Content uploaded by Azi Lev-On
Author content
All content in this area was uploaded by Azi Lev-On on Oct 21, 2014
Content may be subject to copyright.
8 The Democratizing Effects of Search Engine
Use: On Chance Exposures and Organizational
Hubs
Azi Lev-On
University of Pennsylvania
Abstract
In this paper I highlight two implications of the widespread use of search engines,
which are often overlooked by commentators. In the first part of the paper I argue
that search engines are conducive to unplanned exposures to diverse and even op-
posing views. In the second part I argue that search engines indirectly contribute
to emergent political organization, since they allow large numbers of people to lo-
cate and access organizational hubs of collective action. I conclude by pointing to
the democratic significance of these properties.
I
In late 2002 Jiang Mianheng, the son of the former Chinese president and a
powerful political figure, visited the 502 research institute of the Ministry
of Information Industry to see a demonstration of high-speed internet. One
of the engineers typed the name of his father, “Jiang Zemin,” in the Google
search engine box. Three of the top ten results were highly critical of the
senior Jiang. “Evil Jiang Zemin” was the title of the first result. Shortly af-
terwards, according to well-informed sources, Jiang Mianheng instructed
to block the search engine site (Tianliang 2005).
In a New York Times article from April 23, 2006 entitled Google's
China Problem (and China's Google Problem), Clive Thompson com-
2 Azi Lev-On2
ments that authoritarian governments and companies that provide internet
search services are strange bedfellows. As evident from the title of his arti-
cle, Thompson focuses on Google and the Chinese authorities. ‘China’s
Google problem’ refers to the authorities’ discontent with the new capa-
bilities of Chinese citizens to locate and gain access, through search en-
gines, to websites critical of certain governmental policies. ‘China’s
Google problem’ is nicely manifest by the Jiang story above. ‘Google’s
China problem’ is Google’s discontent with the authorities’ demand to
censor and monitor its citizens’ use of the search engine. Such demands
are at odds with the company’s policies and, for some, cannot be recon-
ciled with its motto of ‘don’t be evil.’
Google’s recent policy shift and decision to comply with the authorities’
demands and censor certain search results on its Chinese website led to a
public uproar, and to intense and largely critical press coverage. However,
in this paper I do not focus on ‘Google’s China problem,’ but on ‘China’s
Google problem’ instead. In light of the harsh reaction of authoritarian
governments against indexing and searching sites with arguably no inde-
pendent political agenda, I would reflect on the dilemmas that search en-
gines pose for authoritarian governments, and point to the democratic sig-
nificance of search engine use.
This short essay is not an elaborate case study of either ‘Google’s China
problem’ or ‘China’s Google problem.’ I utilize ‘China’s Google problem’
to illustrate the tensions between authoritarianism and enhanced popular
information-seeking capabilities. The tensions between Google and the
Chinese authorities are especially interesting given Google’s current
dominance in the search market, and the aggressive efforts of the Chinese
authorities to lock local surfers behind a ‘great firewall.’ But the points
made in this paper equally hold for other authoritarian governments and
searching and indexing services.
II
Undoubtedly, search engines have become a vital tool for information-
seeking. Search-engine websites consistently top the lists of popular web-
sites; a recent survey shows that on a given day 56% of surfers use them
(Fallows 2005). In addition to search-engine sites, search boxes are em-
bedded in countless websites, and gradually in personal computers as well.
A common metaphor for the internet is of a huge library, containing
vast amounts of materials from great many sources. But a huge library
with no efficient indexing and searching tools is essentially useless and
8 The Democratizing Effects of Search Engine Use: On Chance Exposures and
Organizational Hubs 3
3
probably counter-productive as well. Search engines effectively create an
index and assist in ‘making order out of chaos’ and in evading information
overload online. Battelle (2005) convincingly argues that we should con-
ceptualize search engines as information intermediaries or brokers, that as-
sist in matching information supply and demand by creating a ‘market-
place’ where information-providers can ‘publicize’ their merchandise and
be located by information-seekers, and information-seekers can obtain lists
of results that are potentially relevant to their queries.1
Famously, the algorithm behind Google, PageRank, emphasizes in-
bound links when it determines the relevance of possible responses to us-
ers’ queries. More precisely, the algorithm holds links from popular sites
‘in greater esteem’ than links from unpopular sites when determining rele-
vance. The idea is that the linking patterns of popular sites provide a good
proxy for users’ needs. In other words, if according to many sites (and par-
ticularly popular sites) a particular site contains information that is relevant
to your query, you are likely to find this site relevant as well.
Search engines which utilize the linking patterns of many other users to
determine relevance have, evidently, a number of advantages over human-
generated indexes (where users categorize and comment on individual
sites), and expert-run answering services (where users provide direct re-
sponses to other users’ queries), in terms of such parameters as the efforts
required from the information-broker, response times, and the number of
sources upon which the answer is based. When PageRank and its cousins
produce ‘organic’ results, which are driven by the linking decisions of in-
dividuals and not tinkered with or compromised by spammers, firms or
governments, they create a rather genuine ‘public choice.’ PageRank and
similar algorithms popularize the search function, basing it on a slightly
‘filtered’ public opinion.
