1. ON THE NATURE OF INTERNET ADDICTION

Clinicians and scholars began claiming in the late 1990s that excessive internet use was a growing problem that should be recognized as an addiction (Griffiths, Reference Griffiths and Gackenbach1998; Thompson, Reference Thompson1996; Young, Reference Young1998b). But there were then, and there continue to be, several controversies surrounding the concept of addiction (Du Plessis, Reference Du Plessis2012; Pickard, Ahmed, & Foddy, Reference Pickard, Ahmed and Foddy2015; Shaffer, Reference Shaffer1997; West & Brown, Reference West and Brown2013) and its application to excessive use of the internet. To begin, there are multiple theories of addiction (West & Brown, Reference West and Brown2013), including choice theories (which claim that addicts are not compelled to use an addictive substance but choose to do so because its perceived benefits are greater than its perceived costs; Ainslie, Reference Ainslie2013; Becker & Murphy, Reference Becker and Murphy1988; Campbell, Reference Campbell2003; Heyman, Reference Heyman2009, Reference Heyman2013; Skog, Reference Skog, Heather and Vuchinich2003); disease theories (originating in the nineteenth century and officially endorsed by the American Medical Association in 1956, these claim addictive substances produce persistent pathological changes in vulnerable individuals that generate powerful craving and weakened self-control; Leshner, Reference Leshner1997; Levine, Reference Levine1978; McLellan, Lewis, O’Brien, & Kleber, Reference McLellan, Lewis, O’Brien and Kleber2000; Volkow & Koob, Reference Volkow and Koob2015; Volkow, Koob, & McLellan, Reference Volkow, Koob and McLellan2016); learning theories (which hold that addiction is a learned behavior acquired through a conditioning process of positive and negative reinforcement; Drummond, Cooper, & Glautier, Reference Drummond, Cooper and Glautier1990; Niaura, Reference Niaura2000); and, most recently, neurobiological theories (versions of disease theories developed in the 1990s, these propose specific molecular [e.g., the neurotransmitters dopamine and gamma-aminobutyric acid] and neurological [e.g., activity in the ventral tegmental area, the nucleus accumbens, and the prefrontal cortex] mechanisms of addiction; Fakhoury, Reference Fakhoury2014; Goldstein & Volkow, Reference Goldstein and Volkow2011; Koob & Simon, Reference Koob and Simon2009).

Moreover, there are also multiple views of how addiction itself should be defined. The two most important clinical definitions of addiction are those provided by the Diagnostic and Statistical Manual of Mental Disorders, 5th edition (DSM-5; American Psychiatric Association, 2013), and the International Statistical Classification of Diseases and Related Health Problems, 11th revision (ICD-11; World Health Organization, 2018). The DSM-5 and the ICD-11 provide similar characterizations of addictions, but while the ICD-11 defines “disorders due to substance use or addictive behaviors” in terms of seven criteria, the DSM-5 proposes eleven criteria, at least two of which do not appear in the ICD-11 list (Basu & Ghosh, Reference Basu and Ghosh2018). Nonclinical definitions are also a subject of controversy. The Merriam-Webster online dictionary gives a medical definition of addiction as a “compulsive physiological need for and use of a habit-forming substance (such as heroin, nicotine, or alcohol) characterized by tolerance and by well-defined physiological symptoms . . . on withdrawal.”Footnote ¹ Yet some researchers argue that addictions do not involve compulsion (Heyman, Reference Heyman2009; Schaler, Reference Schaler2000) and others that addictions need not involve a habit-forming substance (Rosenberg & Feder, Reference Rosenberg and Feder2014; Yau & Potenza, Reference Yau and Potenza2015).

