Fixing Social Media’s Grand Bargain
Jack M. Balkin 

Aegis Series Paper No. 1814

To regulate social media in the twenty-first century, we should focus on its political
economy: the nature of digital capitalism and how we pay for the digital public sphere
we have. Our digital public sphere is premised on a grand bargain: free communications
services in exchange for pervasive data collection and analysis. This allows companies
to sell access to end users to the companies’ advertisers and other businesses.
The political economy of digital capitalism creates perverse incentives for social media
companies. It encourages companies to surveil, addict, and manipulate their end users
and to strike deals with third parties who will further manipulate them.
Treating social media companies as public forums or public utilities is not the proper cure.
It may actually make things worse. Even so, social media companies, whether they like it
or not, have public obligations. They play important roles in organizing and curating public
discussion and they have moral and professional responsibilities both to their end users
and to the general public.
A reinvigorated competition law is one important way of dealing with the problems of social
media, as I will describe later on. This essay, however, focuses on another approach: new
fiduciary obligations that protect end-user privacy and counteract social media companies’
bad incentives.

How Do We Pay for the Digital Public Sphere?
How does the political and economic system pay for the digital public sphere in our Second
Gilded Age?1 In large part, it pays for it through digital surveillance and through finding
ever new ways to make money out of personal data.
Twenty-first-century social media like Facebook or YouTube differ from twentieth-century
mass media like broadcast radio and television in two important respects. First, they are
participatory, many-to-many media. Twentieth-century broadcast media are few-to-many:
they publish and broadcast the content of a relatively small number of people to large
audiences. In the twentieth century, most people would never get to use these facilities
of mass communication to speak themselves. They were largely relegated to the role of
audiences.

National Security, Technology, and Law

A HOOVER INSTITUTION ESSAY

 2

Twenty-first-century social media, by contrast, are many-to-many: they depend on mass
participation as well as mass audiences. They make their money by encouraging enormous
numbers of people to spend as much time as possible on their platforms and produce
enormous amounts of content, even if that contribution is something as basic as commenting
on, liking, or repeating somebody else’s contribution. Facebook and Twitter would quickly
collapse if people didn’t constantly produce fresh content. Search engines, which are key parts
of the digital infrastructure, also depend on people creating fresh links and fresh content that
they can collect and organize.
Second, twenty-first-century social media like Facebook, YouTube, and Instagram rely
on far more advanced and individualized targeted advertising than was available to
twentieth-century broadcast media. Television and radio attempted to match advertisers
with viewers, but there were limits to how finely grained they could target their audiences.
(And newspapers, of course, relied on very broad audiences to sell classified advertisements.)
What makes targeted advertising possible is the collection, analysis, and collation of personal
data from end users. Digital communication leaves collectible traces of interactions, choices,
and activities. Hence digital companies can collect, analyze, and develop rich dossiers of
data about end users. These include not only the information end users voluntarily share
with others, but their contacts, friends, time spent on various pages, links visited, even
keystrokes. The more that companies know about their end users, the more they know
about other people who bear any similarity to them, even if the latter spend less time on
the site or are not even clients. In the digital age, we are all constantly informing, not only
on ourselves, but on our friends and relatives and, indeed, on everyone else in society.
This is not only true of social media, but of a wide range of digital services. The publisher
of a paperback book in the 1960s could tell little about the reading habits of the people
who purchased it, while Amazon can tell a great deal about the reading habits of the
people who use their Kindle service, down to the length of time spent, the pages covered,
the text highlighted and shared, and so on. As the Internet of things connects more and
more devices and appliances to digital networks, surveillance spreads to ever more features
of daily interaction. In general, the more interactive and the more social the service, the
greater the opportunities for data collection, data analysis, and individualized treatment.
Data collection and analysis allow targeted advertising, which allows more efficient advertising
campaigns, which allow greater revenues. But data collection and analysis offer another
advantage: in theory, they give social media opportunities to structure and curate content
for end users that they will find most engaging and interesting. That is important because
advertising revenues depend on the amount of time and attention spent on the site. More
engaging content means more time spent and more attention gained.

Jack M. Balkin  •  Fixing Social Media’s Grand Bargain 

 3

Social media companies have economic incentives to develop algorithms that will promote
content that engages people. That is because companies’ central goal is to gain attention
share. This leads them to collect ever more data about their end users so that they can tailor
content to individual end users to maximize their emotional engagement.2
This creates a problem. Often what engages people the most is material that produces strong
emotional reactions—even if it is polarizing, false, or demagogic. Companies have economic
incentives to expose people to this material. And unscrupulous actors, both domestic and
foreign, have learned to take advantage of this feature of social media. As a result, the same
business model that allows companies to maximize advertising revenues also makes them
conduits and amplifiers for propaganda, conspiracy theories, and fake news.3

The Digital Grand Bargain and its Problems
Social media business models are a special case of the grand bargain that has made the
digital public sphere possible in our Second Gilded Age. The bargain goes something like
this: We will give you miraculous abilities. We will give you social media that allow you to
connect with anyone, anywhere, anytime, in a fraction of a second. We will give you search
engines that find anything you are looking for instantaneously. We will give you new forms
of entertainment that are absorbing, engaging, outrageous, and amusing. We will give you
ever more ways to measure yourself and express yourself to others.
We will give all of this to you, for free! And in return, you will let us surveil you. You will
let us collect and analyze your habits, your locations, your links, your contacts with your
friends, your mouse clicks, your keystrokes, anything we can measure. We will gladly take
all of that and study it, and draw inferences from it, and monetize it, so that we can give
you all these miraculous things. And we will use that data to perform experiments on you
to figure out how to keep you even more focused on our sites and our products, so that you
can produce even more data for us, which we can monetize.
This is the grand bargain of the Second Gilded Age. This is twenty-first-century data
capitalism. And this is also the irony of the digital age: an era that promised unbounded
opportunities for freedom of expression is also an era of increasing digital surveillance and
control. The same technological advances allow both results. The infrastructure of digital
free expression is also the infrastructure of digital surveillance.
What is objectionable about this grand bargain? The most obvious objection is that we must
surrender individual privacy in order to speak. We must make ever more detailed portraits
of our lives available to social media companies and their business partners. Beyond this,
however, lies a deeper concern: the potential for abuse of power. In particular, the digital
grand bargain creates an increased danger of manipulation—both by social media companies
and by those who use social media companies—that is of a different degree and kind than

