After reading the article “Even Good Bots Fight: The Case of Wikipedia” (Article attached), discuss how AI plays a role in communicating with other technology and humans. How do you think technology will change the way bots interact with each other? What would you like to see implemented that would make your life easier?
RESEARCH ARTICLE
Don't use plagiarized sources. Get Your Custom Essay on
violations, greet newcomers, and so on [19]. Our analysis here focuses on editing bots, which
modify articles directly. We analyze the interactions between bots and investigate the extent to
which they resemble interactions between humans. In particular, we focus on whether bots
disagree with each other, how the dynamics of disagreement differ for bots versus humans,
and whether there are differences between bots operating in different language editions of
Wikipedia.
To measure disagreement, we study reverts. A revert on Wikipedia occurs when an editor,
whether human or bot, undoes another editor’s contribution by restoring an earlier version of
the article. Reverts that occur systematically indicate controversy and conflict [20–22]. Reverts
are technically easy to detect regardless of the context and the language, so they enable analysis
at the scale of the whole system.
Our data contain all edits in 13 different language editions of Wikipedia in the first ten
years after the encyclopedia was launched (2001–2010). The languages represent editions of
different size and editors from diverse cultures (see Materials and Methods for details). We
know which user completed the edit, when, in which article, whether the edit was a revert and,
if so, which previous edit was reverted. We first identified which editors are humans, bots, or
vandals. We isolated the vandals since their short-lived disruptive activity exhibits different
time and interaction patterns than the activity of regular Wikipedia editors.
Results
Bots constitute a tiny proportion of all Wikipedia editors but they stand behind a significant
proportion of all edits (Fig 1A and 1B). There are significant differences between different lan-
guages in terms of how active bots are. From previous research, we know that, in small and
endangered languages, bots are extremely active and do more than 50% of the edits, sometimes
Even good bots fight
PLOS ONE | DOI:10.1371/journal.pone.0171774 February 23, 2017 3 / 13
up to 100% [19]. Their tasks, however, are mainly restricted to adding links between articles
and languages. In large and active languages, the level of bot activity is much lower but also
much more variable.
Compared to humans, a smaller proportion of bots’ edits are reverts and a smaller propor-
tion get reverted (Fig 1C and 1D). In other words, bots dispute others and are disputed by oth-
ers to a lesser extent than humans. Since 2001, the number of bots and their activity has been
increasing but at a slowing rate (S1 Fig). In contrast, the number of reverts between bots has
been continuously increasing (Fig 2A). This would suggest that bot interactions are not
becoming more efficient. We also see that the proportion of mutual bot-bot reverts has
remained relatively stable, perhaps even slightly increasing over time, indicating that bot own-
ers have not learned to identify bot conflicts faster (Fig 2B).
In general, bots revert each other a lot: for example, over the ten-year period, bots on
English Wikipedia reverted another bot on average 105 times, which is significantly larger
than the average of 3 times for humans (S1 Table). Bots on German Wikipedia revert each
other to a much lesser extent than other bots (24 times on average). Bots on Portuguese Wiki-
pedia, in contrast, fight the most, with an average of 185 bot-bot reverts per bot. This striking
difference, however, disappears when we account for the fact that bots on Portuguese Wikipe-
dia edit more than bots on German Wikipedia. In general, since bots are much more active
Fig 1. The proportion of Wikipedia editors who are human, vandals, and bots and the type of editorial
activity in which they are involved. A language edition to the left has a higher total number of edits than one
to the right. (A) Bots comprise a tiny proportion of all Wikipedia users, usually less than 0.1% (not visible in the
figure). (B) However, bots account for a significant proportion of the editorial activity. The level of bot activity
significantly differs between different language editions of Wikipedia, with bots generally more active in
smaller editions. (C) A smaller proportion of bots’ edits are reverts compared to humans’ edits. (D) A smaller
proportion of bots’ edits get reverted compared to humans’ edits. Since by our definition, vandals have all of
their edits reverted, we do not show them in this figure.
doi:10.1371/journal.pone.0171774.g001
Even good bots fight
PLOS ONE | DOI:10.1371/journal.pone.0171774 February 23, 2017 4 / 13
editors than humans, the higher number of bot-bot reverts does not mean that bots fight more
than humans. In fact, the proportion of bots’ edits that are reverts is smaller for bots than for
humans (Fig 1C). This proportion is highest for bots in the English and the Romance-language
editions (Spanish, French, Portuguese, and Romanian). Interestingly, although bots in these
languages revert more often compared to bots in other languages, fewer of these reverts are for
another bot (S2 Fig).