While recognizing the controversies surrounding the concept of addiction, we do not have to settle them here. For the purposes of our article, we adopt the view of Griffiths (Reference Griffiths, Szabo and Terry2005; see also Kim & Hodgins, Reference Kim and Hodgins2018; Shaffer, LaPlante, LaBrie, Kidman, Donato, & Stanton, Reference Shaffer, LaPlante, LaBrie, Kidman, Donato and Stanton2004), who argues that all addictions are marked by six common components: salience (the addictive activity, e.g., smoking or taking heroin, dominates the addict’s thoughts, feelings, and behavior), mood modification (the activity produces a “buzz,” “high,” or relaxed, destressed feeling or “flow”), tolerance (the addict must engage in increasing amounts of the activity to achieve its former effects), withdrawal (the addict experiences distress or unpleasant physical effects when unable to access the activity), conflict (the addict has conflicts within oneself or with those around them or experiences other adverse circumstances or damage), and relapse (unable to control oneself, the addict reverts to the activity after trying to stop). These six components overlap with the clinical criteria advocated by the DSM-5 and the ICD-11 and with other proposed lists of diagnostic criteria (Ko, Yen, Chen, Chen, & Yen, Reference Ko, Yen, Yen, Yen, Chen and Yen2005; Tao, Huang, Wang, Zhang, Zhang, & Li, Reference Tao, Huang, Wang, Zhang, Zhang and Li2010; Young, Reference Young1998b). Griffith’s components view provides a well-defined characterization of addiction and has been verified and endorsed by a number of studies (Andreassen, Torbjørn, Brunborg, & Pallesen, Reference Andreassen, Torbjørn, Brunborg and Pallesen2012; Canan, Ataoglu, Nichols, Yildirim, & Ozturk, Reference Canan, Ataoglu, Nichols, Yildirim and Ozturk2010; Griffiths, Szabo, & Terry, Reference Griffiths2005; Kuss, Shorter, van Rooij, Griffiths, & Schoenmakers, Reference Kuss, Shorter, Rooij, Griffiths and Schoenmakers2014; Lemmens, Valkenburg, & Peter, Reference Lemmens, Valkenburg and Peter2009; Nichols & Nicki, Reference Nichols and Nicki2004), and so it is our preferred characterization of internet addictions. Note, however, that although much of the research on internet-related addictions uses diagnostic instruments based on Griffiths’s criteria, some research uses instruments based on the DSM-5 and other criteria.

In addition to the controversies over what addiction is are issues related to the term internet addiction. Internet addiction, of course, does not refer to an addiction to a substance, such as heroin or nicotine, but is an addiction to a behavior, alongside other behavioral addictions, such as addictions to gambling, sex, and exercise (Karim & Chaudhri, Reference Karim and Chaudhri2012; Marks, Reference Marks1990). But until recently, psychiatrists and other medical professionals did not recognize behaviors as potentially addicting (Holden, Reference Holden2001, Reference Holden2010; Karim & Chaudhri, Reference Karim and Chaudhri2012; Walker, Reference Walker1989). It was not until 2013—with the publication of the DSM-5—that the American Psychiatric Association (APA) for the first time recognized a behavioral disorder (i.e., excessive gambling) as an addiction. Here we adopt the position of the DSM-5 that behaviors as well as substances can be addicting (Grant, Potenza, Weinstein, & Gorelick, Reference Grant, Potenza, Weinstein and Gorelick2010; Robbins & Clark, Reference Robbins and Clark2015).