Google and other search engines have been recently criticized for a va-
riety of reasons. Some argue that at times there is no sufficient separation
between the presentation of organic results and paid results, and conse-
quently users may fail to clearly distinguish between the two. Other criti-
cisms refer to biases that result from governmental intervention, as in the
Chinese case. Censoring some organic search results and replacing popular
items with government-approved less popular items obviously bias the
search outcomes. Moreover, when search engines completely remove ‘for-
bidden’ items from the result list, without even leaving a non-functional
1 Battelle (2005, 47) suggests that Google “would like to provide a platform that
mediates supply and demand for pretty much the entire world economy.”
4 Azi Lev-On4
link to the blocked result, users are unaware that such ‘forbidden’ results
even exist.
The above critiques refer to manipulations of the presentation of search
results. Other critiques regard the by-products of the inherent features of
search-engine algorithms. In this regard, it has been argued that search en-
gines assist in transforming the equality of opportunity the internet is so
much praised for, into inequality of outcome, and substantiating the domi-
nance of a small elite of highly-linked sites over users’ attention.2 Research
shows that the Web link structure is highly skewed, where a small number
of sites is heavily linked to, and the overwhelming majority of sites are
quite inaccessible. These skewed linking patterns hold not just for the Web
as a whole, but also for thematic sites that deal with political issues as gun
control, abortion, capital punishment, and general politics directories
(Hindman, Tsioutsiouliklis and Johnson 2003). These phenomena have
been explained as consequences of a ‘rich get richer’ dynamics, which
mainly occurs due to preferential attachment of new outbound links to al-
ready salient websites (Barabási 2002; Huberman 2001).3 It has been ar-
gued that search engines, and especially Google-style popularity-based
search engines, channel surfers primarily to already popular sites, and help
substantiate their centrality. But search engines not only direct people to
already popular sites; assuming that the probability that users link to a par-
ticular site increases if they are routed to this site, search engines indirectly
perpetuate and reinforce the highly inegalitarian distribution of links and
traffic online.
Note that such critiques are reminiscent of long-standing critiques of
direct democracy (that some search engines emulate) regarding its vulner-
ability to administrative and commercial pressures, and its tendency to lead
to majority tyranny. All, or some, of these concerns may be justified to
some degree or another. But they are not our main concern here. Instead,
2 But see Fortunato et al. (2005), who argue that search engines are “directing
more traffic toward less popular sites, even in comparison to what would be ex-
pected from users randomly surfing the Web.”
3 Research shows that skewed distributions, such as power-law distributions, are
ubiquitous online. In addition to the Web link structure and traffic that is corre-
lated with it (Barabási 2002), Drezner and Farrell (2003) found that the distribu-
tion of inbound links to blogs follows a power-law distribution as well. The
highly inegalitarian distributions of links and traffic have profound implications
for web-based organization. The fact that a small number of sites emerge as fo-
cal sites means that users with similar tastes, economic interests or hobbies can
easily converge onto a narrow set of focal sites. Such focal points serve, in es-
sence, as organizational hubs that can be easily discerned by search engines (see
later).
8 The Democratizing Effects of Search Engine Use: On Chance Exposures and
Organizational Hubs 5
5
let us focus on certain advances that search engines generate in democra-
cies, and the flip-side: the concerns they raise in authoritarian regimes.
My aim here is not to deny that the uses of search-engines generate some
by-products that may be at odds with our democratic sensibilities. Such
potential problems coexist with the new promises that are surveyed below.
III
An interesting feature of search engines, which is nicely demonstrated by
the Jiang incident, is that they occasionally generate unplanned and unpre-
dictable exposures to diverse views, even to information that runs counter
to searchers’ prior beliefs. For example, users who want to learn about cel-
lular phones can be directed to websites which focus on their disadvan-
tages and even hazards (Brin and Page 1998), but at other times can be
routed to websites which praise them. Users who champion capitalism or
globalization and want to learn more about these topics can be channeled
to anti-capitalist or anti-globalization sites, respectively.
Keep in mind that offline, the chances of running into opposing views,
especially in political matters, are not promising. Research shows that
people tend to carefully select their conversation partners, and political talk
occurs mostly among friends, family and like-minded others (see Huck-
feldt and Sprague 1995; Kim, Wyatt and Katz 1999; Conover, Searing and
Crewe 2002). Even the voluntary associations that people choose to join,
evolve to become rather homogenous ideologically (Theiss-Morse and
Hibbing 2005, see also Mutz 2006).