Our analysis focuses on interactions in dyads over time. We model the interaction trajecto-
ries in two-dimensional space, where the x-axis measures time and the y-axis measures how
many more times the first editor has reverted the second compared to the second reverting the
first (Fig 3). We analyze three properties of the trajectories: latency, imbalance, and reciprocity.
Latency measures the average steepness of the interaction trajectory, imbalance measures the
distance between the x-axis and the last point of the trajectory, and reciprocity measures the
trajectory’s jaggedness (see Materials and Methods below for definitions).
Analyzing the properties of the interaction trajectories suggests that the dynamics of dis-
agreement differ significantly between bots and humans. Reverts between bots tend to occur at
a slower rate and a conflict between two bots can take place over longer periods of time, some-
times over years. In fact, bot-bot interactions have different characteristic time scale than
human-human interactions (S3 Fig). The characteristic average time between successive
reverts for humans is at 2 minutes, 24 hours, or 1 year. In comparison, bot-bot interactions
have a characteristic average response of 1 month. This difference is likely because, first, bots
systematically crawl articles and, second, bots are restricted as to how often they can make
Fig 2. The number of bot reverts executed by another bot and the proportion of unique bot-bot pairs
that have at least one reciprocated revert for the period 2001–2010. (A) Generally, the number of bot-bot
reverts has been increasing. (B) However, the proportion of reciprocated reverts has not been decreasing
(error bars correspond to one standard error). This suggests that disagreement between bots is not becoming
less common.
doi:10.1371/journal.pone.0171774.g002
Even good bots fight
PLOS ONE | DOI:10.1371/journal.pone.0171774 February 23, 2017 5 / 13
edits (the Wikipedia bot policy usually requires spacing of 10 seconds, or 5 for anti-vandalism
activity, which is considered more urgent). In contrast, humans use automatic tools that report
live changes made to a pre-selected list of articles [24,25]; they can thus follow only a small set
of articles and, in principle, react instantaneously to any edits on those.
Bots also tend to reciprocate each other’s reverts to a greater extent. In contrast, humans
tend to have highly unbalanced interactions, where one individual unilaterally reverts another
one (S4 and S5 Figs).
We quantify these findings more precisely by identifying different types of interaction tra-
jectories and counting how often they occur for bots and for humans, as well as for specific
languages. To this end, we use k-means clustering on the three properties of the trajectories
(latency, imbalance, and reciprocity) and on all bot-bot and human-human interactions longer
than five reverts (the results are substantively similar without the length restriction). We do
not claim that the clusters are natural to the data; rather, we use the clusters to compare the
interactions of the different groups.
Fig 3. Typical interaction trajectories for bot-bot and human-human pairs in English Wikipedia in the
period 2001–2010. The interaction trajectories are constructed as follows: starting from yo = 0, yt = yt-1 + 1 if i
reverts j and yt = yt-1 − 1 if j reverts i at time t; the labels i and j are assigned so that y >= 0 for the majority of the
ij interaction time; to compress the extremes, we scaled the y-axis to the power of 0.5. The panels show the
trajectories of 200 pairs randomly sampled from those who have exchanged more than five reverts. In
addition, we highlight the four longest trajectories in the sample from each of the four trajectory types we
identify. Compared to human-human interactions, bot-bot interactions occur at a slower rate and are more
balanced, in the sense that reverts go back and forth between the two editors.
doi:10.1371/journal.pone.0171774.g003
Even good bots fight
PLOS ONE | DOI:10.1371/journal.pone.0171774 February 23, 2017 6 / 13
The algorithm suggested that the data can be best clustered in four trajectory types (S6 Fig):
• Fast unbalanced trajectories. These trajectories have low reciprocity and latency and high
imbalance. They look like smooth vertical lines above the x-axis.
• Slow unbalanced trajectories. These trajectories have low reciprocity and high latency and
imbalance. They look like smooth diagonal lines above the x-axis.
• Somewhat balanced trajectories. These trajectories have intermediate imbalance and reci-
procity. They are somewhat jagged and cross the x-axis.
• Well balanced trajectories. These trajecto
ries have low imbalance and high reciprocity.
They are quite jagged and centered on the x-axis.
Looking at the prevalence of these four types of trajectories for bots and humans and across
languages, we confirm the previous observations: bot-bot interactions occur at a slower rate
and are more balanced, in the sense that reverts go back and forth between the two bots (Fig
4). Further, we find that bot-bot interactions are more balanced in smaller language editions of
Wikipedia. This could be due to the fact that bots are more active in smaller editions and
hence, interactions between them are more likely to occur. Less intuitively, however, this
observation also suggests that conflict between bots is more likely to occur when there are
fewer bots and when, common sense would suggest, coordination is easier.