Although the APA has not yet officially classified excessive use of the internet as an addiction (American Psychiatric Association, 2013: 795–98; Basu & Ghosh, Reference Basu and Ghosh2018), it is now widely accepted that it is an addiction (Block, Reference Block2008; Ha, Yoo, Cho, Chin, Shin, & Kim, Reference Ha, Yoo, Cho, Chin, Shin and Kim2006; Karim & Chaudhri, Reference Karim and Chaudhri2012; Olsen, Reference Olsen2011; Robbins & Clark, Reference Robbins and Clark2015; Tao et al., Reference Tao, Huang, Wang, Zhang, Zhang and Li2010; Young, Reference Young1996, Reference Young1998b, Reference Young2004; Young & Rogers, Reference Young and Rogers1998).Footnote ² Excessive use of the internet, in particular, is now widely accepted as a behavioral addiction for three reasons. First, and most importantly, excessive use of the internet exhibits the same criteria that mark other substance and behavioral addictions (Andreassen, Reference Andreassen2015; Grant & Chamberlain, Reference Grant and Chamberlain2014; Grant et al., Reference Grant, Potenza, Weinstein and Gorelick2010; Kuss & Griffiths, Reference Kuss and Griffiths2012; Young, Reference Young1998a, Reference Young1998b). Second, functional neuroimaging studies have shown that the same areas of the brain that are active in other substance and behavioral addictions are active in the brains of those who meet the diagnostic criteria for internet addiction (Brand, Young, & Laier, Reference Brand, Young and Laier2014; Han, Bolo, Daniels, Arenella, Lyoo, & Renshaw, Reference Han, Bolo, Daniels, Arenella, Lyoo and Renshaw2011; Holden, Reference Holden2001; Ko et al., Reference Ko, Liu, Hsiao, Yen, Yang and Lin2009; Ko, Liu, Yen, Yen, Chen, & Lin, Reference Ko, Liu, Yen, Yen, Chen and Lin2013; Kuss & Griffiths, Reference Kuss and Griffiths2012; Leeman & Potenza, Reference Leeman and Potenza2013; Lorenz et al., Reference Lorenz, Krüger, Neumann, Schott, Kaufmann and Heinz2013; Olsen, Reference Olsen2011; Park, Han, & Roh, Reference Park, Han and Roh2017). Third, the same molecular pathways (e.g., dopaminergic) that operate in substance addictions are implicated in internet addiction (Hou, Jia, Hu, Fan, Sun, Sun, & Zhang, Reference Hou, Jia, Hu, Fan, Sun, Sun and Zhang2012; Shaffer et al., Reference Shaffer, LaPlante, LaBrie, Kidman, Donato and Stanton2004; Tian et al., Reference Tian, Chen, Zhang, Du, Hou and Chao2014). Although the question of whether internet addiction should be recognized as an addictive disorder is still sometimes debated (Cash, Rae, Steel, & Winkler, Reference Cash, Rae, Steel and Winkler2012; Pies, Reference Pies2009; Ryding & Kaye, Reference Ryding and Kaye2018), here we adopt the emerging consensus that compulsive and excessive use of the internet is as much a true addiction as a substance addiction (Aboujaoude, Reference Aboujaoude2010; Grant et al., Reference Grant, Potenza, Weinstein and Gorelick2010; Hammond, Mayes, & Potenza, Reference Hammond, Mayes and Potenza2014; Kuss & Griffiths, Reference Kuss and Griffiths2012; Shaw & Black, Reference Shaw and Black2008; Spada, Reference Spada2014; Weinstein & Lejoyeux, Reference Weinstein and Lejoyeux2010; Young, Reference Young1998b).

A final issue with the term internet addiction is that, despite the fact that we have been using the term as if it refers to a single kind of addiction, the term seems actually to encompass several distinct addictions. Davis (Reference Davis2001) distinguishes between addiction to the internet in all of its forms, which we will call general internet addiction, and addiction to specific activities that are accessed through the internet, which we will call specific internet addictions, such as social networking, gaming, gambling, information searching, and accessing online pornography. Young (Reference Young1996, Reference Young1998b), one of the first to coin the term internet addiction, later argued that the term should be understood as an umbrella term encompassing five subtypes of specific internet addictions: cyber sexual addiction (cybersex and cyberporn), cyber relationship or social media addiction (online social interactions), net compulsions (gambling, shopping, or day trading), information overload (web surfing and information searching), and computer addiction (game playing) (Young, Pistner, O’Mara, & Buchanan, Reference Young, Pistner, O’Mara and Buchanan1999). In this article, our primary focus will be on the specific internet addiction that Young categorizes as cyber relationship addiction or what others have called social media addiction or social networking addiction (Kuss & Griffiths, Reference Kuss and Griffiths2011). A secondary focus of our discussion, however, will be directed at general internet addiction, particularly because much of the research on internet-related addiction has been on its generalized form.

General internet addiction has a surprisingly high prevalence among both adults and the young. In their meta-analysis of thirty-one international studies, Cheng and Li (Reference Li, Dang, Zhang, Zhang and Guo2014) estimated that 6 percent of the world’s population had become addicted to the internet. A survey by Durkee et al. (Reference Durkee, Kaess, Carli, Parzer, Wasserman and Floderus2012) found that about 4.4 percent of European adolescents were addicted to the internet, while Bányai et al. (Reference Bányai, Zsila, Király, Maraz, Elekes and Griffiths2017) found that 4.5 percent of Hungarian adolescents were addicted. Koukia, Mangoulia, and Alexiou (Reference Koukia, Mangoulia and Alexiou2014) found that the prevalence of internet addiction among Greek university students was 4.5 percent. Anderson (Reference Anderson2001) found that 9.8 percent of US college students who used the internet were addicted, while an online survey by Cooper, Morahan-Martin, Mathy, and Maheu (Reference Cooper, Morahan-Martin, Mathy and Maheu2002) found that 9.6 percent of US respondents were addicted. Thatcher and Goolam (Reference Thatcher and Goolam2005) estimated that as many as nine million Americans were addicted to the internet. Studies of Asian populations have found significantly higher prevalence rates than those of Western groups. The studies of Yen, Yen, Chen, Tang, and Ko (Reference Yen, Ko, Yen, Wu and Yang2007, Reference Yen, Yen, Chen, Tang and Ko2009) concluded that about 18 percent of Chinese high school students and about 12 percent of Chinese college students were addicted to the internet, while Ko et al. (Reference Ko, Yen, Yen, Yen, Chen and Yen2005) found that about 20 percent of Taiwanese adolescents were internet addicted. Internet addiction is clearly a large and global problem.

