serve.

Recent developments in warfare technology—including communications technologies and many forms of automation—have shifted the experience of warfare, both for those who have access to those technologies and for their targets. Earlier in this section we mentioned predictive policing, in which a neighborhood that is assumed (by humans, and the algorithms trained on human data) to be the locus of more crime becomes more surveilled. In modern warfare, drones are used analogously to human police officers, to surveil putative criminals, and (after a period of observation), to kill some of the surveilled individuals. This leaves a population terrorized by targeted killings, and “collateral damage” of civilian deaths and injuries (Wall and Monahan 2015). Individuals and communities indeed change their behaviors, going out less and gathering less, and often developing further reasons to hate the government that has sent the drones (Coyne and Hall 2018).

The panoptical gaze, whether from prison guards or via drones, is designed to enforce certain behaviors, to keep particular groups subjugated and anxious. Being surveilled is harmful to those being surveilled, as noted above, but the act of surveillance is often harmful to those who watch as well. While there has been literature on stress in prison guards (Keinan and Malach-Pines 2007), the growing body of work linking drone control and PTSD shows ways that technology can create new stressors for the cogs in the surveillance machine (Enemark 2019). The story “Codename: Delphi,” by Linda Nagata, included with this textbook engages with this question very directly.

The rhetoric of protection is one of boundaries: between lawful citizens and law-breakers, between inmates and unconstrained citizens, between citizens of our country and our enemies, and between the ill and the healthy. Technology has been presented as a means of distinguishing between lawful and unlawful, as if video of an act (for instance, without context) is sufficient to judge that act. Consider, for instance, a video of a confrontation that omits the initial provocation that led to anger or violence. Zooming out to show a mob, armed militias, or someone in danger, can alter our judgment, as can the backstory for the individuals involved in the confrontation. As we discussed in chapter 3, video data is not information. It requires interpretation and context. The discussion in section 3.2.3 reminds us to think about who does that interpretation. What assumptions do the people or entities who surveil hold about those who are surveilled, and how does their culture of knowledge influence which data is accrued for interpretation?

She’s smiling as he comes up to the car, and he’s smiling, too, and now he’s reaching for the door handle … but the door seems to be locked, and she’s fumbling for the release. “Carly,” she says, turning away from the sight of his face caught there in the window as if Carly were an actual person sitting in the driver’s seat, when, of course, there’s no one there. “Carly, is the child lock on? ”

“I’m sorry,” Carly says, “but this individual is untrustworthy. Don’t you recall what happened last Tuesday evening at 9:19 P.M. in front of the S.P.C.A. facility at 83622 Haverford Drive? ”

“Carly,” she says, “open the door.”

Cindy works as an advocate for the homeless, and her job keeps her busy enough that she relies heavily on the AI of her autonomous vehicle to manage her life for her. The latest advances in AV technology have caused some serious problems for Cindy’s clients, including her favorite client, Keystone. But for the most part, Cindy herself is happy to let the car keep track of her work calendar, her shopping preferences, and her teenage son Jackie’s movements. What does bother Cindy is when the seamless management by the car’s AI crosses into outright interference with Cindy’s goals and desires, like spending more time with Keystone. This interference is a logical extension of what personal AVs have been designed for: it’s just another form of keeping users “safe.” But “safety” can encompass a lot of things, and is sometimes in tension with our immediate pleasure or even our long-term goals. Cindy’s dissatisfactions, and her mixed feelings about the car on which she depends so heavily, urge us to consider what it costs us to enmesh ourselves into convenient, automated systems that keep us safe from ourselves.

Consider an historical example: In the earliest days of the United States, cotton-growing was not the huge enterprise it would later become. In the process of turning raw cotton into cloth, there was a time-consuming step of extracting the fiber from the cotton seeds by hand. Then in 1793, the invention of the cotton gin greatly sped up the extraction process. With this manufacturing bottleneck eliminated, there was suddenly a much higher demand for raw cotton. The next few decades saw a dramatic growth in cotton farming in the Southern states, powered by a dramatic growth in the use of slave labor (Bailey 1994). Although the practices of cotton farming itself did not change directly because of the invention of the cotton gin, there was a huge indirect impact. It was, in fact, because there was no commensurate increase in the efficiency of farming that using the labor of enslaved people became more profitable.

This example illustrates how technologies and labor practices form a complex system in which it is difficult to predict how change in one area will affect others (e.g., will cause a ripple effect). It also illustrates how economic opportunity does not always translate to an improvement in human welfare. We will examine contemporary developments through the lens of traditional employment, whereby a worker has an ongoing agreement with an entity to provide a service in exchange for predictable payments. This may be the most familiar way of arranging labor for many people in the twentieth and early twenty-first centuries, but its role in the future is uncertain; automation is affecting the makeup and availability of jobs and the internet is enabling new ways to seek income outside of employment.

The future effects of artificial intelligence, robotics, and other automation technologies on jobs are hard to predict. Some economists argue that automation may increase employment by, in a ripple effect similar to the cotton gin’s, allowing previously unmet demand for products to be addressed (Autor 2015; Bessen 2018). But a 2018 survey of experts’ predictions showed that there was no consensus over whether automation would result in a net creation or net destruction of available jobs over the coming decades, and to what extent (Winick 2018).

As an example of the relationship between automation and employment, consider a 2017 study on the effects of industrial robots in Germany, one of the leading countries in robotization (Dauth et al. 2017). The study suggests that industrial robots did not cause a net reduction in the overall number of jobs, but they changed which jobs were available: in robot-exposed manufacturing sectors, there were fewer new jobs created, but higher rates of job retention for workers who were already in the industry. Although robots drove a net loss of manufacturing jobs, they also drove a net gain in service-sector jobs. In addition to job availability, robots affected wages in asymmetric ways, with higher pay for some occupations (e.g., scientists and managers) and lower pay for others (e.g., machine operators). Greater productivity from robots led to greater per-worker profitability, but a lack of average increase in wages suggests that the main beneficiaries of this productivity were investors and not the labor class itself.

