ArticlePDF Available

How Constraints Affect Content: The Case of Twitter’s Switch from 140 to 280 Characters



It is often thought that constraints affect creative production, both in terms of form and quality. Online social media platforms frequently impose constraints on the content that users can produce, limiting the range of possible contributions. Do these restrictions tend to push creators towards producing more or less successful content? How do creators adapt their contributions to fit the limits imposed by social media platforms? In this work, we conduct a matched observational study to answer these questions. On November 7, 2017, Twitter changed the maximum allowable length of a tweet from 140 to 280 characters, significantly altering its signature constraint. In our studies we compare tweets with nearly or exactly 140 characters before the change to tweets of the same length posted after the change. This setup enables us to characterize how users alter their tweets to fit the constraint and how this affects their tweets' success. We find that in response to a length constraint, users write more tersely, use more abbreviations and contracted forms, and use fewer definite articles. Also, although in general tweet success increases with length, we find initial evidence that tweets made to fit the 140 character constraint tend to be more successful than similar length tweets written when the constraint was removed, suggesting that the length constraint improved tweet quality.
How Constraints Affect Content:
The Case of Twitter’s Switch from 140 to 280 Characters
Kristina Gligori´
Ashton Anderson
University of Toronto
Robert West
It is often said that constraints affect creative production,
both in terms of form and quality. Online social media plat-
forms frequently impose constraints on the content that users
can produce, limiting the range of possible contributions. Do
these restrictions tend to push creators towards producing
more or less successful content? How do creators adapt their
contributions to fit the limits imposed by social media plat-
forms? To answer these questions, we conduct an observa-
tional study of a recent event: on November 7, 2017, Twitter
changed the maximum allowable length of a tweet from 140
to 280 characters, thereby significantly altering its signature
constraint. In the first study of this switch, we compare tweets
with nearly or exactly 140 characters before the change to
tweets of the same length posted after the change. This setup
enables us to characterize how users alter their tweets to fit
the constraint and how this affects their tweets’ success. We
find that in response to a length constraint, users write more
tersely, use more abbreviations and contracted forms, and use
fewer definite articles. Also, although in general tweet suc-
cess increases with length, we find initial evidence that tweets
made to fit the 140-character constraint tend to be more suc-
cessful than similar-length tweets written when the constraint
was removed, suggesting that the length constraint improved
tweet quality.
1 Introduction
The enemy of art is the absence of limitations. —Orson Welles
It is often thought that constraints affect both the form
and quality of creative content. For example, there is anec-
dotal evidence across many domains that imposing length
constraints can shape and improve the resulting writing: aca-
demic authors edit papers to fit a page limit, poets adhere
to a prescribed verse form or rhyme scheme, and journal-
ists edit articles to fit a word count limit (McPhee 2015).
More broadly, research in several fields—spanning product
design (Joyce 2009; Moreau and Dahl 2005), process man-
agement (Fritscher and Pigneur 2009), and education (Hen-
nessey 1989)—suggests that having too much freedom can
be paralyzing (epitomized by the feeling of staring at a blank
sheet of paper), and that there is a sweet spot with just the
right amount of constraints.
Copyright © 2018, Association for the Advancement of Artificial
Intelligence ( All rights reserved.
Many social media platforms enforce length restrictions
for posts, which is why the latter are frequently called mi-
croposts. For instance, at the time of writing, Instagram cap-
tions were limited to 2,200 characters, LinkedIn updates to
700 characters, Pinterest pins to 500 characters, and Twit-
ter posts, tweets, to 280 characters. In the context of social
media, a less restrictive character limit could conceivably be
either positive or negative. In this work, we investigate the
relationship between length constraints on post length and
the form and quality of the resulting content. Do length con-
straints steer users towards editing their content in a useful
way on social media, or does additional space allow for more
creative and engaging content? Answering this question in-
volves resolving two competing hypotheses:
H1 Imposing constraints has a positive effect on creativity,
influencing users to create succinct content that is more
likely to be appealing to others (Joyce 2009).
H2 Relaxing constraints provides users more space for ex-
pressing opinion and allows for more potentially in-
teresting content. Also, longer posts occupy a larger
portion of the update feeds displayed to users (Rosen
2017), further increasing engagement.
Choosing the right character limit is not trivial, given these
conflicting hypotheses. Resolving them would not only help
elucidate the relation between content and constraints; given
the multi-billion-dollar economy depending on social me-
dia, the question is also of significant financial consequence.
The ideal way to answer this question would be through a
randomized control study: subject randomly sampled users
to various character limits, and observe how the nature and
success of their posts depends on the respective limit. Unfor-
tunately, however, only social-media providers themselves
have full control over their platforms, including the ability
to conduct such A/B tests. Furthermore, even this ideal ex-
periment might suffer from creating an ecosystem with dif-
fering content lengths with clashing norms and conventions.
Studying the effect of constraints is therefore difficult.