Although research on the epidemiology of social media addiction is not as mature as the research on general internet addiction, a few studies have attempted to look at how widespread social media addiction (as distinct from general internet addiction) is in the general population and among younger users. (While some of these studies rely on the Griffiths conceptualization of addiction, we note that others use overlapping but different diagnostic criteria.) Cabral (Reference Cabral2011), for example, surveyed 313 users of social media in the United States and found that 59 percent of them felt they were addicted to social media, while Olowu and Seri (Reference Olowu and Seri2012) surveyed 884 students in Nigeria and found that 27 percent of them felt that they were addicted; the studies of Cabral (Reference Cabral2011) and of Olowu and Seri (Reference Olowu and Seri2012), however, were based on self-reports. A study of Chinese college students by Wu, Cheung, Ku, and Hung (Reference Wu, Cheung, Ku and Hung2013) using a self-designed validated diagnostic instrument found that 12 percent of their sample was addicted. A study of young Peruvian subjects by Wolniczak, Cáceres-DelAguila, Palma-Ardiles, Arroyo, Solís-Visscher, and Paredes-Yauri (Reference Wolniczak, Cáceres-DelAguila, Palma-Ardiles, Arroyo, Solís-Visscher and Paredes-Yauri2013), also using a newly constructed validated diagnostic instrument, found that 8.6 percent of their sample were addicted to social media. A study of 1,870 Indian students using yet another validated diagnostic instrument found that 36.9 percent of social media users in the sample were addicted (Ramesh Masthi, Pruthvi, & Phaneendra, Reference Ramesh Masthi, Pruthvi and Phaneendra2018). Taken together, these studies suggest that social media addiction is an important problem, but because studies have relied on self-reports or newly developed and different instruments, it is difficult to say with precision how extensive the problem is. The lesson to be taken from the studies, however, is that social media addiction is a global issue that appears to be widespread among young people as well as adults.

While the specific mechanisms social media companies use in designing their platforms in ways that have rendered them addictive have changed over time, three of these design elements are common and worth pointing out: first, the use of intermittent variable rewards (or what is sometimes called the slot machine effect) (Griffiths, Reference Griffiths2018; Harris, Reference Harris2019; Williams, Reference Williams2018; Wu, Reference Wu2016); second, design features that take advantage of our desires for social validation and social reciprocity; and third, platform designs that erode natural stopping cues.Footnote ³ We describe these briefly here and will return to them in the next section, but we note that these are not the only addictive mechanisms internet companies use.Footnote ⁴

Wu (Reference Wu2016: 187), paraphrasing the cognitive scientist Stafford (Reference Stafford2006), notes that “the most effective way of maintaining a behavior is not with a consistent, predictable reward, but rather with what is termed ‘variable reinforcement’—that is, rewards that vary in their frequency or magnitude.” When a user opens the Twitter app, the user is brought to a blue loading screen. One might think that this loading screen must be due to slow hardware or internet; however, Morgans (Reference Morgans2017) notes that this delay in loading is yet another way to generate intermittent variable rewards. Pinterest takes a slightly different approach: “As the user scrolls to the bottom of the page, some images appear to be cut off. Images often appear out of view below the browser fold. However, these images offer a glimpse of what’s ahead, even if just barely visible. To relieve their curiosity, all users have to do is scroll to reveal the full picture. . . . As more images load on the page, the endless search for variable rewards of the hunt continues” (Eyal, Reference Eyal2014: 110). More generally, the “pull-to-refresh” feature seen in a number of social media platforms mimics the motion and variable reward schedule of a slot machine (Harris, Reference Harris2019; Williams, Reference Williams2018).

