Examples of negative tech innovations include Facebook’s filter bubble, autonomous vehicles, and predictive policing.

This Article highlights an underappreciated facet of technology's harmful effects: many technologies that generate negative externalities are patented. In an unsettling irony, the patent system significantly subsidizes the development of technologies that impose uncompensated harms on third parties. To better understand this phenomenon, this Section turns to case studies of three patented technologies that produce substantial negative externalities.

1. Facebook's "Filter Bubble" and Ideological Polarization

One prominent example of a patented technology generating negative externalities is Facebook's "filter bubble."72 Over the past two decades, Facebook has developed innovations to personalize content for users based on their interests and connections. This so-called filter bubble technology has several benefits, including enhancing the relevance of content to users, increasing the stickiness of Facebook's website, and raising Facebook's value to advertisers. However, the filter bubble also generates significant harms, particularly in the political sphere. Curated content streams spread misinformation and establish incommensurable "media ecosystems" that expose liberal and conservative users to different universes of information.73 Facebook's algorithms prioritize content linked to "extreme" reactions or numerous posts, which tend to come from a small but vocal fraction of 74 users. Filter bubbles exacerbate political polarization 75 and have even been blamed for contributing to the January 6, 2021, U.S. Capitol insurrection.76 Over two dozen U.S. states have filed suit against Facebook for alleged antitrust violations, some of which relate to Facebook's information-filtering practices.77

While it is difficult to quantify the harms of Facebook's filter bubble, they are likely substantial. Facebook is the dominant source of news and information for many of its users. Filter bubbles foster confirmation biases by providing users with information that supports their existing views and shielding them from opposing perspectives.79 The resulting deterioration of a shared epistemic commons degrades democratic discourse and can lead to political polarization and even violence. Foreign agents have also exploited filter bubbles to spread misinformation; Russia created Facebook pages touting fraudulent information that received 340 million engagements before Facebook deleted them.80 Notwithstanding these harms, Facebook benefits from filter bubbles and has incentives to expand this technology. In the wake of the 2020 U.S. presidential elections, Facebook temporarily altered its algorithms to promote more fact-based information sources, but it soon reversed that decision.81

Notably, Facebook's filter bubble is patented technology. Among Facebook's 9,000 patents are several claiming inventions that 8 personalize content based on a user's interests and connections. 2 Take, for example, U.S. Patent No. 9,110,953 B2. This patent covers methods and systems for "Filtering Content in a Social Networking Service," and 83 it lists Mark Zuckerberg as one of its inventors. The patent claims methods and systems by which an algorithm considers various user 84 attributes to deliver highly relevant content. The patent consists of sixteen pages of technical diagrams, disclosures, and claims that describe the system's operation and various preferred embodiments. It discloses several benefits of this invention, which personalizes content for users and enables more targeted advertising.85 However, the patent does not disclose any mechanism for verifying the information provided to users via this algorithm. Furthermore, it contains no discussion of this technology's wider social ramifications, such as its potential contribution to ideological polarization.

2. Autonomous Vehicles and Job Losses

Autonomous vehicles also generate significant negative externalities. Of course, self-driving vehicles promise significant benefits, such as reducing accidents, enhancing energy efficiency, and increasing mobility for those unable to drive. Alongside these benefits, however, this technology creates external harms for which technology developers are largely unaccountable. For instance, several Tesla cars using its semi-autonomous "Autopilot" driving technology have killed people.86 Lawsuits against Tesla allege design defects; while such lawsuits may force Tesla to internalize some of these costs, Tesla has vigorously denied liability, thus seeking to maintain these costs as externalities. Autonomous vehicles also raise privacy concerns by gathering enormous amounts of data and potentially sharing that data with government entities and third parties.87 At the heart of autonomous vehicles is Artificial Intelligence (AI), which gives rise to a host of negative externalities. As Tesla chief Elon Musk has warned, AI could become "an immortal dictator from which we can never escape."88