Hoover Institution  •  Stanford University

 4

that which existed in the pre-digital era. By “manipulation” I mean techniques of persuasion
and influence that (1) prey on another person’s emotional vulnerabilities and lack of
knowledge (2) to benefit oneself or one’s allies and (3) reduce the welfare of the other person.4
(Successful manipulation can also have ripple effects on third parties, such as family members
and friends, or even fellow citizens.)
The problem with the current business models for social media companies such as Facebook,
Twitter, and YouTube is that they give companies perverse incentives to manipulate end
users—or to allow third parties to manipulate end users—if this might increase advertising
revenues, profits, or both.
Manipulation is not a new problem. In the past, businesses have often appealed to people’s
emotions, desires, and weaknesses and have taken advantage of their relative ignorance.
So have demagogues and political con artists. But the digital world of social media amplifies
the opportunities for manipulation, both by social media companies and by those who use
social media to reach end users.
The digital age exacerbates the twentieth-century problem of manipulation in several
important respects. First, there is the issue of individual targeting. Twentieth-century
influence campaigns were usually aimed at broad groups of individuals, with effects that
were often hit-or-miss. With digital technologies it is now possible to tailor influence
campaigns to individuals or to very small groups. Instead of appealing to the general
emotional vulnerabilities of the public or the vulnerabilities of large demographic groups,
digital companies can increasingly target the specific vulnerabilities and emotional hot
buttons of individuals who may not be aware of precisely how they have been singled out.
Second, there are differences in scale, speed, and interactivity. Digital technologies allow
individualized messages to be targeted to vast numbers of people simultaneously, something
that was not possible with twentieth-century media. Moreover, end users’ responses can be
collected instantaneously, allowing companies to continually fine-tune their approaches,
speeding up the Darwinian evolution of the most successful influence strategies. On top of
this, digital companies now have the ability to perform interactive social science experiments
on us to perfect their abilities to leverage and control our emotions. Facebook, for example,
performed experiments to manipulate the emotional moods of 700,000 end users without
their knowledge.5 It has also experimented with ways of encouraging people to vote. But
such techniques might also be used to discourage people from voting.6 Moreover, these
experiments can affect the behavior of not only end users but also those they come into
contact with.7
Third, there is the problem of addiction. The more digital companies know about people’s
emotional vulnerabilities and predispositions, the more easily they can structure individual
end-user experience to addict end users to the site.8 Social media leverage the data they

Jack M. Balkin  •  Fixing Social Media’s Grand Bargain 

 5

collect about end users to offer periodic stimulation that keeps users connected and
constantly checking and responding to social media. Media have always been designed
to draw people’s attention, but the digital experience can be especially immersive and
pervasive, and thus a more powerful lure than a billboard or magazine advertisement.
Here once again, the digital age far outstrips the powers of twentieth-century media.
One might object that, despite all this, the digital grand bargain remains freely chosen
and welfare-enhancing. End users are free to use or not to use social media, and thus they
are free to decide whether they will subject themselves to experimentation and emotional
manipulation. If the free service is sufficiently valuable to them, they will accept the bargain.
But this overlooks three important features of the emerging system of digital surveillance
that make the assumption of a mutually beneficial arm’s-length bargain highly implausible.
First, we cannot assume that transactions benefit both parties when there is extreme
asymmetry of knowledge, in which one party’s behaviors, beliefs, and activities are known
to the other party while the other party is essentially a black box.
Second, individuals suffer from privacy myopia, a characteristic feature of digital interactions.9
Individuals constantly generate a broad range of information about themselves through digital
interactions, much of which (for example location, social connections, timing of responses,
and rate of keystrokes) they may be only dimly aware. Individuals have no way of valuing or
assessing the risks produced by the collection of particular kinds of information about them
or how that information might be employed in the future. That is because the value of such
information is cumulative and connective. Information that seems entirely irrelevant or
innocuous can, in conjunction with other information, yield surprisingly powerful insights
about individual values, behavior, desires, weaknesses, and predispositions. Because individuals
cannot assess the value of what they are giving up, one cannot assume that their decisions
enhance their welfare. In this environment, the idea of relying on informed consumer choice
to discipline social media companies is a fantasy.
Third, as noted above, information gathered from end users has significant external effects
on third parties who are not parties to the bargain. As digital companies know more about
you, they also can learn more about other people who are similar to you or connected to
you in some respect.10 In the digital age, we do not simply inform on ourselves; we inform
on other people as well. And when a social media company experiments with social moods
or engineers an election, it affects not only its end users but many other people as well.
For all these reasons, it is fatuous to compare the digital grand bargain to a mutually beneficial
arm’s-length economic transaction. If we can pay for digital freedom of expression while
reducing the dangers of digital manipulation, it is worth exploring alternatives.