In addition to generating intermittent variable rewards, social media platforms have introduced reward schemes designed to take advantage of our desire for social validation and reciprocity, among other psychological tendencies and needs. One notable example is Snapchat’s use of “snapstreaks,” a running tally of the number of consecutive days a user has exchanged photographs or “snaps” with another user (Griffiths, Reference Griffiths2018). Teens often face immense pressure to maintain these streaks (Bosker, Reference Bosker2016). Similarly, Facebook’s “like” button taps into social reciprocity (and social validation); as Alter (Reference Alter2017: 128) notes, “it’s hard to exaggerate how much the ‘like’ button changed the psychology of Facebook use.” Most social media platforms have now introduced social reward schemes similar to the “like” button.

Third, the erosion of natural stopping cues is most prominently seen in the use of infinite scrolls (Harris, Reference Harris2019; Williams, Reference Williams2018). Prior to infinite scrolls, when a user arrived at the bottom of a webpage, there was a natural stopping cue—that is, the end of the page. The user at that point would have faced some decisions: whether to press the link to load the next page, whether to exit the platform, and so on. Introducing infinite scrolls removed the opportunity to make such decisions. Now, as the user scrolls, the platform automatically populates the next page, thereby removing stopping cues that would have previously given the user the opportunity to reflect, even for a moment, on whether that user should continue using the platform.

Crucially, the more that users spend time on social media platforms, the more data social media companies have about what works and what does not, which in turn allows them to further refine their platforms. As Alter (Reference Alter2017: 4) puts it, “the people who create and refine tech . . . run thousands of tests with millions of users to learn which tweaks work and which ones don’t—which background colors, fonts and audio tones maximize engagement and minimize frustration. As an experience evolves, it becomes an irresistible, weaponized version of the experience it once was. In 2004, Facebook was fun; in 2016, it’s addictive.”

Given the prevalence of social media addiction (and of internet addictions more broadly) and the ethical significance of the issues raised by addicting users to social media (issues we discuss in detail later), it is clear that social media addiction is a serious problem that managers of social media companies (as well as policy makers, public health officials, educators, and parents) would do well to address.

2. THE IMPERMISSIBILITY OF MAKING SOCIAL MEDIA ADDICTIVE

The research and related literatures on internet addiction, on balance, then, would seem to suggest that there is a substantial social media addiction problem, and one that gives rise to various harms. We turn now to making three distinct, but related, moral arguments about this problem. First, we argue that in light of the kinds of harms associated with internet addictions, it is wrong to use social media platforms to addict users, and these harms are not justified by the benefits those technologies produce. Second, we argue that users of social media platforms are injured in a way that is demeaning, thereby adding insult to the injury. Third, we argue that addicting users to social media constitutes a particularly objectionable form of exploitation. These arguments, we believe, show not only that it is wrong to design social media platforms that addict users but also why it is wrong.

3. A BUSINESS THAT INCENTIVIZES WRONGDOING

Many kinds of businesses (both technology and nontechnology businesses) provide products that addict their users. But addiction is merely a contingent feature of the business model of most of them. For example, a cigarette company would not object if a customer bought its product and threw it in the garbage, used the cigarettes to build model bridges, transformed them into modern art, or used the product in any other way apart from smoking, so long as the customer continued to purchase the product.Footnote ¹⁹ In other words, the cigarette company would be indifferent to whether a customer ever actually smoked its cigarettes, as long as its revenues continued to flow at the same or an increased rate.Footnote ²⁰

Something similar is true even for some addictive technology products that do not have an attention-economy business model. For example, consider subscription-based digital streaming services (e.g., Netflix): the contemporary popularity of the term binge watching is perhaps in large part due to such services. But so long as their customers purchase or renew their memberships, it is immaterial to these subscription streaming services whether or not they binge watch a given television series. This is not to say that these subscription-based streaming services do not employ mechanisms that render their platforms addictive: automatically rolling over into the next episode is a feature designed to keep users on the platform by eliminating natural stopping cues (e.g., having to end an episode and click into a new one). But the point is that it is not a necessary feature of the business model of companies with subscription-based streaming services that customers continue to watch the companies’ shows. As long as customers renew or purchase their memberships, their failure to binge watch is not a significant problem for these companies. Perhaps it is even beneficial to subscription-based streaming services; assuming a company pays royalties on a per use basis, the company could lower its costs, and it would perhaps even be able to narrow its bandwidth infrastructure costs. To be clear, we are not saying that all of its customers would continue to buy and renew their subscriptions to these streaming services if they did not find the content addictive; rather, we are pointing out that making a platform addictive is not an essential feature of the subscription-based streaming service business model.