Discussion
Our results show that, although in quantitatively different ways, bots on Wikipedia behave and
interact as unpredictably and as inefficiently as the humans. The disagreements likely arise
from the bottom-up organization of the community, whereby human editors individually
create and run bots, without a formal mechanism for coordination with other bot owners.
Fig 4. The prevalence of the four types of trajectories for bots and humans and for different language editions of
Wikipedia. The darker the shading of the cell, the higher the proportion for that type of trajectory for the language. Bot-bot
interactions occur at a slower rate and are more balanced, in the sense that reverts go back and forth between the two bots.
Further, bot-bot interactions are more balanced in smaller language editions of Wikipedia.
doi:10.1371/journal.pone.0171774.g004
Even good bots fight
PLOS ONE | DOI:10.1371/journal.pone.0171774 February 23, 2017 7 / 13
Delving deeper into the data, we found that most of the disagreement occurs between bots that
specialize in creating and modifying links between different language editions of the encyclo-
pedia. The lack of coordination may be due to different language editions having slightly dif-
ferent naming rules and conventions.
In support of this argument, we also found that the same bots are responsible for the major-
ity of reverts in all the language editions we study. For example, some of the bots that revert
the most other bots include Xqbot, EmausBot, SieBot, and VolkovBot, all bots specializing in
fixing inter-wiki links. Further, while there are few articles with many bot-bot reverts (S7 Fig),
these articles tend to be the same across languages. For example, some of the articles most con-
tested by bots are about Pervez Musharraf (former president of Pakistan), Uzbekistan, Estonia,
Belarus, Arabic language, Niels Bohr, Arnold Schwarzenegger. This would suggest that a sig-
nificant portion of bot-bot fighting occurs across languages rather than within. In contrast, the
articles with most human-human reverts tend to concern local personalities and entities and
tend to be unique for each language [26].
Our data cover a period of the evolution of Wikipedia when bot activity was growing. Evi-
dence suggests that this period suddenly ended in 2013 (http://stats.wikimedia.org/EN/
PlotsPngEditHistoryTop.htm). This decline occurred because at the beginning of 2013 many
language editions of Wikipedia started to provide inter-language links via Wikidata, which is a
collaboratively edited knowledge base intended to support Wikipedia. Since our results were
largely dictated by inter-language bots, we believe that the conflict we observed on Wikipedia
no longer occurs today. One interesting direction for future research is to investigate whether
the conflict continues to persist among the inter-language bots that migrated to Wikidata.
Wikipedia is perhaps one of the best examples of a populous and complex bot ecosystem
but this does not necessarily make it representative. As Table 1 demonstrates, we have investi-
gated a very small region of the botosphere on the Internet. The Wikipedia bot ecosystem is
gated and monitored and this is clearly not the case for systems of malevolent social bots, such
as social bots on Twitter posing as humans to spread political propaganda or influence public
discourse. Unlike the benevolent but conflicting bots of Wikipedia, many malevolent bots are
collaborative, often coordinating their behavior as part of botnets [27]. However, before being
able to study the social interactions of these bots, we first need to learn to identify them [28].
Our analysis shows that a system of simple bots may produce complex dynamics and unin-
tended consequences. In the case of Wikipedia, we see that benevolent bots that are designed
to collaborate may end up in continuous disagreement. This is both inefficient as a waste of
resources, and inefficacious, for it may lead to local impasse. Although such disagreements
represent a small proportion of the bots’ editorial activity, they nevertheless bring attention to
the complexity of designing artificially intelligent agents. Part of the complexity stems from
the common field of interaction—bots on the Internet, and in the world at large, do not act in
isolation, and interaction is inevitable, whether designed for or not. Part of the complexity
stems from the fact that there is a human designer behind every bot, as well as behind the envi-
ronment in which bots operate, and that human artifacts embody human culture. As bots con-
tinue to proliferate and become more sophisticated, social scientist will need to devote more
attention to understanding their culture and social life.
Materials and methods
Data
Wikipedia is an ecosystem of bots. Some of the bots are “editing bots”, that work on the arti-
human and bot layers. The multi-layer network was pruned by removing self-reverts, as well
as reverts between a bot and its owner.