Hoover Institution  •  Stanford University

 6

Public Options
Proposals for reform of social media abound these days. One kind of proposal argues that
we should counter the power of social media and search engines by treating them as state
actors. Courts should apply standard First Amendment doctrine to them and treat them as
public forums, which require complete content and viewpoint neutrality. If social media
cannot choose what we see, they cannot manipulate us.
This solution fails to grapple with the central problems of the grand bargain. First, treating
social media as public forums would only affect the ability of social media themselves to
manipulate end users. It would do nothing to prevent third parties from using social media
to manipulate end users, stoke hatred, fear, and prejudice, or spread fake news. And because
social media would be required to serve as neutral public forums, they could do little to stop
this. Second, even if social media do not curate feeds, they still collect end-user data. That
end-user data, in turn, can be harvested and sold to third parties, who can use it on the site
or elsewhere. (That is why, for example, requiring social media companies to offer a tiered
service in which people pay not to receive commercial advertisements does not really deal
with the underlying problem of surveillance, data collection, and manipulation.)
Perhaps equally important, the proposal is unworkable. Social media—and search engines—
must make all sorts of editorial and curational judgments that the First Amendment forbids
government entities to make.
For example, social media sites might want to require that end users use their real names
or easily identifiable pseudonyms in order to limit trolling and abuse. They might decide to
ban hate speech or dehumanizing speech, especially if they operate around the world. They
might choose to ban graphic violence, nudity, or pornography. They might choose to ban
advocacy of violence or illegal conduct, or the promotion of suicide. They might decide to
ban certain types of harassment or incitement even if that harassment or incitement does not
immediately lead to a breach of the peace.11 They might ban certain forms of advertising. All
of these regulations would be unconstitutional if a government imposed them in a public
forum. More generally, we should accept that social media will have to make sometimes
quite complicated decisions to discipline abusive trolls, maintain civility norms, demote the
ranking of postings by conspiracy theorists and hate mongers, and, in cases of serial abuse,
terminate accounts. Many of these policies would be unconstitutional if we applied the
same standards to social media that the First Amendment applies to municipal streets and
parks.
At a more basic level, it is impossible to manage a search engine or a social media site without
curation, which involves a wide range of content-based judgments about what content to
promote and what to demote.12 It is also impractical and self-defeating to manage a social
media site without moderation, which requires the imposition of a wide range of civility

Jack M. Balkin  •  Fixing Social Media’s Grand Bargain 

 7

rules that the First Amendment forbids governments from imposing in public discourse.
Moreover, creating individualized social media feeds and search engine results inevitably
requires content-based judgments. As described below, social media and search engines
sometimes make bad decisions about these matters, but the solution is not to impose a set
of doctrinal rules crafted for municipal streets and parks.
A second, related proposal argues that we should treat social media sites and search engines
as public utilities because they perform what are clearly public functions. But public utility
regulation—for example, of water and power utilities—generally focuses on two issues:
access to essential services and fair pricing. Neither of these is particularly relevant. Social
media and search engines want everyone to participate and they offer their services for free.
If the goal of the public utility metaphor is to prevent content discrimination, it faces the
same problems as treating digital media as state actors.
A third and quite different approach is public provisioning. Instead of treating existing
private companies as arms of the state, governments could provide their own public options:
government-run social media and search engines. For reasons stated above, these would
not really work very well if they had to be organized as public forums and moderation
was forbidden. There are potential solutions, however. The government could provide
only a basic telecommunications system for social media messages and then allow various
groups and businesses to create their own private moderation systems on top, from which
individuals could choose. The government might also create an open system in which
third parties could develop applications that allow people to design their own personalized
feeds.
A government-provided search engine that is as efficient and effective as Google’s is a
somewhat harder lift and the cost of public provisioning for social media and search engines
might be prohibitive. But public provisioning poses a far larger problem: state surveillance.
Instead of Facebook and Google scooping up your personal data, the government would.
The Fourth Amendment might not prohibit this under existing doctrines, because people
willingly give the information to the public entity. Therefore any public provisioning system
would have to be accompanied by very strict self-imposed restrictions on collection, analysis,
and use. I am deeply skeptical that law enforcement and national security officials would
willingly forgo access to all of this information.

Professional and Public-regarding Norms
We should not treat social media companies and search engines as state actors subject to
the First Amendment. Yet we can still criticize them for arbitrariness and censorship. How
is that possible if, as I have just explained, these companies must engage in content- and
viewpoint-based judgments to do their jobs?

Hoover Institution  •  Stanford University

 8

We can criticize social media companies in three ways, none of which requires us to treat
them as state actors.
First, we can criticize them for being opaque and non-transparent and for denying basic norms
of fair process. This happens when social media do not state their criteria for governance
clearly in advance and do not offer reasoned explanations for their decisions.
Second, we can criticize them for being arbitrary—for not living up to their own community
guidelines and terms of service. They should apply their own rules without fear or favor to the
rich and to the poor, the high and low alike. Twitter and Facebook, to name two examples,
have often been lax with violations of their terms of service by famous or well-known people
and strict with violations by people who are not famous or well known.13 This allows the
more powerful and famous to abuse the less powerful with impunity and it creates blind
spots in enforcement.
Third, and perhaps more important, we can criticize social media companies for failing to
live up to norms of professionalism and expertise—that is, for failing to live up to the norms
of the kind of entity they purport to be.
Here is an analogy. People criticize major newspapers and media outlets all the time. They
criticize them for biased coverage, they criticize them for giving a platform to people who
make stupid or evil arguments, and they criticize them for failing to educate the public
about the issues of the day.
In most cases, people understand that these criticisms aren’t supposed to lead to government
regulation of newspapers and mass media. People understand that these companies have a
First Amendment right to exercise editorial discretion as they see fit, even if they exercise
it badly. Nevertheless, they hold these companies to a higher standard than ordinary
individuals expressing their opinions. The public rightly assumes that media companies
should live up to certain professional standards that are both public-regarding and connected
to democratic life. These include, among other things, providing the public with important
information necessary to self-government, striving to cover the news accurately and fairly,
engaging in professional fact-checking, adhering to professional standards of journalistic
ethics, and so on.
Many media organizations fail to live up to these standards, often spectacularly so. And some
media organizations have essentially given up on professionalism, fairness, and accuracy.
But people generally understand that this is a valid reason to criticize them, not to exculpate
them. Media companies hold themselves out as adhering to professional and public-regarding
norms. Therefore people in a democracy feel that they have a right to criticize them when,
in their estimation, media fail to live up to those norms. Perhaps equally important, because
the norms are public-regarding, citizens in a democracy feel that they have a right to debate