But attention-economy businesses—of which social media businesses are the paradigmatic example and our primary focus—have a business model that exhibits an important difference: it hinges on keeping users active on a platform for prolonged periods of time. The longer a user is active and engaged on a social media platform, the more profitable it is for the social media company. This is because the longer the user remains engaged with the platform, the more likely it is that this user will be exposed to, influenced by, and engaged with advertisements, and so the more the social media company can charge its advertisers (Lanier, Reference Lanier2018; McNamee, Reference McNamee2019; Price, Reference Price2018). Users of social media, unlike users of cigarettes, alcohol, or junk food, are not the source of the companies’ revenues. The revenues of social media companies come from advertisers, not users. As the familiar slogan goes, with social media, you are not the customer, you are the product. Footnote ²¹ Thus built into the business model of social media is a strong incentive to keep users online for prolonged periods of time, even though this means that many of them will go on to develop addictions (Alter, Reference Alter2017; Price, Reference Price2018). And, as we have argued, the significant harms of social media addiction have a temporal dimension: they are primarily related to the amount of time the person who becomes addicted spends on social media.Footnote ²² Given the arguments from the previous section—that addicting users to social media is impermissible because it inflicts unjustified harms in a way that is demeaning and objectionably exploitative—social media businesses have a strong incentive to engage in wrongdoing.

To be clear, we are not claiming that ad-based businesses are the only ones with a strong incentive to capture their customers’ attention for as long as possible. We are arguing, rather, that insofar as a company’s business model is an attention-economy business model (of which the ad-based models of social media companies are a paradigmatic example), this model generates a strong incentive to design websites in ways that addict users. The prime incentive for an attention-economy internet business is to get its users to devote prolonged periods of time to its website, and devoting their time to the website, for those who subsequently become addicted, is a primary source of the harms associated with their addiction.

The attention-economy business model is not novel, of course: both radio and TV programming have long run on such a model (Wu, Reference Wu2016). Worries about addiction to radio and TV were also raised when these technologies first came to market (Meerloo, Reference Meerloo1954; Sussman & Moran, Reference Sussman and Moran2013). And as ad-based attention-economy businesses, TV and radio also have an incentive to addict. But, as we have argued, technologies such as adaptive algorithms allow social media companies to target and continuously maximize their addictive potential at the individual level in ways that radio and television, currently, at least, cannot do.

Not all social media users, of course, become addicted, and there are a variety of reasons why particular individuals are vulnerable to becoming addicted to various behaviors, including genetics, environmental factors, and individual vulnerabilities (Browne et al., Reference Browne, Hing, Rockloff, Russell, Greer and Nicoll2019; Kim & Hodgins, Reference Kim and Hodgins2018). But a particularly important cause of internet addiction in general, and social media addiction in particular, are the design elements that internet companies embed in their platforms (Alter, Reference Alter2017; Price, Reference Price2018). Some of the design elements that addict users to internet platforms were originally developed by engineers who drew on behavioral psychology to keep gamblers seated before the computerized slot machine monitors (electronic gambling machines or EGMs) that have largely replaced other forms of gambling in casinos (Abbott, Reference Abbott2017; Breen & Zimmerman, Reference Breen and Zimmerman2002; Schüll, Reference Schüll2014). According to Yücel, Carter, Harrigan, van Holst, and Livingstone (Reference Yücel, Carter, Harrigan, Holst and Livingstone2018: 20), these EGMs “are intentionally designed with carefully constructed design elements … that modify fundamental aspects of human decision-making and behaviors, such as classical and operant conditioning, cognitive biases, and dopamine signals.” Having been developed in the gambling industry, it was an easy step to adapt these design techniques to early computer games and then to the design of internet websites (Alter, Reference Alter2017: 136–39; Courtwright, Reference Courtwright2019). But web design researchers have gone on to develop new addicting technologies that combine engineering and behavioral psychology to ensure website users will be “persuaded” to behave as the designer wants (Alter, Reference Alter2017; Fogg, Reference Fogg2003; Lanier, Reference Lanier2018; Price, Reference Price2018). Engineers trained in university programs where such techniques are researched, developed, and taught (e.g., at the Persuasive Technology Lab of Stanford University, located in Silicon Valley) are hired by Silicon Valley companies to use those techniques—for example, variable reinforcement (Cash et al., Reference Cash, Rae, Steel and Winkler2012)—to design websites that entice users to remain engaged for ever longer periods of time (Andersson, Reference Andersson2018; Leslie, Reference Leslie2016; Simone, Reference Simone2018) and that ultimately addict many of them (Alter, Reference Alter2017; Lanier, Reference Lanier2018; Young & de Abreu, Reference Young and Abreu2011). Users who are vulnerable become “hooked” (Eyal, Reference Eyal2014). The morally significant harms suffered by those who become addicted to social media, then, are the result of the design decisions of people who work in the companies that own and create those platforms.