Twitter’s November 2017 switch. In order to circumvent
these difficulties, we take advantage of a recent event to con-
duct an observational study of how length constraints affect
microposts: on November 7, 2017, Twitter suddenly and un-
expectedly increased its maximum tweet length from 140 to
280 characters.
arXiv:1804.02318v2 [cs.SI] 10 Apr 2018
Before switch
[136, 140]
After switch
[91, 100]
Figure 1: Left: Schema of study setup. Right: Histogram of
tweet lengths, before and after the switch.
Figure 2: Engagement (retweets and favorites) as a function
of length, with 95% confidence intervals. Left: Probability
of receiving at least one engagement. Right: Mean number
of engagements (for tweets with at least one engagement).
The switch was reportedly introduced to allow users to ex-
press their thoughts without running out of characters, thus
preventing them from finishing a thought (Rosen and Ihara
2017). This change in length constraint—which we hence-
forth refer to as the switch—constitutes an exogenous event
that was most likely unexpected to most Twitter users, so we
may reasonably assume that user behavior did not change
in anticipation of this event. After carefully controlling for
certain factors, differences between posts tweeted before vs.
after the switch can inform us about the effect of different
length constraints.
Research questions. Given that our aim is to study the ef-
fect of length constraints on the style and success of content
in online social media platforms, we seek to answer the fol-
lowing questions:
RQ1 Do length constraints lead to characteristic changes in
the writing style of posts?
RQ2 Do length constraints push users toward creating con-
tent that other users are more likely to engage with?
2 Methodology1
Research design. In this paper, our main methodological
contribution is a matched observational study design that al-
lows us to use the exogenous shock to the Twitter ecosystem
to investigate the relationship between length constraints and
content form and quality, even in the absence of full exper-
imental control of Twitter. Before Twitter’s switch from a
140-character limit to a 280-character limit, tweets that are
nearly or exactly 140 characters in length are likely to have
been explicitly “squeezed” by the user to comply with the
character limit. After the switch, tweets of this length are
less likely to have been affected by the character limit, since
they are far short of the new 280-character maximum. Our
basic design is thus to compare various properties of tweets
1More details in the appendix.
of a length between 136 and 140 characters2just before vs.
right after the switch (Fig. 1, left). The major difference be-
ing the presence vs. absence of the 140-character limit, com-
paring tweets of this length from before vs. after the switch
(corresponding to the upper, red arrow in Fig. 1, left) lets
us isolate the impact of the character limit. By comparing
content written very close in time, we minimize the likeli-
hood of external factors changing in the time during which
our studied tweets were written. In a control study, we also
compare tweets with 91 to 100 characters before vs. after the
switch, as these tweets are unaffected by the character limit
during both time periods (lower, gray arrow in Fig. 1, left).
Data. In order to execute this research design, one would
ideally analyze all tweets in our character range of interest
that were authored right before and right after the switch. In
lieu of access to complete Twitter data, we approximate this
ideal. We first collect tweets from the 1% sample Twitter
supplies via its Spritzer API, for the time between April and
June 2017. From this set of tweets we sample 100K users,
where the probability of being sampled is proportional to the
number of 140-character-long tweets they posted. This en-
sures that we are likely to sample users who disproportion-
ately generate tweets for which the 140-character constraint
is relevant. We proceed to collect these users’ timelines us-
ing the Twitter user timeline API, gathering up to the 3,200
most recent tweets per user. We simultaneously keep collect-
ing new tweets these users post. All tweets we analyze were
posted between November 2017 and January 2018.3
Controlling for users and topics. With the goal of mitigat-
ing potential confounds in our comparison of tweets before
vs. after the switch, we perform two analyses. First, we con-
trol for users by including, for each user, the same number
of tweets written before the switch as after the switch (but
this number may vary across users). As we are interested
in matching tweets actually written by a particular user, we
do not consider retweets of other users’ tweets. Second, we
also conduct an analysis in which we control for topics by
including, for each unique set of hashtags, the same number
of tweets before as after the switch with that set of hashtags.
Controlling for temporal effects. By design, we compare
tweets posted before and after the switch, thus introducing
a time gap between the observed content. This raises two
problems: first, retweet and favorite counts depend on how
much time has passed since posting; and second, one might
observe differences in the number of retweets and favorites
as a side effect of a user’s becoming more popular with time.
We address the first issue by making sure to collect a tweet at
least 48 hours after it was posted, since by that time 99.99%
of retweets happen (Sysomos Retweets Study 2010). The
second issue is ruled out by making sure that the number
of a user’s followers did not increase by more than 1% be-
tween the user’s first and last tweets in our dataset, when
controlling for users.
2We use a 5-character range because it is not always possible to
strictly optimize for the limit: omitting one word to meet the con-
straint may result in a tweet of fewer than exactly 140 characters.