Jack M. Balkin  •  Fixing Social Media’s Grand Bargain 

 9

what those professional norms should be, whether or not the media companies assume
them or live up to them.
Social media companies and search engine companies are not newspapers. Even so, they are
more than just run-of-the-mill companies. They do more than just serve ads or sell widgets.
They perform a public service—three connected services, in fact. First, they facilitate public
participation in art, politics, and culture. Second, they organize public conversation so that
people can easily find and communicate with each other. Third, they curate public opinion
through individualized results and feeds and through enforcing terms-of-service obligations
and community guidelines.
These digital companies are the twenty-first-century successors of twentieth-century mass
media companies, even though their functions are somewhat different. The public, not
surprisingly, has come to view them as having a public-oriented mission.
In fact, these companies encourage this understanding through the ways they talk about
themselves. The Twitter Rules, for example, begin with the statement, “We believe that
everyone should have the power to create and share ideas and information instantly,
without barriers. In order to protect the experience and safety of people who use Twitter,
there are some limitations on the type of content and behavior that we allow.”14 This is
a statement of public-regarding, professional norms for facilitating public participation,
organizing public discussion, and curating public opinion. Facebook and YouTube have
made similar statements of purpose and justifications for their community guidelines,
although their policies differ in some respects.15
Whether they imagined it or not at the outset, these companies have taken on a public
function. People may therefore criticize them—and should criticize them—if they feel that
these companies are acting contrary to appropriate professional norms.
Moreover, because these companies have taken on these three tasks—facilitating public
participation, organizing public conversation, and curating public opinion—they may also
impose basic civility norms against abuse, threats, and harassment. They may also ban hate
speech or speech that denigrates people if they think that this kind of speech will undermine
the public-regarding purposes of the site. Social media companies may do this even if the First
Amendment would prevent the federal government from imposing the same civility norms
on a government-operated social media site.
But if social media companies decide to govern their sites through imposing civility norms
and regulating harassment and abuse, they should abide by the two other basic norms stated
above. First, they should be transparent about what they are doing and why they are doing
it. Second, they should not be arbitrary in their governance.

Hoover Institution  •  Stanford University

 10

Social media companies have been only fitfully successful at meeting these obligations.
Understood charitably, we might say that they are at the very beginning of a long process of
learning how to be responsible professionals. They have been wildly successful as technology
companies, but professionalism is more than technological expertise. Professional judgments
may require the application of norms that do not scale well. Sometimes applying these
norms will require judgment and individualized assessment as well as algorithmic sorting
and bright-line rules. Doing this costs more in human resources and attention than purely
technological solutions. To the extent that this is the case, social media companies should
absorb the extra costs of being professionals and living up to professional norms. Although
their efforts have been halting and often inadequate, social media companies are slowly
beginning that arduous process. In the meantime, civil society can play an important role
by continuing to criticize social media companies and by encouraging them to live up to
their public responsibilities.

Reforming Social Media
I have already said that we should not use the law to force these companies to behave as
public-regarding professionals any more than we can force major newspapers to adhere to
proper journalistic standards. Does this mean that law has no role to play? No. The law may
encourage these public-regarding norms in certain limited ways consistent with the First
Amendment.
Instead of directly aiming at the editorial policies of social media companies, reform
proposals should focus instead on the grand bargain that has turned the infrastructure
of digital free expression into the infrastructure of digital surveillance and control. Social
media companies will continue to cause a host of social problems as long as their business
models cause them not to care about these problems.
There are two central ways to change their behavior. The first is to reshape the organization
of social media companies. This is the task of antitrust and pro-competition law, which
have grown moribund in the Second Gilded Age and need a serious rethinking.
Social media companies’ perverse incentives derive from their business models—selling end
users’ information to advertisers and manipulating and addicting end users so that they
spend more time on social media and are thus more accessible to advertisers. Because a small
number of social media dominate end users’ attention, they also have a stranglehold over
digital advertising. People who wish to advertise online must operate primarily through
Facebook’s and Google’s advertising networks. This reduces revenues for many news and
media sites that are crucial to the health and vibrancy of the digital public sphere.
Increased enforcement of existing antitrust laws and a series of new pro-competition policies
might have two salutary effects. First, these reforms might restructure how digital advertising

Jack M. Balkin  •  Fixing Social Media’s Grand Bargain 

 11

operates, ameliorating the current bottleneck and freeing up revenues for a wider range of
media companies. Second, reform of competition policy and stricter antitrust enforcement
might break up the largest companies into smaller companies that can compete with each
other or create a space for new competitors to emerge. (Facebook and Google have often
bought up potential competitors before they could grow large enough to threaten them.)
More social media companies mean more platforms for innovation and more different
software features and affordances. More companies might also make it more difficult for
foreign hackers to disrupt the digital public sphere. All other things being equal, it may be
harder to hack twelve Facebooks than only one.16 Finally, more different kinds of companies
might also provide more models for social spaces and communities and a wider variety of
speech policies.
This last point is especially important. I have just argued that social media companies must
be allowed to enforce civility norms and regulate or even ban a wide range of speech that state
actors may not touch. But modern democracies increasingly rely on social media to perform
the public functions of organizing public opinion and facilitating public discussion. Therefore
it is very important to ensure that there are many social media applications and businesses
in order to prevent a small number of powerful for-profit companies from dominating how
public opinion is organized and governed.
Moreover, social media companies often enforce their terms of service imperfectly and
arbitrarily and they may make many questionable judgments. Some, like Facebook, attempt
to impose the same standards around the world.17 Finally, civil society organizations, mass
media, politicians, and governments have and will put increasing pressure on social media to
ban speech that they do not like and expel speakers who offend them. All of them, in various
ways, will try to coax social media into serving their political or ideological agendas. These
are all reasons for using pro-competition laws to ensure a healthy number of competing firms
organizing public discourse. Precisely because people will demand that huge multinational
corporations ban speech they do not like, it is important to have many Facebooks, not just one.
If we expect social media sites to enforce civility norms, we also need multiple social media
sites serving different values and different publics.

Information Fiduciaries
A second approach to reform is to make social media companies internalize the costs they
impose on society through surveillance, addiction, and manipulation by giving them new
social responsibilities. The short-term goal is to counteract the most egregious examples of
bad behavior. The long-term goal is to create legal incentives for social media companies
to develop professional cultures and public-oriented norms for organizing and curating
public discussion. To do this, I propose reaching back to some very old ideas in the law that
governs the professions: namely, the idea of fiduciary obligation.