Search engines, on the other hand, enable easy access, with a click of a
mouse, to vast amounts of information generated by many sources. But
easy access cannot by itself counter the filtering mechanisms of everyday
discourse. Let us imagine an information environment in which extensive
amounts of information exist alongside refined tailoring abilities of con-
tent, i.e. people can use search engines to carefully select those items that
correspond to their worldview from the massive amounts of information,
and screen out all the rest. In such environments, refined search and tailor-
ing abilities may generate exposures only to information confirming and
reassuring users’ prior views, consistent with users’ homophile informa-
tion-seeking patterns offline (Mutz 2006).
But I argue that, at present, search engines do not allow for such re-
fined filtering capabilities, and at times even unintentionally expose users
to opposing views. While unfolding the reasoning, let us keep in mind that
three ‘components’ are involved in the process of retrieving information
6 Azi Lev-On6
through search engines: the user, the search engine, and the information
available online. The figure below shows a highly simplified version of
mediated processes of searching and retrieving information from a data-
base, where users (‘demand’) retrieve information from databases, using
an ‘intermediary.’
Fig. 8.1. An illustration of Mediated Database Information Retrieval Processes
The ‘intermediary’ phase of this process is depicted in the drawing as a
human and can be, for example, a family-member, a friend or an expert.
But it can also be non-human; for example, the intermediary can be Pag-
eRank or another algorithm that fetches information from the database at
the request of the ‘demanding’ person.
In an ‘ideal retrieval process,’ queries are perfectly framed and articu-
lated by the users (‘demand’). The intermediary does not only have access
to the wording of the query, but has a ‘deeper’ understanding of users’ in-
tentions which enables it to ask for clarifications or suggest modifications
to the query before accessing the database. The database itself is perfectly
indexed, such that the intermediary can have a direct access to all the rele-
vant information (for another account of a ‘perfect search’ see Battelle
2005, ch. 7). Think of an intelligent agent that can, upon command in natu-
ral language, “fetch all arguments for limiting immigration”, or “provide a
summary of the recent successes of pro-life efforts”, or “suggest an argu-
ment why gay marriages are morally right” or wrong. Such an ‘ideal
search’ allows users, if so they wish, to craft their own ideological uni-
verse out of the vast amounts of information available online, and effec-
tively filter out all traces of diverse and opposing views.
But there are a number of obstacles for such an ‘ideal search,’ when it
is carried out through search engines. Below I focus on three such obsta-
cles involving imperfect database indexing, limited intermediary qualifica-
tions to recognize the intentions of searchers and fine-tune the query, and
ill-formulations of queries. I claim that such obstacles prevent users from
perfectly tailoring their ‘ideological universe,’ and given the massive
8 The Democratizing Effects of Search Engine Use: On Chance Exposures and
Organizational Hubs 7
7
amounts and diversity of information online, they can even facilitate expo-
sure to diverse and opposing views. Let us review these obstacles in some
detail.
Let us start with the ‘supply side’ of the retrieval process, and comment
on the current absence of a comprehensive and reliable universal tagging
system (i.e. a semantic web) – the lack of a network of keywords that
properly describe the content of online documents. The current absence re-
sults from the lack of a central authority or a ‘central librarian’ to classify
online documents, a feature which is inherent to the internet. Note that re-
cently there have been some suggestions for collaborative classification of
documents, where users generate keywords that are associated with indi-
vidual sites. Tagging content collaboratively is an instance of what I else-
where call second-order collaborations (Lev-On and Hardin 2007), and is
increasingly used in a variety of websites. 4
The ill-classification of the online ‘database’ makes it difficult for the
‘intermediary’ (whether a search engine or otherwise) to locate relevant
content. Moreover, it makes it difficult to discriminate content based on
ideological affiliations in order to design and maintain, for example, a
‘progressive universe’ or a ’conservative web’ that can be queried through
search engines. In other words, it makes it difficult to perfectly tailor the
ideological affiliation of sites towards which users are channeled.
Where the first obstacle for an ‘ideal search’ is associated with the
‘supply’ side of the process, the second obstacle involves the interaction
between the intermediary and the ‘demand’ side – the searcher -- and re-
gards the comparatively limited abilities of the intermediary to have a
‘deep understanding’ of the intentions of searchers.
Let us think of queries along the lines of ‘fetch all the arguments and
court rulings against stem-cell research.’ one can direct such queries to an
‘online answering service’ composed of experts; alternatively, one can
post a query to newsgroups or virtual communities with known ideological
affiliations. Compared to such alternatives, the results obtained from
search engines can be pale. The alternatives have clear advantages over
search engines in terms of the usage of natural language, the ability to in-
duce intentions from the context and wording of the query, and the uses of
4 By ‘secondary collaborations’ we refer to a ‘family’ of institutions that aggre-
gate large amounts of individual selections and generate social choices. ‘Secon-
dary collaborations’ can be used to produce reputations, edit and rate content,
moderate discussions and provide reviews and recommendations of products
and services. Note that collaborative tagging may face such problems as im-
proper (and even malicious) tagging, and inter-personal disagreement on tag-
ging.