3Data available at to 280
Matched for user Matched for topic Control study
Linguistic feature Before After Before After Before After
Number of hashtags 0.53 0.54 0.22 0.22
Number of emojis 0.25 0.25 0.37 0.29 0.31 0.32
Number of abbreviations 0.17 0.13 0.09 0.07 0.12 0.12
Number of definite article 0.68 0.71 0.66 0.69 0.48 0.48
Number of indefinite articles 0.55 0.55 0.44 0.45 0.38 0.39
Number of and 0.38 0.42 0.25 0.32 0.32 0.32
Number of &0.16 0.11 0.18 0.14 0.03 0.03
Number of missing spaces after punctuation 0.59 0.64 0.64 0.68 0.48 0.52
Number of auxiliary verbs and negations in long form 0.73 0.75 0.68 0.66 0.53 0.52
Number of auxiliary verbs and negations in contracted form 0.53 0.44 0.52 0.41 0.39 0.29
Fraction of lexical words (nouns, verbs, adjectives, adverbs) 56.45% 56.39% 55.44% 55.62% 57.34% 57.92%
Lexical variation 0.9089 0.9081 0.9136 0.9121 0.9405 0.9410
Lexical sophistication 0.0014 0.0014 0.0014 0.0014 0.0013 0.0013
Number of words per sentence 17.81 17.86 15.86 15.85 14.82 14.90
Matched for user
Before After
I am 0.011 0.013
I’m 0.057 0.051
have 0.123 0.128
’ve 0.033 0.026
will 0.065 0.068
’ll 0.027 0.021
would 0.033 0.034
’d 0.014 0.011
Table 1: Results of linguistic analysis (RQ1). Left: Linguistic features averaged across tweets, before vs. after switch, when
matching for users and topics (Sec. 3.2), and in control study (Sec. 3.3). Right: Mean number of occurrences per tweet of
expanded and contracted word forms. Bold values are significant with p<0.00365 according to Kolmogorov–Smirnov tests.
Matched for user Matched for topic Control study (tweet length [91,100])
Retweets Favorites Retweets Favorites Retweets Favorites
Before After Before After Before After Before After Before After Before After
Probability of at least one 21.61% 20.38% 45.50% 45.52% 28.78% 29.94% 52.29% 52.17% 19.10% 18.72% 52.04% 52.94%
Quartiles (25%, 50%, 75%) (1, 2, 7) (1, 2, 7) (1, 3, 8) (1, 3, 8) (1, 2, 5) (1, 2, 4) (1, 2, 6) (1, 2, 6) (1, 2, 5) (1, 2, 5) (1, 2, 6) (1, 2, 6)
Table 2: Results of success analysis (RQ2). Top: Probability of receiving at least one engagement (retweet or favorite). Bot-
tom: Quartiles of numbers of engagements (for tweets with at least one engagement).
Constraint consistency over time. We ensure that all con-
tent we analyze was generated under the same character
counting policy. Additionally, we are primarily interested in
how text is altered under a particular restriction. We thus
limit ourselves to tweets not containing URLs and @ replies.
Pagination. Twitter users have long been working around
the 140-character limit by splitting a long piece of text into
a sequence of length-compliant tweets. These tweet threads
are usually annotated with the position of the tweet in its se-
quence, e.g.,2/3. They are usually not intentionally altered
in order to fit the character limit requirement using mecha-
nisms we are interested in. Thus, we disregard such tweets.
3 Results
3.1 General impact of length constraints
We begin with an analysis of the general impact of tweet
length limits. We follow a data-driven approach and ana-
lyze 4M tweets written before, and 1.9M written after, the
switch. Tweets are filtered according to our overall study
setup (Sec. 2).
Fig. 1 (right) contains a histogram of tweet lengths before
and after the switch. The spike at exactly 140 characters be-
fore the switch is indicative of users dealing with the char-
acter limit constraint. Additionally, the fact that there is no
corresponding spike after the switch reflects the fact that the
spike was indeed induced by the constraint—once it is lifted,
140 characters is an unexceptional tweet length.
Fig. 2 presents the probability of obtaining at least one
engagement (retweet or favorite) and the average number
of engagements (given at least one engagement), as a func-
tion of tweet length. We observe that, both before and after
the switch, longer tweets are more successful on average,
as captured by the number of retweets and favorites. This
demonstrates the existence of a length effect consistent with
hypothesis H2 (Sec. 1). This finding is, however, subject to
several potential confounds; e.g., tweet length might corre-
late with the importance of what the user has to say, or with
the popularity of the user herself, which might in turn be
the real cause for the success of the respective tweet. Our
matched studies, described next, address these confounds.
3.2 Matched studies
We compare tweets of length [136,140]posted before, with
tweets of a length in the same range posted after, the switch.
Linguistic analysis (RQ1). We are interested in linguis-
tic aspects that might conceivably be altered in an attempt
to meet length constraints: hashtags, emojis, abbreviations
and acronyms used in the Twitter community (Beal, Vangie
2016), articles, conjunctions, spaces after punctuation, and
auxiliary verbs in their long, contracted, or incorrectly con-
tracted forms. We also evaluate stylistic features of text, in-
cluding the frequency of lexical words, lexical variation, and
sophistication as defined by Lu (2012). Moreover, we com-
pute the number of words per sentence, an indicator of read-
ability. Here, we are not interested in content that fails to
elicit any engagement at all, as it is unlikely to reflect the op-
timizing mechanisms of interest. We thus keep only tweets
that have received at least one retweet or favorite, resulting
in 12K (4K) tweet pairs when matching for users (topics).