Hoover Institution  •  Stanford University

 12

We should treat social media companies—and many other digital media companies as well—
as information fiduciaries toward their clients and end users.18 As information fiduciaries, digital
companies should have duties of care, confidentiality, and loyalty toward the people whose
data they collect, store, and use. This reform is a natural outgrowth of the grand bargain that
has enabled free expression in the digital age.
Because of digital companies’ increasing capacities for surveillance and control, they must
take on new legal responsibilities. Put simply, digital companies know a lot about us, and
they can use that knowledge in many ways—but we don’t know a lot about them. Moreover,
people increasingly depend on a wide range of digital services that observe them and collect
data about them. That makes people increasingly vulnerable to these companies. Because
the companies’ operations are not transparent, people have to trust that these services will
not betray them or manipulate them for their own ends. Digital companies that create and
maintain this dependence and vulnerability should be considered information fiduciaries
toward their end users.
There is plenty of precedent for this idea. For centuries, the law has recognized that certain
people hold power over others who are vulnerable to them, dependent on them, and have to
trust them. It created the idea of fiduciary obligations for just these situations.19 For example,
the law has long maintained that the clients or patients of doctors and lawyers are in special
relationships of dependence and vulnerability. We need to trust these professionals with
sensitive personal information about ourselves, but the people we trust could use this same
information to harm us and enrich themselves in many different ways. Therefore the law
treats professionals like doctors, lawyers, accountants, and estate managers as fiduciaries.
Fiduciary relationships require good faith and loyalty toward people whom the relationships
place in special positions of vulnerability. Accordingly, fiduciaries have special duties of care,
confidentiality, and loyalty toward their clients and beneficiaries.
Because social media companies collect so much data about their end users, use that data
to predict and control what end users will do, and match them with third parties who
may take advantage of end users, they are among the most important examples of the new
information fiduciaries of the digital age. We should apply these traditional obligations to
the changed conditions of a new technological era.
Facebook is not your doctor or lawyer. YouTube is not your accountant or estate manager.
We should be careful to tailor the fiduciary obligations to the nature of the business and
to the reasonable expectations of consumers. That means that social media companies’
fiduciary duties will be more limited.
Social media companies and search engines provide free services in exchange for the right
to collect and analyze personal data and serve targeted ads. This by itself does not violate
fiduciary obligations. Nevertheless, it creates a perpetual conflict of interest between end

Jack M. Balkin  •  Fixing Social Media’s Grand Bargain 

 13

users and social media companies. Companies will always be tempted to use the data they
collect in ways that increase their profits to their end users’ disadvantage. Unless we are to
ban targeted advertising altogether (which I would oppose and which raises serious First
Amendment problems) the goal should be to ameliorate or forestall conflicts of interest and
impose duties of good faith and non-manipulation. That means that the law should limit
how social media companies can make money off their end users, just as the law limits how
other fiduciaries can make money off their clients and beneficiaries.
As information fiduciaries, social media companies have three major duties: duties of care,
duties of confidentiality, and duties of loyalty. The duties of care and confidentiality require
fiduciaries to secure customer data and not disclose it to anyone who does not agree to
assume similar fiduciary obligations. In other words, fiduciary obligations must run with the
data. The duty of loyalty means that fiduciaries must not seek to advantage themselves at
their end users’ expense and they must work to avoid creating conflicts of interest that will
tempt them to do so. At base, the obligations of loyalty mean that digital fiduciaries may
not act like con artists. They may not induce trust on the part of their user base and then
turn around and betray that trust in order to benefit themselves.
To see what these obligations would mean in practice, we can use the Cambridge Analytica
scandal that propelled the issue of social media regulation to public attention in the spring
of 2018.
Although the facts are complicated, they essentially involved Facebook’s decision to allow
third parties to access its end users’ data.20 Facebook allowed researchers to do this for free
and took a cut of the profits for business entities. This allowed it to leverage its central
resource—consumer data—to increase profits.
Aleksandr Kogan, a data scientist, used a personality quiz to gain access to Facebook’s end-user
data. He thereby obtained not only the data of the 300,000 people who logged in using their
Facebook credentials, but also all of their Facebook friends, an estimated 87 million people.21
In fact, Kogan was actually working for Cambridge Analytica, a for-profit political consulting
company. Cambridge Analytica used the end-user data to produce psychological profiles that,
in turn, it would use to target political advertisements to unsuspecting Facebook users. In
fact, these practices were only the tip of a far larger iceberg. Facebook made a series of unwise
decisions to allow a range of business partners access to its end users’ social graphs and thus
make them vulnerable to various kinds of manipulation.22
As an information fiduciary, Facebook violated all three of its duties of care, confidentiality,
and loyalty. It did not take sufficient care to vet its academic and business partners. It did
not ensure that it only gave access to data to entities that would promise to maintain the
same duties of care, confidentiality, and loyalty as Facebook. It did not take sufficient steps
to audit and oversee the operations of these third parties to ensure that they did not violate