8 Azi Lev-On8
interactivity. These features allow such ‘intermediaries’ a fine-grained un-
derstanding of the intentions behind a formal query.
In the case of search engines, however, the interface is essentially tex-
tual and there are minimal interaction and feedback between the ‘demand’
side and the ‘intermediary.’ As a result, there are fewer opportunities for a
fine-grained understanding of the intention behind a formal query when
using search engines. At the current state of search engine technology,
then, mapping users’ intentions to relevant answers, especially for more
complicated queries, can be highly imperfect (see Battelle 2005).
The third and last obstacle for an ‘ideal search’ process regards the
searchers - the ‘supply side’ – and how they formulate and articulate their
intentions. A number of studies on information-seeking behaviors online
reveal that users compose very short queries, hardly use advanced search-
ing options, view a very small number of documents per query, and almost
never view more than one page of results (see Spink and Jansen 2004, Ma-
chill et al. 2004). Spelling mistakes and non-grammatical formulations are
frequent (Hargittai 2006).5
Such information-seeking patterns limit searchers’ abilities to retrieve
only information tailored to their views and filter out information that op-
poses them, and reduce the effectiveness of searching strategies. Note that
the first two obstacles for an ‘ideal search’ -- regarding content tagging
and intention guessing -- can be better addressed when search technolo-
gies improve and are better able, for example, to approximate natural lan-
guage or to capitalize on a comprehensive semantic Web. But improper
use of searching tools and inadequate framing of search queries will con-
tinue to limit users’ abilities to retrieve information, even after technologi-
cal capabilities improve.
In summary, I argue that due to such factors as the absence of a com-
prehensive and reliable system of keywords, the difficulties of deciphering
searchers’ intention by intermediaries, and far-from-optimal popular
search patterns, it is difficult to craft an ideal search, and searchers cannot
easily limit themselves to sealed ideological spaces online. If users had the
abilities to limit their horizons in such ways, they would indeed be able to
efficiently craft their own ideological echo chambers and totally prevent
exposure to opposing or diverse views, substantiating Sunstein’s (2001)
fears. But since agents can find it very difficult to limit their horizons in
such ways, and given the large amount of information and the variety of
sources online, when agents use search engines they can be directed to un-
5 The somewhat paradoxical argument here is, in effect, that illiteracy has its vir-
tues… at least in the narrow domain of generating chance exposures to diverse
and opposing views while using search engines.
8 The Democratizing Effects of Search Engine Use: On Chance Exposures and
Organizational Hubs 9
9
expected places, even to (popular) sites presenting arguments that counter
their views.6
Earlier we commented on the role of search engines in bridging the
supply and demand of information online (Battelle 2005). While search
engine do make information markets more efficient, they still imperfectly
bridge demand and supply. The combination of imperfect matching and
tailoring abilities, with abundance and diversity of information, seems
conducive to drive people to diverse and even opposing views (see Lev-On
and Manin 2007).
IV
In the previous section I argued that search engines are conducive to
chance exposures to diverse and even opposing views. In the following
section I argue that search engines also assist in generating and maintain-
ing organizational hubs that are instrumental for collective action.
Let us think of collective actions such as citizen-based campaigns to re-
evaluate and reconsider public policies (i.e. Leach 2005 on such a web-
supported campaign which aimed at revising immunization policies), or
orchestrated demonstrations and rallies, or community efforts to revise lo-
cal development plans. Typically such collective efforts are of interest to
large numbers of people, but at the absence of organizational infrastructure
such causes may not attract and mobilize enough support, and may become
latent (i.e. Olson 1965).
Orchestrating such collective efforts entails costs to both organizers
and activists. Organizers need to make decisions about mobilization of re-
sources, alliance formations, protest scheduling, location and coordination,
and so on. Activists and sympathizers need to locate particular events, re-
ceive relevant information and forward it to relevant others, and decide
where they can contribute effectively. Particularly, successful collective
6 Elsewhere we argue that occasional unplanned exposures should be seen as
‘happy accidents’ -- that some randomness are instrumental for adequate delib-
eration (Lev-On and Manin 2007, Sunstein 2001). Nevertheless, we do not wish
to argue that search engine should produce only random outcomes. Such search
engines would attract very little traffic, and will be conducive to chance expo-
sures of very few surfers. A necessary condition for mass exposure to opposing
views through search engines is, of course, that many people actually use the
search engines. And they use them, obviously, because they think that they are
likely to obtain valuable information through the search engines. This is, obvi-
ously, not the case with ‘random’ search engines.
10 Azi Lev-On10
efforts require the existence of easily accessed focal points to which organ-
izers, activists and sympathizers can converge to post and retrieve informa-
tion in order to coordinate their efforts.
I argue that search engines contribute to such collective efforts by ex-
posing popular organizational hubs, and directing traffic to them. The new
abilities of many people to locate organizational hubs of collective action
are, arguably, especially important for ‘unprivileged’ or ‘disorganized’ in-
terests. It may be difficult – if not impossible- to find information about
and join such collective efforts, that oftentimes lack a clear and easily-
accessible organizational ‘address’, offline.