Matching for users and for topics, respectively, we com-
pare the above linguistic features before vs. after the switch
using Kolmogorov–Smirnov significance tests. Given that
we test multiple hypotheses (one per feature), the signifi-
cance threshold is adjusted via Dunn– ˇ
ak correction (Abdi
2007), where the adjusted significance threshold is calcu-
lated as 1 (1α)1/m. We set α=0.05 and m=14, result-
ing in an adjusted significance threshold of 0.00365.
Whether we control for users or for topics, constrained
tweets systematically differ in linguistic features (Table 1).
They contain fewer hashtags, fewer articles, and more words
in their short forms: more abbreviations, more auxiliary
verbs in their contracted forms, more &, and fewer and.
These discrepancies are indicative of edits taking place
given a 140-character limit constraint. We do not observe
more missing spaces after punctuation in constrained tweets.
This signals that users are unwilling to use punctuation in an
incorrect way in order to fit a constraint.
Success analysis (RQ2). When quantifying engagement, we
evaluate how much other users engage with a tweet in terms
retweets and favorites. Controlling for users and for topics,
respectively, we evaluate success before vs. after the switch.
The results of this study are presented in Table 2. We see
that, when controlling for users, constrained tweets are 6.0%
(relative) more likely to elicit engagement in the form of at
least one retweet, compared to unconstrained tweets. Note
that the trend is reversed when controlling for topics (in-
stead of users), but the difference is smaller (3.9% rela-
tive), and controlling for users is arguably a stricter crite-
rion than controlling for topics, which, taken together, indi-
cates that length restrictions slightly improve the success of
tweets, in line with hypothesis H1 (Sec. 1).
3.3 Control study for ruling out confounds
As our analysis compares two different time periods to each
other, one might object that the observed stylistic changes
were in fact caused by a community-wide drift of norms
between the two time periods. Although unlikely given the
condensed time frame, we aim to rule out this alternative ex-
planation with a control study in which we repeat our anal-
ysis (matching for users; 9K tweet pairs), but this time com-
paring tweets of a length much below the 140-character limit
(91 to 100 characters) before vs. after the switch (cf. Sec. 2).
Such short tweets are unlikely to have been optimized for
the length constraint. As a consequence, if any differences
are observed for [91,100]tweets, the same mechanisms
might be at work for [136,140]tweets, and it might be these
mechanisms—rather than the 140-character constraint—that
also cause the differences among [136,140]tweets. How-
ever, we observe only one statistically significant difference
in linguistic features for the [91,100]range (Table 1), hinting
at the length constraint as the cause for the other features.
As for success, we observe that [91,100]tweets posted
before the change are 2.0% (relative) more likely to be
retweeted (Table 2), a much smaller difference than for
[136,140]tweets (6.0% relative), which we interpret as sup-
port for our above finding that length constraints slightly in-
crease tweet success. However, more studies are necessary
to confirm these findings, ideally with more data.
4 Discussion
In this paper, we address the question of how users adapt
their contributions to fit the limits imposed by social media
platforms and whether these restrictions tend to push users
towards producing more or less successful content. We de-
velop a matched observational methodology that involves
comparing carefully selected sets of tweets posted just be-
fore vs. just after the switch.
We find that tweets constrained by a 140-character limit
contain fewer hashtags and definite articles, and more words
in abridged form: more abbreviations, more contracted aux-
iliary verbs, more &, and fewer and, implying that, when
subject to a length constraint, users write more tersely.
We also found initial evidence that length-constrained
tweets are slightly more successful in terms of the engage-
ment they receive from other users. However, future work
needs to develop a better understanding of this phenomenon
and determine whether the observed findings are robust. We
anticipate that collecting more data from the time after the
relaxation of the constraint will be helpful to this end.
In our studies, we assume that tweets in the [136,140]
range posted after the switch are not constrained. We cannot
rule out the existence of users adhering to the old constraint
despite the switch (e.g., out of sheer nostalgia, or because
they are bots generating tweets via scripts written before the
switch), but in the absence of such hypothetical users, the
amplitude of the observed effects would be even larger.
Future work. This paper presents the first in what we hope
will be a series of studies regarding the impact of Twitter’s
switch from 140 to 280 characters. In particular, our obser-
vational studies should be followed by experimental studies,
in order to corroborate the initial findings presented here.
Also, our results are from right around the switch, when
things were potentially still in flux. It would be interesting to
revisit our findings in the future once the Twitter ecosystem
has settled into a new steady state. Finally, we hope that fu-
ture work will generalize our findings to constrained content
production beyond Twitter.