Our aim has not been to argue that social media firms intentionally addict their users; rather, we have focused exclusively on characterizing the moral dimensions of the act of social media firms designing platforms in ways that result in addicting users (whatever the intentions might have been of the people designing these platforms). Making claims about any given agent’s intentions with respect to an action requires a kind of evidence that we have not aimed to provide. (We may note, however, that there is a plausible case to be made that some individuals at social media companies at least knew that users would become addicted and thus harmed by their website designs.)Footnote ²³ Our chief aim in this section has been to discuss some of the distinctive incentives created by the business model of social media companies and to highlight how these incentives—particularly the incentive to monopolize the user’s time—have led to the time-related harms inflicted on those who become addicted to social media.

4. IMPLICATIONS FOR THEORY AND PRACTICE

Some implications for both theory and practice are now worth noting. Several decades ago, the importance of not divorcing business ethics and engineering ethics was most prominently raised by the case of Ford Pinto (Danley, Reference Danley2005). Our article provides further support for why certain issues in design ethics and engineering ethics are not only of tangential relevance to business ethics but themselves raise distinctly business ethics issues. While the specific issues of how exactly to design a social media platform may turn on questions of engineering and its ethics, many of these decisions are made by the company’s managers and are prompted by the incentive structure of the company. Scholars should not see engineering ethics questions as divorced from business ethics, and vice versa.

Second, well-intentioned teachers often encourage children to become more tech savvy, with an eye to preparing them for college and beyond. But this emphasis on technology in K–12 education, given the high addictive potential of social media, should be carried out with full awareness of its costs. More than that, it is not self-evident that a more technologically advanced class is a more pedagogically advanced class. Nor is it obvious that use of technology allows our children to become better—however we may understand this term—graduates or even citizens of our communities. Indeed, the founders and executives of many Silicon Valley tech companies—employees of the very firms that create the most addictive platforms—have opted to send their children to low-tech schools that do not integrate computers, tablets, or other electronic devices into their curricula (Archibald, Reference Archibald2018).Footnote ²⁴

Furthermore, although much research has focused on the so-called digital divide (the disparity in access to the technology needed for educational and professional success between low- and high-income communities) (Rideout & Robb, Reference Rideout and Robb2019; van Dijk, Reference Dijk2006), there is a different kind of digital divide—call it the digital use divide—where teens in low-income communities are exposed to nearly two hours more per day of screens than teens in wealthier communities (Rideout & Robb, Reference Rideout and Robb2019). While it is important to ensure that children of low-income communities have access to the resources required for educational and professional success, understanding the digital use divide takes on added importance, given the potential to addict.

Third, the design features firms use to make their platforms more addictive could be used for the opposite purpose: to empower users to have a healthier relationship with social media. Importantly, these are fixes tech companies could implement with relative ease. For example, Apple has implemented features into its most recent iOS operating system to alert the user to his or her phone’s usage statistics (e.g., hours spent on the device, number of instances a user turned on his or her phone). Something similar could be done for social media. Harris (Reference Harris2018) suggests other helpful design features, including alerting users to the estimated time they would spend were they to log in to a given website, alerting them to how long ago they logged in, and more. These suggestions are akin to the sorts of suggestions Sunstein and Thaler (Reference Sunstein and Thaler2008) make in their discussion of nudges: in the same way that use of an opt-out on an organ donor form dramatically increases the number of donors, social media companies could assume that users opt out of the use of addictive aspects of technology unless they explicitly opt in (see also Goldstein, Johnson, Herrmann, & Heitmann, Reference Goldstein, Johnson, Herrmann and Heitmann2008).