Hoover Institution  •  Stanford University

 14

the interests of its end users. It allowed third parties to manipulate its end users for profit.
And when it discovered what had happened, many years later, it did not take sufficient steps
to claw back its end users’ data and protect them from further breaches of confidentiality
and misuse.
Fiduciary obligations matter most in situations in which social media companies have
powerful market incentives not to protect their end users: for example, when social media
companies give access to data to third-party companies without adequate safeguards to prevent
these third parties from manipulating end users. Fiduciary obligations also matter when social
media companies perform social science experiments on their end-user base. Social media
companies are not part of universities and therefore are not bound by human-subjects research
obligations. As information fiduciaries, however, they would have legal duties not to create an
unreasonable risk of harm to their end users or to the public for their own advantage. They
would have duties, just as university scientists do, to minimize harm and prevent overreaching
and manipulation by their employees and contractors.
Finally, if social media companies are information fiduciaries, they should also have a
duty not to use end-user data to addict end users and psychologically manipulate them.
Social media companies engage in manipulation when end users must provide information
in order to use the service and when companies use this information to induce end-user
decision making that benefits the company at the expense of the end user and causes harm
to the end user. Because this creates a conflict of interest between the company and its end
users, it violates the duty of loyalty.
It may be useful to compare the fiduciary approach with the privacy obligations of the
European Union’s General Data Protection Regulation (GDPR). There is considerable overlap
between the two approaches. But the most important difference is that the GDPR relies
heavily on securing privacy by obtaining end-user consent to individual transactions. In
many respects, it is still based on a contractual model of privacy protection. Contractual
models will prove insufficient if end users are unable to assess the cumulative risk of granting
permission and therefore must depend on the good will of data processors. The fiduciary
approach to obligation does not turn on consent to particular transactions, nor is it bounded
by the precise terms of a company’s written privacy policy or terms of service, which are easy
for companies to modify. Rather, the fiduciary approach holds digital fiduciaries to obligations
of good faith and non-manipulation regardless of what their privacy policies say.
The fiduciary approach is also consistent with the First Amendment. That is because it aims
at regulating the relationships of vulnerability and trust between information fiduciaries
and those who must trust them.23
The First Amendment treats information gained in the course of a fiduciary relationship
differently from other kinds of information. Tell a secret to a person in the street and he

Jack M. Balkin  •  Fixing Social Media’s Grand Bargain 

 15

or she can publish it tomorrow and even use it against your interests. But when you reveal
information to a fiduciary—a doctor, nurse, or lawyer—he or she has to keep it confidential
and cannot use it against you. Information gained in the course of a fiduciary relationship—
and that includes the information that social media companies collect about us—is not
part of the public discourse that receives standard First Amendment protection. Instead, the
First Amendment allows governments to regulate fiduciaries’ collection, collation, use, and
distribution of personal information in order to prevent overreaching and to preserve trust and
confidentiality.24 The same principle should apply to the new information fiduciaries of the
digital age.
There may be close cases in which we cannot be sure whether a company really is acting
as an information fiduciary. To deal with these situations, Jonathan Zittrain and I have
proposed that Congress offer digital companies a different grand bargain to protect end
users’ privacy.25 It would create a safe harbor provision for companies that agree to assume
fiduciary obligations. The federal government would preempt state regulation if digital
media companies accept the obligations of information fiduciaries toward their end users.
Offering this exchange does not violate the First Amendment.
For the most part, the fiduciary approach leaves social media companies free to decide how
they want to curate and organize public discussion, focusing instead on protecting privacy
and preventing incentives for betrayal and manipulation. It affects companies’ curation and
organization of public discourse only to the extent that companies violate their duties of
care, confidentiality, and loyalty.
The fiduciary approach has many advantages. It is not tied to any particular technology. It
can adapt to technological change. It can be implemented at the state or the federal level,
and by judges, legislatures, or administrative agencies.
The fiduciary approach also meshes well with other forms of consumer protection, and it
does not exclude other reforms, like GDPR-style privacy regulation. In particular, it does not
get in the way of new pro-competition rules or increased antitrust enforcement as described
above. That is because it does not turn on the size of an organization (although Congress
might choose to regulate only the larger sites in order to encourage innovation and avoid
barriers to entry). It also does not turn on the presence or absence of monopoly power. It
applies whether we have twelve Facebooks or only one. Indeed, even if we had a wide range
of social media companies, all harvesting, analyzing, and using end-user data, there would
still be a need for fiduciary obligations to prevent overreaching.
The fiduciary approach pays attention to deeper causes. It directs its attention to the
political economy of digital media. It focuses on repairing the grand bargain that pays for
the digital public sphere in the Second Gilded Age.

Hoover Institution  •  Stanford University

 16

Notes
1 ​During the First Gilded Age, which ran from the end of Reconstruction to the beginning of the twentieth
century, technological innovation created huge fortunes in the hands of a small number of entrepreneurs and
produced increasing inequalities of wealth and deep political corruption. Waves of immigration and increasing
racial tensions led to the emergence of populist demagogues. American government was increasingly for sale,
and many people despaired for the future of American democracy. The corruption and inequality of the First
Gilded Age led to the reforms of the Progressive Era and, eventually, the New Deal. For a general history of the
period, see H. W. Brands, American Colossus: The Triumph of Capitalism, 1865–1900 (New York: Anchor, 2010).
2 ​See Zeynep Tufecki, “Facebook’s Surveillance Machine,” New York Times, March 19, 2018, accessed
September 27, 2018, https://­w ww​.­nytimes​.­com​/­2018​/­03​/1­ 9​/­opinion​/f­ acebook​- ­cambridge​-­analytica​.­html.
(“Facebook makes money, in other words, by profiling us and then selling our attention to advertisers, political
actors and others. These are Facebook’s true customers, whom it works hard to please.”) These business
models and the incentives they create are examples of what Shoshana Zuboff calls “surveillance capitalism.”
Shoshana Zuboff, “Big Other: Surveillance Capitalism and the Prospects of an Information Civilization,” Journal of
Information Technology 30 (April 2015): 75 (defining “surveillance capitalism” as a “new logic of accumulation”
and a “new form of information capitalism [that] aims to predict and modify human behavior as a means to
produce revenue and market control”).
3 ​See, e.g., Paul Lewis, “ ‘Fiction Is Outperforming Reality’: How YouTube’s Algorithm Distorts Truth,” Guardian,
February 2, 2018, accessed September 27, 2018, https://­w ww​.­theguardian​.­com​/t­ echnology​/­2018​/­feb​/­02​/­how​
-­youtubes​-­algorithm​- ­distorts​-­truth (explaining how YouTube’s algorithm to engage viewers promotes conspiracy
theories).
4 ​This definition of manipulation focuses on leveraging another’s lack of knowledge and emotional vulnerability
to benefit oneself at the expense of another’s welfare. This is not the only way to define the concept. See,
e.g., Cass R. Sunstein, The Ethics of Influence: Government in the Age of Behavioral Science (New York: Cambridge
University Press, 2016), 82 (a technique of influence is “manipulative to the extent that it does not sufficiently
engage or appeal to [a person’s] capacity for reflection and deliberation”). That definition, however, raises the
problem of how to distinguish manipulation from a wide range of ordinary techniques of marketing. A third
approach would focus on real or objective interests: manipulation is persuasion that leverages lack of knowledge
and emotional vulnerability to cause people to act against their real interests, however those are defined.
This approach raises the question of how we know what people’s real or objective interests are.
5 ​“Facebook Admits Failings over Emotion Manipulation Study,” BBC News, October 3, 2014, accessed
September 27, 2018, https://­w ww​.­bbc​.c­ om​/n
­ ews​/­technology​-­29475019 (“the company was widely criticised
for manipulating material from people’s personal lives in order to play with user emotions or make them sad”).
6 ​Jonathan Zittrain, “Engineering an Election,” Harvard Law Review Forum 127 (June 20, 2014): 335–36
(noting that experiment caused an additional 340,000 votes to be cast).
7 ​Ibid., 336 (describing the “ripple effects” of experiments).
8 ​See Mike Allen, “Sean Parker Unloads on Facebook: ‘God Only Knows What It’s Doing to Our Children’s
­ nloads​
Brains,’ ” Axios, November 9, 2017, accessed September 27, 2018, https://­w ww​.­axios​.­com​/­sean​-­parker​-u
-­on​-­facebook​-­god​-­only​-­knows​-­what​-­its​-­doing​-­to​-­our​-­childrens​-­brains​-­1513306792​-­f855e7b4​-­4e99​-­4d60​-­8d51​
-­2775559c2671​.­html (quoting statement by former president of Facebook that social media applications are
designed to “exploit a vulnerability in human psychology” using psychological methods to “consume as much
of your time and conscious attention as possible” and keep users locked into the site); Paul Lewis, “ ‘Our Minds
Can Be Hijacked’: The Tech Insiders who Fear a Smartphone Dystopia,” Guardian, October 6, 2017, accessed
September 27, 2018, https://­w ww​.­theguardian​.c­ om​/­technology​/­2017​/­oct​/­05​/­smartphone​-­addiction​-­silicon​-­valley​
-­dystopia (interviewing former employees at Google and Facebook who report that technologies are designed to
addict users and monopolize their attention).