Let me note that search engines are, of course, not always successful in
exposing organizational hubs. When agents rely on search engines to ob-
tain information about collective actions, the search engines determine
what the Web consists of for those seeking to contribute. If a search does
not return a link for a certain site, say a grassroots effort to change public
policy, then the seeker might never know that such an effort exists. On the
other hand, after a site gains momentum and becomes popular, search en-
gines make the popular effort even more noticeable for large numbers of
surfers, and provide potential contributors with a powerful gateway to col-
lective action.
Organizational hubs can have two main functions. First, they can en-
able intra-site communication, either in the form of documents and organ-
izational information (about timing of protests, for example), or in the
form of interactive conversations. Second, they also include links that,
when followed, can easily route people to other relevant sites.7 Search en-
gines function primarily in the second capacity, i.e. they direct agents to
other focal sites.
What does it mean that a site serves as an ‘organizational hub’ and di-
rects traffic to other sites? To illustrate this, think of the ‘Slashdot effect’.
Slashdot is one of largest virtual communities. The community is so suc-
cessful that it is famous for generating a ‘Slashdot effect:’ right after a link
to an interesting story published elsewhere becomes available, massive
numbers of users flood the original site. This sudden and heavy traffic
sometimes crashes the linked sites’ servers (the crash is the ‘Slashdot ef-
fect’). Search engines serve a similar function of revealing sites relevant
7 To clarify the distinction, think of a parallel distinction between topical blogs
that post information about a particular theme, and filter blogs that primarily post
links to sites that post information about such a theme.
8 The Democratizing Effects of Search Engine Use: On Chance Exposures and
Organizational Hubs 11
11
for collective action to large numbers of searchers who are interested in
such efforts.8
On the internet it is much easier to establish such organizational hubs
than offline, due to factors such as the reduction of gate keeping and setup
costs.9 Such organizational hubs can be set up by a variety of agents, such
as civil society organizations, interest groups, parties, social movements,
or just single individuals who take it upon themselves to initiate such col-
lective actions. But although almost anyone can establish a website which
aims at organizing collective action, such sites get varying amounts of at-
tention and are far from being similarly successful. Locating the sites that
genuinely serve as focal points for collective action remains an intricate
task.
Why do search engines efficiently expose organizational hubs? As
noted above, Google and other search engines rely heavily on popularity to
determine the relevance of search results. Thus, Google and its cousins
serve as sensitive barometers that reveal, in our case, the sites that many
people think are important access points for a certain collective effort.
Typically, they channel users to popular sites that many people found rele-
vant and important enough to link their sites to. For example, if one looks
for information on a community protest against a development plan, the
results obtained from the search engine are sites that, according to many
people, include important information about the local protest. Search en-
gines also enable an easy path to these access points, and direct traffic
primarily (but not exclusively) to such focal sites.10
8 Admittedly, communities are generally better able to route potential contributors
to relevant collaborative projects, since they (unlike search engines) can include
large pools of agents who select to join the community and have some interest
or expertise relevant to the focal theme of the community. The combination of
scale, self-selection and some ‘local expertise’ means that community members
are more likely to be, as a general rule, motivated and to take an active interest
in a relevant collective effort, than just an aggregation of search-engine users.
9 Elsewhere, in a manuscript co-written with Russell Hardin, we argue that inter-
net communication is conducive for such large-scale collaborations. We argue
that much of the success of such collaborations should be attributed to the avail-
ability of the internet as a shared communicative and organizational platform,
the large and excessive number of potential contributors attracted to focal col-
laborations, and the reduction of costs of both individual contributions and the
social organization of production (Lev-On and Hardin 2007).
10 A suggested above, search engines even perpetuate the popularity of such focal
sites (assuming that more popular and accessible sites are linked-to more often
than less popular and accessible sites).
12 Azi Lev-On12
Think of a movement like the Falun Gong, which is now banned in
China and operates from outside its borders (and, also, is blocked by the
‘great firewall’.) Searching for activities organized by Falun Gong in the
uncensored version of Google does not direct users to obscure sites that in-
cidentally mention ‘Falun Gong;’ instead, it directs them to sites including
relevant information about the movement and its activities. Many (proba-
bly most) people that use Google to look for ‘Falun Gong’ (or related
keywords) are routed to the same small set of relevant destinations. Search
engines, then, allow many surfers to easily distinguish popular sites from
unpopular sites, and converge into a small set of focal sites.
Organizers of collective action increasingly capitalize on the centrality-
enhancing property of search engines. Often they ask supporters to install
links from their personal sites to the site that organizes collective action.