Abdi, H. 2007. Bonferroni and ˇ
ak corrections for multiple
comparisons. In Encyclopedia of Measurement and Statistics,
103–107. Sage Publications.
Beal, Vangie. 2016. Twitter dictionary: A guide to understanding
Twitter lingo.
Fritscher, B., and Pigneur, Y. 2009. Supporting business model
modelling: A compromise between creativity and constraints. In
Proc. International Workshop on Task Models and Diagrams.
Hennessey, B. A. 1989. The effect of extrinsic constraints on
children’s creativity while using a computer. Creativity Research
Journal 2(3):151–168.
Joyce, C. K. 2009. The Blank Page: Effects of Constraint on
Creativity. PhD thesis, UC Berkeley.
Lu, X. 2012. The relationship of lexical richness to the quality of
ESL learners’ oral narratives. The Modern Language Journal
McPhee, J. 2015. Omission: Choosing what to leave out. The
New Yorker. September 14, 2015.
Moreau, C. P., and Dahl, D. W. 2005. Designing the solution: The
impact of constraints on consumers’ creativity. Journal of
Consumer Research 32(1):13–22.
Rosen, A., and Ihara, I. 2017. Giving you more characters to
express yourself.
Rosen, A. 2017. Tweeting made easier.
Sysomos Retweets Study. 2010. Replies and retweets on Twitter.
A Appendix
The first four pages of this paper were published in the
proceedings of ICWSM 2018. In this appendix, which did
not appear in the proceedings, we provide additional de-
tails about the implementation of our observational study
and make some additional remarks.
A.1 Implementation details
Controlling for users and topics. In the main studies, we
build pairs of tweets posted by the same user or containing
the same set of hashtags. Note that a single user might post
multiple such pairs of tweets. When matching, we choose
the tweets such that the time distance from the switch is min-
imized: before the change, we keep the most recent tweet,
and after the change, we keep the least recent tweet. As a
result, selected pairs of tweets are tightly distributed around
the switch. Three quarters of matched tweets posted before
the switch are posted after October 9, and the median is Oc-
tober 26. In the case of tweets posted after the switch, three
quarters are posted before December 9, and the median is
November 23.
Constraint consistency over time. Twitter has been gradu-
ally distancing itself from the original hard limit of 140 char-
acters. First, they stopped counting media (including URLs
and quoted tweets) and @ replies toward the character count,
with these becoming metadata instead. However, after relax-
ing the character limit, such content is now counted again.
Therefore, in order to study the impact of the exact identical
constraint during all time periods, we exclude tweets con-
taining URLs, quoted tweets, and @ replies altogether when
sampling tweets for our studies. In order to rule out any fur-
ther potential variations of the counting policy over time,
and since it is impossible to determine the version of the ap-
plication that was used to send a given tweet, we moreover
consider only tweets that were posted using the Twitter web
client. Furthermore, since the character limit is not equal for
all languages (Rosen 2017) and in order to maintain a mean-
ingful interpretation of lexical features, we limit ourselves
to tweets in English.
Pagination. In our analysis, we detect and disregard pagi-
nated tweets in the following way. From the gathered data,
we detect 1,722 instances of pagination by detecting con-
secutive tweets posted by the same user containing a label
in form of position of a tweet in a sequence/total number of
tweets in a sequence,e.g.,1/3, 2/3, 3/3. We infer that pagina-
tion is relatively rarely used, as less than 0.1% of all tweets
in our dataset are paginated. Furthermore, 90% of subse-
quent paginated tweets are posted within four minutes of
each other. In the subsequent analysis, we disregard tweets
the same user posts within this time frame, but we observe
that varying the time frame, and thus the percentage of de-
tected paginations, does not qualitatively affect our findings.
A.2 Note on a user interface update
Alongside the change of the character limit, Twitter updated
its user interface such that it no longer counts down the num-
ber of characters as the user types. Instead, a circle fills in as
the limit is approached. As a consequence, the user is not
aware of the exact number of characters left until only 20
(out of the total of 280) remain. As this change was made
contemporaneously with the switch from 140 to 280 char-
acters, we cannot rule out the possibility that our results are
affected by this user interface change. However, regardless
of whether users see an exact character count or a half-filled
circle, it would likely be clear that tweets in the [136,140]
range that we study are not near the 280-character limit af-
ter the switch. Moreover, by being shown a circle instead of
an exact character count, users after the switch cannot real-
ize how close exactly they are to the deprecated 140-charac-
ter limit and whether they have passed it, which effectively
rules out optimization for the old 140-character limit after
the switch due to nostalgia or hipsterdom. For the above rea-
sons, we argue it is reasonable to assume that the change
from displaying an exact count to displaying a circle has a
negligible affect on our results.