Jack M. Balkin  •  Fixing Social Media’s Grand Bargain 

 17

9 ​See, e.g., Ryan Calo, “The Boundaries of Privacy Harm,” Indiana Law Journal 86, no. 3 (October 4, 2011): 1131,
1149 (“Many consumers have little idea how much of their information they are giving up or how it will be used”);
A. Michael Froomkin, “The Death of Privacy?” Stanford Law Review 52 (2000): 1461, 1502 (“Consumers suffer
from privacy myopia: they will sell their data too often and too cheaply”); Daniel J. Solove, “Privacy and Power:
Computer Databases and Metaphors for Information Privacy,” Stanford Law Review 53 (2001): 1393, 1452 (“It is
difficult for the individual to adequately value specific pieces of personal information”).
10 ​See Tufekci, “Facebook’s Surveillance Machine,” explaining that Facebook collects “shadow profiles” on
nonusers: “even if you are not on Facebook, the company may well have compiled a profile of you, inferred from
data provided by your friends or from other data. This is an involuntary dossier from which you cannot opt out
in the United States.” Social media users may unwittingly imperil each other’s privacy. The Cambridge Analytica
scandal revealed that when Facebook users logged in to a third-party app using their Facebook credentials, they
shared the social graphs of all of their Facebook friends without the latter’s consent. See Alexandra Samuel, “The
Shady Data-Gathering Tactics Used by Cambridge Analytica Were an Open Secret to Online Marketers. I Know,
Because I Was One,” The Verge, March 25, 2018, accessed September 27, 2018, https://­w ww​.­theverge​.­com​/­2018​/­3​
/­25​/­17161726​/­facebook​- ­cambridge​-­analytica​- ­data​- ­online​-m
­ arketers. (“The tactic of collecting friend data, which
has been featured prominently in the Cambridge Analytica coverage, was a well-known way of turning a handful
of app users into a goldmine.”)
11 ​For examples of what social media sites regulate, see Facebook, Community Standards, https://­w ww​
.­facebook​.­com​/­communitystandards; and Twitter, The Twitter Rules, https://­help​.t­ witter​.­com​/­en​/r­ ules​-a
­ nd​
-­policies​/­t witter​-­rules (both accessed September 27, 2018).
12 ​Tarleton Gillespie, Custodians of the Internet: Platforms, Content Moderation, and the Hidden Decisions
That Shape Social Media (New Haven, CT: Yale University Press, 2018).
13 ​See Kate Klonick, “The New Governors: The People, Rules, and Processes Governing Online Speech,” Harvard
Law Review 131 (April 10, 2018): 1598, 1654–55 (2018) (noting that social media companies may disproportionately
favor people with power over other end users).
14 ​Twitter, the Twitter Rules.
15 ​Facebook, Community Standards, “We recognize how important it is for Facebook to be a place where people
feel empowered to communicate, and we take our role in keeping abuse off our service seriously. That’s why we
have developed a set of Community Standards that outline what is and is not allowed on Facebook. . . . ​The goal
of our Community Standards is to encourage expression and create a safe environment,” YouTube, Policies and
Safety, accessed September 27, 2018, https://­w ww​.­youtube​.­com​/­y t​/­about​/­policies​/­#community​-­guidelines,
“When you use YouTube, you join a community of people from all over the world. . . . ​Following the guidelines
below helps to keep YouTube fun and enjoyable for everyone.”
16 ​Sally Hubbard, “Fake News is a Real Antitrust Problem,” CPI Antitrust Chronicle, December 2017: 5, accessed
September 27, 2018, https://­w ww​.­competitionpolicyinternational​.­com​/­wp​-­content​/­uploads​/­2017​/­12​/­CPI​
-­Hubbard​.­pdf.
17 ​Klonick, “New Governors,” 1642, describing Facebook’s goal of applying its norms worldwide and the resulting
compromises; Julia Angwin and Hannes Grassegger, “Facebook’s Secret Censorship Rules Protect White Men
from Hate Speech but Not Black Children,” ProPublica, June 28, 2017, accessed September 27, 2018, https://­
www​.­propublica​.­org ​/­article​/­facebook​-­hate ​-­speech​-­censorship​-­internal​-­documents​-­algorithms (describing
Facebook’s attempts to enforce its hate speech rules worldwide and the arbitrariness of its categories).
18 ​See Jack M. Balkin, “Information Fiduciaries and the First Amendment,” UC Davis Law Review 49, no. 4
(April 2016): 1183; Jack M. Balkin, “The Three Laws of Robotics in the Age of Big Data,” Ohio State Law Journal 78
(2017): 1217.
19 ​See generally Tamar Frankel, Fiduciary Law (New York: Oxford University Press, 2011).