Interaction trajectories
We model the interaction trajectories in two-dimensional space, where the x-axis measures time
and the y-axis measures the difference between the number of times i has reverted j and the
number of times j has reverted i. To construct the trajectories, starting from y0 = 0, yt = yt-1 + 1
if i reverts j at time t and yt = yt-1 − 1 if j reverts i at time t; the labels i and j are assigned so that
y >= 0 for the majority of the ij interaction time. We analyze three properties of the trajectories:
• Latency. We define latency as the mean log time in seconds between successive reverts:
μ(log10 Δt).
• Imbalance. We define imbalance as the final proportion of reverts between i and j that were
not reciprocated: |ri − rj| / (ri + rj), where ri and rj are the number of times i reverted j and j
reverted i, respectively.
• Reciprocity. We define reciprocity as the proportion of observed turning points out of all
possible: (# turning points) / (ri + rj− 1), where ri and rj are the number of times i reverted j
and j reverted i, respectively. A turning point occurs when the user who reverts at time t is
different from the user who reverts at time t+1.
K-means clustering
To identify the number of clusters k that best represents the data, we apply the elbow and sil-
houette methods on trajectories of different minimum length. The rationale behind restricting
the data to long trajectories only is that short trajectories tend to have extreme values on the
three features, thus possibly skewing the results. According to the elbow method, we would
like the smallest k that most significantly reduces the sum of squared errors for the clustering.
According to the silhouette method, we would like the k that maximizes the separation dis-
tance between clusters and thus gives us the largest silhouette score.
Although the elbow method suggests that four clusters provide the best clustering, the sil-
houette method indicates that the data cannot be clustered well (S8 Fig). We do not necessarily
expect that trajectories cluster naturally; rather, we employ clustering in order to quantify the
differences between the interactions of bots versus humans across languages. We hence ana-
lyze the clustering with k = 4. This clustering also has the advantage of yielding four types of
trajectories that intuitively make sense.
Supporting information
S1 Fig. The number of bots, the number of edits by bots, and the proportion of edits done
by bots between 2001 and 2010. Between 2003 and 2008 the number of bots and their activity
have been increasing. This trend, however, appears to have subsided after 2008, suggesting
that the system may have stabilized.
(TIFF)
S2 Fig. For the majority of languages, bots are mainly reverted by other bots, as opposed to
human editors or vandals. English and the Romance languages in our data present excep-
tions, with less than 20% of bot reverts are done by other bots.
(TIFF)
Even good bots fight
PLOS ONE | DOI:10.1371/journal.pone.0171774 February 23, 2017 10 / 13
S3 Fig. Bot-bot interactions have different characteristic time scale than human-human
interactions. The figures show the distribution of interactions for a particular latency, where
we define latency as the mean log time in seconds between successive reverts. (A) Bot-bot
interactions have a characteristic latency of 1 month, as indicated by the peak in the figure. (B)
Human-human interactions occur with a latency of 2 minutes, 24 hours, or 1 year.
(TIFF)
S4 Fig. Bot-bot interactions are on average more balanced than human-human interac-
tions. We define imbalance as the final proportion of reverts between i and j that were not
reciprocated. (A) A significant proportion of bot-bot interactions have low imbalance. (B) The
majority of human-human interactions are perfectly unbalanced.
(TIFF)
S5 Fig. Bots reciprocate much more than humans do also at a smaller timescale. We mea-
sure reciprocity as the proportion of observed turning points out of all possible. (A) A signifi-
cant proportion of bot-bot interactions have intermediate or high values of reciprocity. (B)
The majority of human-human interactions are not reciprocated.
(TIFF)
S6 Fig. Four types of interaction trajectories suggested by the k-means analysis. The left
panels show a sample of the trajectories, including bot-bot and human-human interactions
and trajectories from all languages. The right panels show the distribution of latency, imbal-
ance, and reciprocity for each type of trajectory. The three properties measure the average
steepness, the y-value of the last point, and the jaggedness of the trajectory, respectively. (A)
Fast unbalanced trajectories have low reciprocity and latency and high imbalance. (B) Some-
what balanced trajectories have intermediate imbalance and reciprocity. (C) Slow unbalanced
trajectories have low reciprocity and high latency and imbalance. (D) Well balanced trajecto-
ries have low imbalance and high reciprocity.
(TIFF)
S7 Fig. The number of articles with a certain number of bot-bot and human-human
reverts. (A) Few articles include more than 10 bot-bot reverts. The most contested articles
tend to be about foreign countries and personalities. Further, the same articles also re-
appear in different languages. (B) There are many articles that are highly contested by
humans. The most contested articles tend to concern local personalities and entities. It is
rare that a highly contested article in one language will be also highly contested in another
language.