A.3 Final remark about engagement
We also make an interesting observation with respect to the
relation between tweet length and other users’ engagement
with a tweet: Fig. 2 shows that the success of tweets is
strongly correlated with their length, both before and af-
ter the switch, but also that tweets of a length up to 140
characters posted after the switch are vastly less popular
than tweets of the same length posted before the switch. In-
deed, it seems that the total amount of engagement caused
by the entirety of all tweets stays constant before vs. af-
ter the switch—it is merely distributed over a wider range
of lengths after the switch. It is interesting to see Twitter’s
switch in this light, given that it was officially motivated by
the claim that tweets longer than 140 characters lead to more
engagement (Rosen 2017).
... (g) High character count: A higher character count positively influenced the reactions on both platforms, which was confirmed by a study by Gligorić et al. (2018) [50] for Twitter posts. A possible reason could be a preference of subscribers for detailed, informative posts. ...
... (g) High character count: A higher character count positively influenced the reactions on both platforms, which was confirmed by a study by Gligorić et al. (2018) [50] for Twitter posts. A possible reason could be a preference of subscribers for detailed, informative posts. ...
... Löffler (2014) [52] also recommended that organizations from the marketing sector publish longer Facebook posts for more shares. According to Vries et al. (2012), the same applies to Twitter, regardless of topic and user [50]. ...
Full-text available
Social networks expand the communication tools of nature conservation. Nonetheless, to date there is hardly any scientific literature on nature conservation communication in social networks. For this reason, this paper examines 600 Facebook and Twitter posts of three German nature conservation organizations: Federal Agency for the Conservation of Nature (Bundesamt für Naturschutz, BfN), Naturschutzbund Deutschland e. V. (NABU), and World Wide Fund for Nature (WWF) Germany. Using the Mann–Whitney U method and Spearman’s rank correlation analysis, it reveals how post design affects communication success and provides respective recommendations for German conservation organizations. Communication success was divided into four indicators: reactions, comments, shares, and overall engagement as a synthesis of the three. On Facebook, the use of hashtags, images, and many characters (up to 1500) leads to higher success, whereas emojis and videos can reduce it. On Twitter, links, images, and longer posts promote user interactions. Emojis have a positive influence on comments and overall engagement, but a negative influence on reactions and shares. In addition, hashtags reduce overall engagement on Twitter. These results are discussed with reference to similar studies from other political fields in order to provide recommendations for conservation organizations. A validation and expansion of the presented results is recommended due to the growing relevance of digital nature conservation communication.
... Second, the platforms through which information travels have particular designs which may affect the fidelity of the information. In other words, platforms and their design constraints matter (Malik and Pfeffer 2016;Gligorić, Anderson, and West 2018). For example, Malik and Pfeffer (2016) use temporal data from Facebook and Netflix to give proofof-concept to the notion of platform effects, demonstrating how use differs on both platforms before and after a significant design change that alters user exposure to information. ...
Full-text available
The public interest in accurate scientific communication, underscored by recent public health crises, highlights how content often loses critical pieces of information as it spreads online. However, multi-platform analyses of this phenomenon remain limited due to challenges in data collection. Collecting mentions of research tracked by Altmetric LLC, we examine information retention in the over 4 million online posts referencing 9,765 of the most-mentioned scientific articles across blog sites, Facebook, news sites, Twitter, and Wikipedia. To do so, we present a burst-based framework for examining online discussions about science over time and across different platforms. To measure information retention we develop a keyword-based computational measure comparing an online post to the scientific article's abstract. We evaluate our measure using ground truth data labeled by within field experts. We highlight three main findings: first, we find a strong tendency towards low levels of information retention, following a distinct trajectory of loss except when bursts of attention begin in social media. Second, platforms show significant differences in information retention. Third, sequences involving more platforms tend to be associated with higher information retention. These findings highlight a strong tendency towards information loss over time - posing a critical concern for researchers, policymakers, and citizens alike - but suggest that multi-platform discussions may improve information retention overall.
The internet and development of information technologies brought about the emergence of digital communication tools. In this vein, social media have become a phenomenon in terms of creating informative, interactive, and participatory platforms for the individuals. The social media tools have become prominent not only for public relations or communications experts, but also for politicians, scholars, groups, brands, organizations, etc. One of the effective social media tools is Twitter, which has been focus of political communication research due to its tendency of creating discussion platform that allows the users to involve in and interact with each other. This study focused on how Twitter creates the two-way interaction for the users and what the main components of this interaction are. In addition, the contribution of Twitter to organizational promotion was also another concern of the study. In this context, the research focuses on both inter-organizational and individual levels. Multiple case study technique was used as research technique. Five different cases were analyzed.