For example, when a visitor opposed to a local development project em-
beds in her website an icon that is linked to the website of a group that or-
ganizers the opposition for the development plans, this act increases the
popularity of the group’s website. Even if organizers do not think strategi-
cally when asking contributors to install such links, this practice assists in
making the site more popular and, as a consequence, more easily located.11
Search engines, then, indirectly assist in organizational efforts by ex-
posing focal organizational hubs and routing people there, providing a
channel for people with similar interests to seamlessly coordinate their ef-
forts.
V
So far I argued that search engines contribute to unintended exposure to
diverse and opposing views, and indirectly contribute to the organization
of collective action. Why are such contributions significant to democracy?
To answer this question in a nutshell I will draw on insights from democ-
ratic theory. Space limitations will obviously make the remaining discus-
sion somewhat sketchy.
Elsewhere I argued (with Bernard Manin, 2007) that exposure to di-
verse and especially opposing views contributes to the deliberative quali-
ties of democratic discussion.12 There is a long tradition of liberal theory
11 There is a notable family resemblance between such practices and practices of
search-engine optimization, i.e. strategic inflation of inbound links and similar
techniques which aim at pushing a site up the search-engine result list, and gain-
ing the attention of search-engine users.
12 The following few sections borrow from Lev-On and Manin (2007).
8 The Democratizing Effects of Search Engine Use: On Chance Exposures and
Organizational Hubs 13
13
praising the benefits of diverse and conflicting views for adequate delib-
eration (for a recent exploration see Mutz 2006, especially ch. 3). Mill
(1991:26, emphasis mine) who discusses this topic extensively in his ‘On
Liberty’, praises the benefits that can occur when opposing views confront
each other, and argues that even “[T]he most intolerant of churches, the
Roman Catholic Church, even at a canonization of a saint, admits, and pa-
tiently listens to a ‘devil’s advocate’.” Empirical evidence support some of
the theoretic assertions, and show that exposure to opposing views is in-
strumental for deliberation as it generates such qualities as lack of polari-
zation and radicalization, knowledge gains, more considered opinions, sat-
isfaction from the deliberative process, and enhanced feelings of efficacy
(Price and Cappella 2002; Iyengar, Fishkin and Luskin 2003; Muhlberger
2005).
But exposing agents to opposing views during deliberation entails a
number of challenges. First, typically there are substantial opportunity
costs for the deliberating agents, as deliberation takes time and cognitive
resources that may be devoted to other issues, more aligned with the
deliberants’ interests and concerns.
Second, debates with an adversarial character need ‘enhanced’ promo-
tion and organization, since they require participants to face conflict and
generate talk across cleavages. But research shows that people tend to
avoid the psychic discomfort of expressing opposing views and becoming
involved in contentious discussions. Furthermore, as mentioned earlier, re-
search shows that when people do talk about politics, they do so primarily
with like-minded others.
Democratic deliberation, then, is a complex public good whose facilita-
tion has to overcome a number of obstacles: opportunity costs, generating
cross-cleavage communication, overcoming conflict avoidance. But orga-
nizing exposure to diversity of views, and especially to opposing views, is
difficult to generate in the course of our daily lives. Mill’s interpretation of
the role of the advocatus diaboli is a mistake: the presence of a devil’s ad-
vocate is required precisely because no one may spontaneously take the
other side. This is where search engines get in. I argued that search engines
can facilitate exposure to diverse and even opposing views, even against
people’s intentions. Widely used to seek and obtain political information,
search engines can thus enrich democratic deliberation, and are a welcome
addition to the few spheres in which unplanned exposures to diverse and
opposing views are viable.
While exposure to diverse and opposing views may be essential for
certain models of democracy, other models emphasize political organiza-
tion over deliberation. Realist models of democracy propose, with
14 Azi Lev-On14
Schattschneider (1960, 139), that as a general rule “conflict, competition,
leadership, and organization are the essence of democratic politics,” and
that “the possibility of contestation by conflicting interests is sufficient to
explain the dynamic of democracy” (Przeworski 1991, 10). Notably, plu-
ralist models of democracy depict it as a process of mutual adjustments be-
tween a variety of organized partisan interests. Democratic pluralism em-
phasizes the importance of a variety of competitive channels to influence
policy, and the need to enable multiple groups to organize and influence
the policy-making process (see notably Lindblom 1965, Dahl 1967).
However, scholars realize that the competition in actually-existing de-
mocracies is highly imperfect, due to such factors as high organization and
entry costs. As a result of the disparity of organizational abilities between
different groups, policy areas are dominated by those groups that are better
financed and organized, where unorganized interests can sink into oblivion
and latency.
By now it has become common wisdom that internet communication
drastically reduces the costs of establishing organizations to promote a va-
riety of causes that were previously squeezed out of the political market-
place. Consequently, it is much easier to generate effective voice for
causes that would not otherwise be actively promoted. Internet communi-
cation supports novel intermediaries that supplement existing intermediar-
ies to generate an ‘advocacy explosion’ (Shapiro 1999; Bimber 2003), by
expanding the organizational abilities of a variety of actors to frame and
articulate issues, mobilize support and effectively make political demands.