(TIFF)
S8 Fig. Performance of the k-means clustering algorithm for different number of clusters
and for sub-samples with different minimum length of trajectories. (A) The elbow method
requires the smallest k that most significantly reduces the sum of squared errors for the cluster-
ing. Here, the method suggests that four clusters give the best clustering of the data. (B) The sil-
houette method requires the k that maximizes the separation distance between clusters, i.e. the
largest silhouette score. Here, the method suggests that the clustering performs worse as the
number of clusters increases.
(TIFF)
S1 Table. Descriptive statistics for the bot-bot layer and the human-human layer in the
multi-layer networks of reverts. Bots revert each other to a great extent. They also reciprocate
each other’s reverts to a considerable extent. Their interactions are not as clustered as for
human editors. Still, both for bots and humans, more senior editors tend to revert less senior
Even good bots fight
PLOS ONE | DOI:10.1371/journal.pone.0171774 February 23, 2017 11 / 13
Copyright of PLoS ONE is the property of Public Library of Science and its content may not
be copied or emailed to multiple sites or posted to a listserv without the copyright holder’s
express written permission. However, users may print, download, or email articles for
individual use.
What Will You Get?
We provide professional writing services to help you score straight A’s by submitting custom written assignments that mirror your guidelines.
Premium Quality
Get result-oriented writing and never worry about grades anymore. We follow the highest quality standards to make sure that you get perfect assignments.
Experienced Writers
Our writers have experience in dealing with papers of every educational level. You can surely rely on the expertise of our qualified professionals.
On-Time Delivery
Your deadline is our threshold for success and we take it very seriously. We make sure you receive your papers before your predefined time.
24/7 Customer Support
Someone from our customer support team is always here to respond to your questions. So, hit us up if you have got any ambiguity or concern.
Complete Confidentiality
Sit back and relax while we help you out with writing your papers. We have an ultimate policy for keeping your personal and order-related details a secret.
Authentic Sources
We assure you that your document will be thoroughly checked for plagiarism and grammatical errors as we use highly authentic and licit sources.
Moneyback Guarantee
Still reluctant about placing an order? Our 100% Moneyback Guarantee backs you up on rare occasions where you aren’t satisfied with the writing.
Order Tracking
You don’t have to wait for an update for hours; you can track the progress of your order any time you want. We share the status after each step.
From brainstorming your paper's outline to perfecting its grammar, we perform every step carefully to make your paper worthy of A grade.
Preferred Writer
Hire your preferred writer anytime. Simply specify if you want your preferred expert to write your paper and we’ll make that happen.
Grammar Check Report
Get an elaborate and authentic grammar check report with your work to have the grammar goodness sealed in your document.
One Page Summary
You can purchase this feature if you want our writers to sum up your paper in the form of a concise and well-articulated summary.
Plagiarism Report
You don’t have to worry about plagiarism anymore. Get a plagiarism report to certify the uniqueness of your work.
Free Features $66FREE
Most Qualified Writer
$10FREE
Plagiarism Scan Report
$10FREE
Unlimited Revisions
$08FREE
Paper Formatting
$05FREE
Cover Page
$05FREE
Referencing & Bibliography
$10FREE
Dedicated User Area
$08FREE
24/7 Order Tracking
$05FREE
Periodic Email Alerts
$05FREE
Our Services
Join us for the best experience while seeking writing assistance in your college life. A good grade is all you need to boost up your academic excellence and we are all about it.
On-time Delivery
24/7 Order Tracking
Access to Authentic Sources
Academic Writing
We create perfect papers according to the guidelines.
Professional Editing
We seamlessly edit out errors from your papers.
Thorough Proofreading
We thoroughly read your final draft to identify errors.
Thorough Proofreading
We thoroughly read your final draft to identify errors.
Delegate Your Challenging Writing Tasks to Experienced Professionals
Work with ultimate peace of mind because we ensure that your academic work is our responsibility and your grades are a top concern for us!
We Analyze Your Problem and Offer Customized Writing
We understand your guidelines first before delivering any writing service. You can discuss your writing needs and we will have them evaluated by our dedicated team.
Clear elicitation of your requirements.
Customized writing as per your needs.
We Mirror Your Guidelines to Deliver Quality Services
We write your papers in a standardized way. We complete your work in such a way that it turns out to be a perfect description of your guidelines.
Proactive analysis of your writing.
Active communication to understand requirements.
We Handle Your Writing Tasks to Ensure Excellent Grades
We promise you excellent grades and academic excellence that you always longed for. Our writers stay in touch with you via email.
Thorough research and analysis for every order.
Deliverance of reliable writing service to improve your grades.