A key distinction is that “tasks are automated, not jobs” (Bessen et al. 2020). Rarely does a technology replace every function of a job at once; instead, the parts of the workday originally spent doing a task manually may be spent interacting with a machine to complete that task or may be freed up for a different kind of task. This shift may be rewarding for workers if it eliminates tedium and allows them to work on more engaging tasks; workers who enjoy tasks requiring social skills or creativity may particularly flourish, as these capacities have been among the hardest to automate.

However, a technology that seems empowering in isolation can sometimes degrade rather than enhance workers’ capabilities because of the way they are integrated into existing systems. One manifestation of this effect is “deskilling,” the erosion of workers’ rich sets of individual skills and experiences as their work is restructured in a way that gives them less discretion and autonomy (Hoff 2011). For instance, some commentators have partially blamed a number of contemporary air disasters on the technology-enabled deskilling of pilots; they argue that heavy automation of airplane controls, and reliance on predictable simulations for training, have reduced the depth of skill needed to become a pilot. From a safety perspective, this has left pilots unprepared for novel situations when the automated controls break down; from a social perspective, it has left some pilots feeling like piloting as a professional category has been diminished (Langewiesche 2019).

Up to now, we have presupposed that employment is the main organizing structure for human labor, but this is by no means a permanent reality, especially with the emergence of digitally mediated platform work; as of 2022, people could make money by, for instance, providing freelance transportation services through Uber or Lyft, completing crowdsourced small tasks on Amazon’s Mechanical Turk, or gaining sponsorships as social media “influencers.” Although some of these types of work are new, their structure is similar to the offline contract work that had already been on the rise beforehand, and they inherit some of the same debates about to what extent they are empowering or disempowering for workers. On the one hand, they offer lower barriers to entry than traditional employment and greater worker control over the time, place, and amount of work; but on the other hand, they can leave a worker with fewer protections and benefits than traditional employment (e.g., in the United States, companies must offer health insurance for certain types of employee, but not for contractors) and fewer communication pipelines for the worker to influence how the company is run (Vallas and Schor 2020).

Digitization has also changed traditional employment by enabling remote work, from answering work emails on a phone after coming home from an office to doing the entire job from a home PC in place of the office. As we discussed in the section “What’s the Response, and Why,” this has led to a change in how we organize ourselves and how close one must live to their place of work. A common thread in many kinds of remote work and platform work is constant connectivity (Turkle 2008): the notion of a neat division of the worker’s time into “working hours” and “personal life” has eroded. The availability of work in new times and places can open opportunities for workers who would otherwise have difficulty commuting, such as those who live in remote rural areas, those who cannot afford a car they would need to get to work, and those who cannot drive due to disabilities (Vallas and Schor 2020). However, the increased availability of the worker, plus established ideals about how a diligent worker should behave, can combine to create increased pressure to work longer hours than before and create stress when job expectations interrupt a worker’s family life (Stich 2020). This phenomenon is common enough that psychologists have given it a name, “telepressure” (Barber and Santuzzi 2015).

Telepressure may be amplified by a worker’s own self-image but also by the need to give a good impression to a supervisor. In a traditional office, managers’ access to workers’ activities comes naturally from proximity. The classical “scientific management” model that we described in section 5.4.3, in our sidebar on digital Taylorism (Drury [1922] 1968), encourages managers to use any opportunity for surveillance to its fullest extent, under the assumption that workers will behave lazily whenever they can escape accountability. We have seen this notion of surveillance in the workplace before in chapter 4, when we discussed open offices, as well as some of the complications of any type of surveillance in “Surveillance” in section 5.5.5. In remote work this surveillance takes on different qualities: because workers are often not observable by default and many digital surveillance measures can be circumvented, workers are incentivized to make an active effort to be noticed—both by managers, to have accomplishments credited, and by peers, to try to replicate the community structure of the traditional office (Hafermalz 2020). In such a setting, unlike under traditional surveillance, workers can curate what is observed about them.

I got a ride to Noe Valley from a sunken-eyed kid with a hipster chignon and wisps of blond beard floating around his chin. With me in the back seat were grocery deliveries for some next gig: a fresh baguette that filled the electric car with the smell of hot bread, a giant jackfruit that must have weighed twenty pounds and cost more than that day’s commissions. These were the people I’d forgotten, daily, nightly, the poor kids running the city out of the corners of my eyes.

Andrew is a wealthy San Francisco lawyer, and his wife Madeleine is a musician who comes from money. Though Andrew spends the story in a state of emotional distress about how he uses his neural implant, this anxiety plays out on a backdrop of easy material comfort: his laundry is done for him, his groceries and hair products are delivered to his doorstep, and he is transported wherever he needs to be without the unpleasantness of traffic jams on the surface. Andrew himself doesn’t pay much attention to these conveniences, preferring instead to obsess about what part of his own life he has hidden from himself in Aleph’s privacy mode; it does not occur to him (nor possibly to us, until we join him in privacy mode in the final scene) that these things are related. Although the gig work technologies that keep this future San Francisco running remain largely unseen, “Lacuna Heights” gives us an up-close look at how knowledge- and attention-management technologies like the Aleph—or even modest, now-ordinary technologies like a newsfeed-organizing tool or a tag blocker—can shape how we view labor … or how we don’t.