Fourth, insofar as social media firms continue to render their products more addictive, this fact should be made plain to their users. This is especially so given that, as we discussed, the platforms themselves are increasing in addictive potential due to the use of adaptive algorithms. Imagine if every time you bought coffee from your neighborhood café, the coffee, without your knowledge, spiked in addictiveness. This, obviously, would be troubling. But it might be made less bad if the café were to tell you that it would increase the addictiveness of your coffee each time you purchased coffee there. Tech firms, similarly, owe it to their users to make it clear that they are employing the users’ usage data in ways that will not only make the experience better but may also make the platform more addictive. Moreover, doing so would help to lessen the force of the insult we discussed in section 2.2 on adding insult to injury. In short, even if technology firms continue to addict users, they ought to be transparent to users about the ways in which they are incorporating tools of behavioral psychology to design mechanisms that may elicit addiction.

Fifth, we should take seriously the possibility of ridding ourselves of social media until it takes on a form with a dramatically changed incentive structure and is designed to empower users in their decisions regarding its use (Lanier, Reference Lanier2018; Newport, Reference Newport2019). And policy makers have an important role to play.Footnote ²⁵ Although it is unlikely that policy makers would (or even should) pursue measures as drastic as prohibiting social media, policy makers should lower the barriers users face to exit social media. If a user wants to quit Facebook, for example, this is the process the user must go through at present. First, the user must click an unlabeled down arrow at the top right of the screen, and then “Settings.” After doing so, the user is presented with a menu of thirteen options, including “Privacy,” “General,” “Security and Login,” “Your Facebook Information,” and “Blocking.” The user would need to know that the correct option to choose is “Your Facebook Information.” Once that is selected, the user is brought to a menu of an additional five options, one of which says “Deactivate and Delete.”

One might think the process is now complete. Not quite. The user is then presented with a choice to “Deactivate Account or Permanently Delete”—with the default, preselected option being the former. Suppose the user selects “Deactivate Account.” After doing so, the user is brought to a different page, at the top of which it asks, “Are you sure you want to deactivate your account?” followed by five algorithmically curated photographs of that user’s friends. Above each photograph, it notes that the friend will miss the user; so, for example, above Anjali’s photo, it says “Anjali will miss you,” along with a prompt to send Anjali a message (which would then thwart the deactivation process).

Suppose the user remains on course. Facebook next requires the user to select from one of ten reasons for leaving. Each of the listed reasons generates a pop-up window. For example, suppose the user’s selected reason for leaving is “I spend too much time using Facebook”; that option pops up a window with the following response: “One way to control your interaction with Facebook is to limit the number of emails you receive from us. You can control what emails you receive [by clicking this link].” Let’s suppose the user exits this window and continues the deactivation process. Once the user selects his or her reason for deactivating, the user must then select a further box to opt out of receiving future emails from Facebook (and select whether the user wants to keep using Facebook’s messenger platform). Suppose the user opts out of receiving emails from Facebook and declines to continue using Facebook’s messenger platform. Now the user can press the deactivate button. After doing so, once again, a notification pops up asking whether the user is sure about wanting to deactivate. If the user selects selects in the affirmative, the user can then press “Deactivate Now,” which will conclude the deactivation process.

Suppose that, after deactivating, the user wants to return to Facebook. What must the individual do to reactivate? Facebook states, “You can reactivate your Facebook account at any time by logging back into Facebook or by using your Facebook account to log in somewhere else.” That is, simply log back in. Given that many users habitually log in, some users may inadvertently log in. And once a user has logged back in (inadvertently or not), if the user wants to deactivate, the user must restart the entire deactivation process.

Suppose that, instead of deactivating, the user wants to permanently delete. The user needs to go through the same process discussed earlier for deactivating, but the user is instead ultimately brought to a screen that notes, “Your account is scheduled for permanent deletion. Facebook will start deleting your account in 30 days.” If at any point during that thirty-day window the user inadvertently logs back in, the deletion is canceled, and the user must begin the process all over, with a renewed thirty-day period. This recommendation for policy makers is thus a simple one: require lower barriers to exit.