Full-text available
Social media data have been widely used to gain insight into human mobility and activity patterns. Despite their abundance, social media data come with various data biases, such as user selection bias. In addition, a change in the Twitter app functionality may further affect the type of information shared through tweets and hence influence conclusions drawn from the analysis of such data. This study analyzes the effect of three Twitter app policy changes in 2015, 2017, and 2019 on the tweeting behavior of users, using part of London as the study area. The policy changes reviewed relate to a function allowing to attach exact coordinates to tweets by default (2015), the maximum allowable length of tweet posts (2017), and the limitation of sharing exact coordinates to the Twitter photo app (2019). The change in spatial aspects of users’ tweeting behavior caused by changes in user policy and Twitter app functionality, respectively, is quantified through measurement and comparison of six aspects of tweeting behavior between one month before and one month after the respective policy changes, which are: proportion of tweets with exact coordinates, tweet length, the number of placename mentions in tweet text and hashtags per tweet, the proportion of tweets with images among tweets with exact coordinates, and radius of gyration of tweeting locations. The results show, among others, that policy changes in 2015 and 2019 led users to post a smaller proportion of tweets with exact coordinates and that doubling the limit of allowable characters as part of the 2017 policy change increased the number of place names mentioned in tweets. The findings suggest that policy changes lead to a change in user contribution behavior and, in consequence, in the spatial information that can be extracted from tweets. The systematic change in user contribution behavior associated with policy changes should be specifically taken into consideration if jointly analyzing tweets from periods before and after such a policy change.
Connecting the affordance framework in computer-mediated communication to public relations theories, this essay proposes an affordance perspective on dialogic communication and digital public relations in general. We argue that 1) the enactment of organization-public dialogue on digital platforms requires certain combinations of media affordances; 2) the lens of affordances facilitates a non-dichotomous examination of the “dialogic communication vs. digital media” debate; 3) the fifth dialogic principle “ease of interface” should be conceptually expanded to “favorable affordances,” which asserts that public relations practice should evaluate digital media platforms’ various action possibilities and consider their inherent potentials for organization-public relationship building; and 4) research on digital public relations should incorporate affordance theory to achieve cross-platform theorization.
Due to the worldwide accessibility to the Internet along with the continuous advances in mobile technologies, physical and digital worlds have become completely blended, and the proliferation of social media platforms has taken a leading role over this evolution. In this paper, we undertake a thorough analysis towards better visualising and understanding the factors that characterise and differentiate social media users affected by mental disorders. We perform different experiments studying multiple dimensions of language, including vocabulary uniqueness, word usage, linguistic style, psychometric attributes, emotions’ co-occurrence patterns, and online behavioural traits, including social engagement and posting trends. Our findings reveal significant differences on the use of function words, such as adverbs and verb tense, and topic-specific vocabulary, such as biological processes. As for emotional expression, we observe that affected users tend to share emotions more regularly than control individuals on average. Overall, the monthly posting variance of the affected groups is higher than the control groups. Moreover, we found evidence suggesting that language use on micro-blogging platforms is less distinguishable for users who have a mental disorder than other less restrictive platforms. In particular, we observe on Twitter less quantifiable differences between affected and control groups compared to Reddit.
Full-text available
Social network services such as Twitter are important venues that can be used as rich data sources to mine public opinions about various topics. In this study, we used Twitter to collect data on one of the most growing theories in education, namely Self-Regulated Learning (SRL) and carry out further analysis to investigate What Twitter says about SRL? This work uses three main analysis methods, descriptive, topic modeling, and geocoding analysis. The searched and collected dataset consists of a large volume of relevant SRL tweets equal to 54,070 tweets between 2011 and 2021. The descriptive analysis uncovers a growing discussion on SRL on Twitter from 2011 till 2018 and then markedly decreased till the collection day. For topic modeling, the text mining technique of Latent Dirichlet allocation (LDA) was applied and revealed insights on computationally processed topics. Finally, the geocoding analysis uncovers a diverse community from all over the world, yet a higher density representation of users from the Global North was identified. Further implications are discussed in the paper.
This paper unpacks how dynamic political and media systems shape the kinds of frames political actors champion, when and how they express support for frames and the implications of both for individual claimsmaking. To do so, we conduct a rigorous qualitative analysis of discourse during a two-week period in which the Florida legislature considered and passed the Marjory Stoneman Douglas High School Public Safety Act after a shooter killed 17 people in Parkland, Florida. We systematically explore how two framing dynamics – competition and amplification – shape what frames political actors champion and the relative effects of these dynamics on individual claimsmaking in 438 letters to the editor and op-eds appearing in mainstream outlets, 4,962 emails sent to Florida Governor Rick Scott, and 1,000 tweets. We find that amplification and competition shape the relative visibility of frames and the frequency with which individuals use these frames in their claimsmaking. Generally speaking, gun control and progressive groups selectively amplified frames associated with the emerging, student-led Never Again Marjory Stoneman Douglas movement and legislative frames that were consistent with their goals. This seems to have increased the visibility of these ideas in mainstream outlets and influenced claimsmaking insofar as individuals drew on amplified frames across the forums relatively frequently. This was not true of frames opposing gun control. Gun rights groups bickered with politicians and among themselves. As a result, gun rights frames were less prevalent in mainstream discourse and in individual claimsmaking.
Conference Paper
Full-text available
Diagrams and tools help to support task modelling in engineering and process management. Unfortunately they are unfit to help in a business context at a strategic level, because of the flexibility needed for creative thinking and user friendly interactions. We propose a tool which bridges the gap between freedom of actions, encouraging creativity, and constraints, allowing validation and advanced features.