Arguably, the internet contributes to making the market for intermediaries
more competitive, and hence to improving competition in democracies.
Search engines take an indirect role and make a modest contribution to
the enhancement of political organization and competition, as they support
the creation of focal organizational hubs that are necessary for collective
action. Search engines expose those central sites that many agents value as
organizational hubs, and allow many others an easy route to the same set
of focal sites. Thus, they contribute to the reduction of organizational costs
of a variety of interests.
To emphasize the importance of centrality, let us return shortly to Mill
(1991, 424), who in his Considerations on Representative Government (in
a discussion of the tensions between central and local authorities) argues
that:
Power may be localised, but knowledge, to be most useful, must be centralised;
there must be somewhere a focus at which all its scattered rays are collected, that
the broken and coloured lights which exist elsewhere may find there what is nec-
essary to complete and purify them.
8 The Democratizing Effects of Search Engine Use: On Chance Exposures and
Organizational Hubs 15
15
In the internet, a highly decentralized environment of political information,
search engines constitute such focal points for the ‘scattered rays’ of
knowledge. They also serve as focal points to locate collaborative projects.
Still, the ‘purification’ that search engines allow is imperfect, and hence
they can generate unplanned encounters to diverse and opposing views,
much more effectively than offline.
While arguably advantageous for democracies, the two properties that I
surveyed (unintentional exposure to diverse and opposing views, and indi-
rect support for political organization) may easily be perceived as threats
by authoritarian regimes. Let us go back to the Chinese case I opened with.
If Chinese citizens were able to seek information about Jiang Zemin in the
uncensored version of Google, they could at times come upon information
praising him, but at other times get exposed to information smearing the
leader (as the earlier story shows), largely depending on the popularity of
the sites containing the information. Note that information critical of the
leader can become available not only to Jiang’s opponents who seek such
information to support their prior opinions but also to loyal supporters, and
even to innocent elementary school students seeking information for a
short presentation about the leader’s legacy.
More importantly, anecdotal evidence suggest that the harsh Chinese
monitoring of the internet is also motivated, in large part, by fears from
unleashing popular or factional organizations through such novel tech-
nologies. For example, the ruthless crackdown of Falun Gong was trig-
gered by a large unauthorized gathering of between 10,000 to 15,000 sup-
porters outside the central leadership compound in Beijing in April 25,
1999. The gathering was orchestrated primarily online. Lin (2001) argues
that this has been the largest reasonable-size unauthorized gathering in the
history of modern China on which the authorities failed to receive prior in-
formation. This case alerted authorities to the ability of internet-supported
movements to organize mass meetings and demonstrations while escaping
the attention of the security services.13 Clive Thompson, in his analysis of
‘China’s Google problem,’ also points to the acuteness of fears from web-
based political organization by the Chinese authorities. Thompson (2006,
71) quotes Zhao Jing, “China’s most famous political blogger” whose blog
has been shut down in the time of writing, who claims that “If you talk
every day online and criticize the government, they don’t care… [b]ecause
13 For extensive reviews of the movement and its uses of the internet, see Bell and
Boas 2003 and Chase and Mulvenon 2002.
16 Azi Lev-On16
it’s just talk. But if you organize- even if it’s just three or four people-
that’s what they crack down on. it’s not speech; it’s organizing.”
VI- Conclusions
I argued that search engines indirectly advance political organization, and
generate unintentional exposures to diverse and opposing views. They thus
cater to the concerns of both deliberative democrats aiming at enriching
the deliberative qualities of democratic discussion, and pluralist democrats
who are concerned about making the political marketplace more open, in-
clusive and competitive.
On the other hand, Authoritarian governments aim at avoiding unpre-
dictability and chance exposures to critical information, and at depressing
emergent organization. In this, The Chinese government closely follows
not Mill, but Hobbes’ key advice to governments to keep a keen eye on
dangers originating from dissemination of ‘seditious doctrines’ and coor-
dination of anti-establishment powers (see Hardin 1991). The internet and
search engines are perceived as particularly disruptive. As argued above, at
times search engines expose people to ‘unwarranted’ information, even
against their intentions. Search engines can also be used by many people to
locate and converge on organizational hubs. Sometimes search engines do
both these things -- they expose many people, even against their intentions,
to hubs containing information that authorities disapprove of.
Authoritarian governments, then, need to monitor and regulate search
engines in order to suppress exposure to ‘unwarranted information’ and
prevent unauthorized emergent organization. The political importance of
search engines is clearly demonstrated by the actions of the Chinese gov-
ernment. It is equally important for advocates of open and democratic so-
cieties to constantly monitor the functioning of search engines, and to ver-
ify that they continue to support and enrich the informational and
organizational infrastructure of democracy.
References
Barabási AL (2002) Linked: The new science of networks. Perseus, Cambridge,
MA.
Battelle J (2005) The search. Portfolio, New York.