Hoover Institution  •  Stanford University

 18

20 ​See Carole Cadwalladr and Emma Graham-Harrison, “How Cambridge Analytica Turned Facebook ‘Likes’ into
a Lucrative Political Tool,” Guardian, March 17, 2018, accessed September 27, 2018, https://­w ww​.­theguardian​.­com​
/­technology​/­2018​/­mar​/­17​/­facebook​-­cambridge​-­analytica​-­kogan​-­data​-­algorithm; Carole Cadwalladr and Emma
Graham-Harrison, “Revealed: 50 Million Facebook Profiles Harvested for Cambridge Analytica in Major Data
Breach,” Guardian, March. 17, 2018, accessed September 27, 2018, https://­w ww​.t­ heguardian​.­com​/­news​/­2018​/m
­ ar​
/­17​/­cambridge​-­analytica​-­facebook​-­influence​-­us​- ­election; Paul Lewis, “  ‘Utterly Horrifying’: Ex-Facebook Insider
Says Covert Data Harvesting Was Routine,” Guardian, March 20, 2018, https://­w ww​.­theguardian​.­com​/­news​/­2018​
/­mar​/­20​/­facebook​-­data​-­cambridge​-­analytica​-­sandy​-­parakilas.
21 ​See Michael Riley, Sarah Frier, and Stephanie Baker, “Understanding the Facebook-Cambridge Analytica Story:
QuickTake,” Washington Post, April 9, 2018, accessed September 27, 2018, https://­w ww​.­washingtonpost​.­com​
/­business​/­understanding​-­the​-­facebook​-­cambridge​-­analytica​-­story​-­quicktake​/­2018​/­0 4​/­09​/­0f18d91c​-­3c1c​-­11e8​
-­955b​-­7d2e19b79966​_ ­story​.­html (estimating that 300,000 people participated and that 87 million users had their
data harvested).
22 ​Lewis, “Covert Data Harvesting Was Routine” (quoting a former Facebook employee who explained that under
the company’s policies, “a majority of Facebook users” could have had their data harvested by app developers
without their knowledge).
23 ​On the First Amendment issues, see Balkin, “Information Fiduciaries and the First Amendment,” 6.
24 ​Ibid.
25 ​Jack M. Balkin and Jonathan Zittrain, “A Grand Bargain to Make Tech Companies Trustworthy,” Atlantic,
October 3, 2016, accessed September 27, 2018, https://­w ww​.­theatlantic​.­com​/­technology​/­archive​/­2016​/­10​
/­information​-­f iduciary​/­502346.

Jack M. Balkin  •  Fixing Social Media’s Grand Bargain 

 19

The publisher has made this work available under a Creative Commons Attribution-NoDerivs license 3.0. To view a copy
of this license, visit http://creativecommons.org/licenses/by-nd/3.0.
Hoover Institution Press assumes no responsibility for the persistence or accuracy of URLs for external or third-party
Internet websites referred to in this publication, and does not guarantee that any content on such websites is, or will
remain, accurate or appropriate.
Copyright © (2018) by the Board of Trustees of the Leland Stanford Junior University
The preferred citation for this publication is: Jack Balkin, “Fixing Social Media’s Grand Bargin,” Hoover Working
Group on National Security, Technology, and Law, Aegis Series Paper No. 1814 (October 16, 2018), available at https://­
www​.­lawfareblog​.­com​/­advanced​-­persistent​-­manipulators​-­and​-­social​-­media​-­nationalism​-­national​-­security​-­world​
-­audiences.

Hoover Institution  •  Stanford University

 Working Group on National Security,
Technology, and Law

About the Author

Jack M. Balkin
Jack M. Balkin is Knight Professor
of Constitutional Law and the First
Amendment at Yale Law School
and the founder and director
of Yale’s Information Society
Project. He is a member of the
American Academy of Arts and
Sciences and founded and edits the
group blog Balkinization (http://­
balkin​.­blogspot​.­com​/ )­ . His most
recent book is Living Originalism
(Cambridge, MA: Belknap Press,
2011).

Hoover Institution, Stanford University
434 Galvez Mall
Stanford, CA 94305-6003
650-723-1754

The Working Group on National Security, Technology, and
Law brings together national and international specialists
with broad interdisciplinary expertise to analyze how
technology affects national security and national security
law and how governments can use that technology to
defend themselves, consistent with constitutional values
and the rule of law.
The group focuses on a broad range of interests, from
surveillance to counterterrorism to the dramatic impact
that rapid technological change—digitalization,
computerization, miniaturization, and automaticity—are
having on national security and national security law.
Topics include cybersecurity, the rise of drones and
autonomous weapons systems, and the need for—and
dangers of—state surveillance. The group’s output will also
be published on the Lawfare blog, which covers the merits
of the underlying legal and policy debates of actions
taken or contemplated to protect the nation and the
nation’s laws and legal institutions.
Jack Goldsmith and Benjamin Wittes are the cochairs of
the National Security, Technology, and Law Working Group.
For more information about this Hoover Institution Working
Group, visit us online at http://www.hoover.org/research-teams
/national-security-technology-law-working-group.

Hoover Institution in Washington
The Johnson Center
1399 New York Avenue NW, Suite 500
Washington, DC 20005
202-760-3200