Full-text available
Across a variety of domains, consumers often choose to act as the designer of their own solution, sourcing the necessary components and assembling the parts to meet their specific goals. While thinking creatively is an integral part in the daily life of every consumer, surprisingly little research in marketing has examined the factors influencing such processes. In our research, we examine how input and time constraints influence the way in which consumers process information during a creative task and how those processes, in turn, influence the creativity of the solution. Paradoxically, we find that input constraints encourage more creative processing, provided the individual is not under significant time constraints. (c) 2005 by JOURNAL OF CONSUMER RESEARCH, Inc..
This study was an examination of the relationship of lexical richness to the quality of English as a second language (ESL) learners' oral narratives. A computational system was designed to automate the measurement of 3 dimensions of lexical richness, that is, lexical density, sophistication, and variation, using 25 different metrics proposed in the language acquisition literature. This system was used to analyze large-scale data from the Spoken English Corpus of Chinese Learners (Wen, Wang, & Liang, 2005) together with the vocd utility of the Computerized Language Analysis programs (MacWhinney, 2000), which offers an additional measure of lexical variation, the D measure (Malvern, Richards, Chipere, & Durán, 2004; McKee, Malvern, & Richards, 2000). This comprehensive analysis allowed us to identify measures that correlate strongly with the raters' judgments of the quality of ESL learners' oral narratives, as well as to understand the relationships among these measures. This research provides ESL teachers and researchers with a robust tool for assessing the lexical richness of ESL language samples and insights into how lexical richness measures may be effectively used as indices of the quality of ESL learners' speaking task performance.
This dissertation is about how constraint—restrictions to freedom that limit and direct search—influences creativity. Freedom is often associated with creativity, yet recent work in the decision making literature suggests that too much freedom can be paralyzing when it provides too many choices. This dissertation examines how the extent of constraint imposed on a task, when conceptualized as a continuum, affects creative processes and outcomes. It employs a multi-method, multi-level approach through three studies. Study 1 was a controlled laboratory experiment centered around a written product design task where constraint was manipulated by varying task instructions. A curvilinear effect of constraint on creativity was identified such that a moderate degree of constraint was more conducive to creativity than either a high or a low degree. These effects were not explained by alternative explanations such as time allocation during the task, or decreased intrinsic motivation. Studies 2 and 3 examined the role of constraint in 43 new product development teams. Through quantitative analysis, Study 2 found that the degree of constraint that new product development teams voluntarily imposed on their projects at the beginning of the semester predicted the creativity of their product proposals more than ten weeks later. The results held up even when controlling for task conflict. Study 3 examined the same 43 teams through a series of three multi-method case studies. Grounded-theory analysis gave qualitative support to the theory proposed in Chapter 2 and revealed several emergent themes that were not anticipated, namely: assumption-constrained creativity, uncovering latent conflict, and confirmation-constrained creativity. The study resulted in new predictions about how constraint affects creative teams, and a novel framework for conceptualizing creativity as a hypothesis-testing activity. These findings suggest that while some amount of choice is important for encouraging creativity, too much can be counterproductive, which runs counter to many popular theories of creativity. This dissertation should provide encouragement to organizations that are institutionally embedded, have scarce resources, or are otherwise restricted.
Two studies were conducted to examine the impact of extrinsic constraints on children's performance while using a computer. Open‐ended tasks were presented and possible differential effects of rewards and evaluations imposed by human (experimenter) and non‐human (computer) sources were explored in relation to product creativity, motivation, and affective response. Both studies provide support for the hypothesis that reward and evaluation will undermine creativity and motivation, even when a computer serves as the source of these constraints. In addition, a developmental trend emerged with older children being far more adversely affected than their younger counterparts. The theoretical and applied significance of these results are discussed.
Supporting business model modelling: A compromise between creativity and constraints
  • Vangie Beal
  • B Fritscher
, Vangie 2016] Beal, Vangie. 2016. Twitter dictionary: A guide to understanding Twitter lingo. [Fritscher and Pigneur 2009] Fritscher, B., and Pigneur, Y. 2009. Supporting business model modelling: A compromise between creativity and constraints. In TAMODIA.
Omission: Choosing what to leave out. The New Yorker
  • J Mcphee
McPhee, J. 2015. Omission: Choosing what to leave out. The New Yorker. September 14, 2015.
Giving you more characters to express yourself
  • A Rosen
  • I Ihara
and Ihara 2017] Rosen, A., and Ihara, I. 2017. Giving you more characters to express yourself. [Rosen 2017] Rosen, A. 2017. Tweeting made easier. [Sysomos Retweets Study 2010] Sysomos Retweets Study. 2010. Replies and retweets on Twitter.
Twitter dictionary: A guide to understanding Twitter lingo
  • Vangie Beal
Beal, Vangie. 2016. Twitter dictionary: A guide to understanding Twitter lingo.