Bell MR, Boas TC (2003) Falun Gong and the internet: Evangelism, community,
and struggle for survival. Nova Religio 6: 277-293.
8 The Democratizing Effects of Search Engine Use: On Chance Exposures and
Organizational Hubs 17
17
Bimber B (2003) Information and American democracy: Technology in the evolu-
tion of political power. Cambridge University Press, New York.
Brin S, Page L (1998) The anatomy of large-scale hypertextual web search engine.
In: Proceedings of the seventh World Wide Web conference.
Chase MS., Mulvenon JC (2002) You've got dissent! Chinese dissident use of the
Internet and Beijing's counter-strategies. RAND, Santa Monica, CA.
Conover PJ, Searing DD, Crewe IM (2002) The deliberative potential of political
discussion. British Journal of Political Science 32: 21-62.
Dahl RA (1967) Pluralist democracy in the United States: Conflict and consent.
Rand McNally, Chicago.
Drezner D, Farrell H (2004) The power and politics of blogs. Paper presented at
the annual meeting of the American Political Science Association, Chicago.
Fallows D (2005) Search engine users. Pew Internet & American Life Project.
http://www.pewinternet.org/pdfs/PIP_Searchengine_users.pdf
Fortunato S, Flammini A, Menczer F, Vespignani A (2006) The egalitarian effect
of search engines. http://arxiv.org/PS_cache/cs/pdf/0511/0511005.pdf
Hardin R (1991) Hobbesian political order. Political Theory 19: 156-180.
Hargittai E. Hurdles to information seeking: Spelling and typographical mistakes
during users' online behavior. Forthcoming in Journal of the Association of
Information Systems.
Hindman M, Tsioutsiouliklis K, Johnson JA (2003) Googlearchy: How a few
heavily-linked sites dominate politics online.
http://www.princeton.edu/~mhindman/googlearchy--hindman.pdf
Huberman BA (2001) The laws of the web: Patterns in the ecology of information.
MIT Press, Cambridge, MA.
Huckfeldt R, Sprague J (1995) Citizens, politics, and social communication. Cam-
bridge University Press, New York.
Iyengar S, Luskin R, Fishkin J (2003) Facilitating informed public opinion: Evi-
dence from face-to-face and on-line deliberative polls.
http://cdd.stanford.edu/research/papers/2003/facilitating.pdf
Kim J, Wyatt RO, Katz E (1999) News, talk, opinion, participation: The part
played by conversation in deliberative democracy. Political Communication
16: 361-385.
Leach M (2005) MMR mobilisation: citizens and science in a British vaccine con-
troversy. http://www.drc-
citizenship.org/docs/publications/citizens_and_science/WP/wp247.pdf
Lev-On A, Manin B (2007). Happy accidents: Deliberation and online exposure to
opposing views. Forthcoming In: Davies T (ed) Online deliberation: Design,
research and practice.
Lev-On A, Hardin R (2007). Internet-based collaborations and their political
significance. Manuscript submitted for publication.
Lin N (2001) Social capital: A theory of social structure and action. Cambridge
University Press, New York.
Lindblom CE (1965) The intelligence of democracy: Decision making through
mutual adjustment. Free Press, New York.
18 Azi Lev-On18
Machill M, Neuberger C, Schweiger W, Wirth W (2004). Navigating the internet:
A study of German-language search engines. European Journal of Communi-
cation 19: 321-347.
Mill JS (1991) John Stuart Mill: On Liberty and Other Essays, ed. Gray J. Oxford
University Press, New York.
Muhlberger P (2005) The virtual agora project: A research design for studying
democratic deliberation. Journal of Public Deliberation 1(1), article 5.
http://services.bepress.com/jpd/vol1/iss1/art5
Mutz DC (2006) Hearing the other side: Deliberative versus participatory democ-
racy online. Cambridge University Press, New York.
Olson M (1965) The logic of collective action. Harvard University Press, Cam-
bridge, MA.
Price V, Cappella JN (2002) Online deliberation and its influence: The electronic
dialogue project in campaign 2000. Information Technology and Society 1:
303-328.
Przeworski A (1991) Democracy and the market: Political and economic reforms
in eastern Europe and latin America. Cambridge University Press, New York.
Schattschneider EE (1960) The semisovereign people: A realist's view of democ-
racy in America. Holt, Rinehart and Winston, New York.
Shapiro A (1999) The control revolution. Public Affairs, New York.
Spink A, Jansen NJ (2004) A study of web search trends. Webology: An Interna-
tional Electronic Journal 1(2)
Sunstein CR (2001) Republic.com. Princeton University Press, Princeton, NJ.
Theiss-Morse E, Hibbing JR (2005) Citizenship and civic engagement. Annual
Review of Political Science 8: 227–249.
Thompson C. Google's China problem (and China's Google problem). New York
Times Magazine, April 23, 2006.
Tianliang Z. Google "kowtows" to the Chinese government. Epoch Times, Febru-
ary 8, 2005.