One significant negative externality of AI-based automation in general, and autonomous vehicles in particular, is job losses.89 The threat of automation to replace human workers has attracted significant public policy concern. The landscape is complex, as automation can enhance worker productivity and increase demand for skills in some contexts. In other contexts, however, automation eliminates jobs.91 Lower-paid, lower-skilled, and less-educated workers are most susceptible to job losses.92 Focusing on autonomous vehicles, it is estimated that when self-driving cars and trucks reach saturation, job losses among U.S. drivers will amount to 25,000 per month or 300,000 per year. The White House Council of Economic Advisors estimates that autonomous vehicles threaten 2.2 to 3.1 million driving-based jobs in the United States. 94

Many of the technologies underlying autonomous vehicles are patented.95 An informal search revealed over 135,000 patents or patent applications in the United States mentioning the term "autonomous vehicles."96 On a massive scale, patent law is encouraging the development of technologies that promise significant benefits but that also impose substantial harms. Take, for example, U.S. Patent No. 9,244,462 B2, which covers "vehicle trajectory planning for autonomous vehicles."97 This patent, owned by Nissan, claims methods for determining an initial vehicle trajectory and then calculating and executing an optimized vehicle trajectory based on changed parameters. Among other benefits, this method can "smooth out" abrupt changes in direction, such as U-turns, lane changes, and turns at intersections. In doing so, it reduces "occupant discomfort, motion sickness, and decreased confidence in the autonomous system."98 Not surprisingly, however, amid twelve pages of technical disclosures, there is no mention of the broader social impact of this technology and potential job losses that may arise from autonomous vehicles.

3. Big Data and Government Overreaching

A final example of negative externalities from patented technologies involves methods for analyzing enormous amounts of data, commonly known as big data.99 This innovation has numerous applications, from allowing Target to anticipate consumers' purchasing needsloo to enabling health officials to track transmissions of coronavirus from people's cell phones.101 One of the most significant government uses of big data is to predict the occurrence of crime. As Professor Elizabeth Joh has chronicled, law enforcement authorities are using big data and predictive analytics to forecast when and where 10 2 crime will occur. Such predictions allow officials to optimize the allocation of law enforcement resources to various geographic areas. Notably, however, such technology may enable government overreaching that imposes unaccounted-for harms on third parties.o3

In particular, predictive policing based on big data may violate Fourth Amendment safeguards against unreasonable searches and seizures. While standard investigative detentions require at least reasonable suspicion based on a totality of the circumstances,104 it is unclear if predictive analysis from big data can support such determinations.105 Furthermore, as Professor Joh argues, predictive policing systems can perpetuate bias: "[R]eliance on arrest rates is surely problematic because arrests themselves are discretionary decisions that, if used as the basis to justify more attention, may simply reinforce unjustified police stereotypes that certain neighborhoods need heavier police attention."106 Additionally, in borderline cases, police officers may over-rely on prediction models based on the perceived objectivity of the algorithms.107 In short, predictive policing technology can generate negative externalities in the form of biased and 08 excessive arrests.1 Predictive policing technology, which generates negative externalities, is patented. One of these patents is U.S. Patent No. 8,949,164 B1, which covers a "predictive policing system" that uses a crime prediction server, historical crime data, and crime forecasting 9 algorithm to forecast crimes for at least one geographic region.10 Notably, this patented system incorporates the "broken windows" theory of policing. This theory holds that "untended" behavior, such as broken windows and broken-down cars, can undermine community controls.110 (The predictive policing system literally prompts police officers to "scan for broken glass.")'I This inclusion is troubling in light of recent research debunking the broken windows theory and linking it to increased incidences of stop and frisk, racial profiling, and police misconduct.112 The patent discusses several benefits of the invention, which can help reduce crime and improve the allocation of law 3 enforcement resources." At no point, however, does the patent mention the potential harms of this technology, such as the possibility of biased predictions of criminal activity, or offer any correctives. As with the other patented technologies profiled here, inventors benefit from highly valuable exclusive rights, but they are not accountable (in the patent system) for the harms their inventions impose on third parties.