CC BY-NC-ND
0CrossMark
Available online at www.sciencedirect.com
ScienceDirect
Procedia Manufacturing 3 (2015) 5496 - 5503
6th International Conference on Applied Human Factors and Ergonomics (AHFE 2015) and the
Affiliated Conferences, AHFE 2015
Designing for retweets - a study on Twitter interface design focusing on retweetability
Yu-Hsiu Hunga,*7 Der-Shuan Hwua, Caroline Arkensonb, Yi-Chin Leea
a Department of Industrial Design, National Cheng Kung University, No.1, University Road, Tainan City 701, Taiwan (R.O.C.) bKTH Royal Institute of Technology, SE-100 44 Stockholm Sweden
Abstract
500 million messages - tweets - are being sent daily by 255 million monthly active Twitter users, which makes Twitter a great platform for sharing information. One way to share information on Twitter is by using the retweet function,which allows a user to share an exact copy of a tweet for his or her own followers to see. Twitter users retweet to spread information to a new audience or show one's interest in something. As online communication is important, we wanted to investigate a tweet's retweetability and see how redesigning the way a tweet is being presented could improve retweeting. We created four example tweets in both English and Chinese versions to see how the age of a tweet and its number of retweets affect its retweetability. Through a survey, we let Swedish and Taiwanese students rate the tweets based on the likeliness that they would retweet it. Overall, the retweetability of our tweets were low as they mostly were being rated with "very unlikely" and "unlikely" to be retweeted by the participants. We found that age of a tweet and number of retweets are not important factors when it comes to retweetability, but that the content of a tweet and an included link do matter. Based on our findings, we present a design proposal on how to preview links to easily let Twitter users find out the content of a link.
© 2015TheAuthors.PublishedbyElsevierB.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.Org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of AHFE Conference
Keywords: Retweet-ability; Interface design; Twitter; Social web
* Corresponding author. Tel.: +886-6-2757575; fax: +886-6-274-6088. E-mail address: idhfhung@mail.ncku.edu.tw
2351-9789 © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.Org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of AHFE Conference
doi: 10.1016/j.promfg.2015.07.699
1. Introduction
Twitter is a social media platform through which users can share "an expression of a moment or idea" to its followers in a message containing of maximum 140 characters. The message, a so-called tweet, can contain text, pictures, videos or links. Users can reply to, retweet or favourite an existing tweet and use a hashtag when composing a tweet in order to assign it a topic. [1] 500 million messages - tweets - are being sent daily by 255 million monthly active Twitter users, which makes Twitter a great platform for sharing information. One way to share information on Twitter is by using the retweet function, which allows a user to share an exact copy of a tweet for his or her own followers to see.[2, 3] ° Why do Twitter users choose to retweet? Amongst the reasons is the will to spread tweets to a new audience, to show one's role as a listener, to agree with someone or validate the thoughts of others .As an act of friendship retweet could promote loyalty and trust between friends[4]. In addition to personal factors, previous studies have shown that retweet a positive advantage in the multi > Toriumi et al. (2013) revealed retweet could relieve anxiety of individual. [5] Retweet was used as a communication or dissemination of information pipeline [6] ° Online communication is a big part of the society we currently live in and expressing oneself and taking part in conversations happen both online and offline - making online communication an important aspect to improve.
As retweeting is a way for users to take part in online conversations it is important to facilitate this act, to allow the user to easily fulfil their motivation behind retweeting. In a previous research on retweets in 2010 it was found that the previous number of retweets did not affect the tweet's retweetability. This was, however, just when the direct retweet function had been introduced and most of the retweets were made as new tweets using a syntax similar to "RT @user copy of message".[7] It would be interesting to see if things have changed since then as there has been a shift in the act of retweeting; the direct retweet function is the one most commonly used today.
When it comes to age as a variable, age of a Twitter account has been studied [4], but not age of the actual tweet. Sometimes online content go viral quickly, and sometimes it goes viral long after it was published online. It has, however, been shown that the probability that a user will retweet something is higher if the tweet appears higher up in a user's Twitter feed, [1] but that does not necessarily mean that an older tweet has lower retweet probability as the tweets in the beginning of the feed might be retweets of old tweets.
Much of the previous research on retweetability has focused on predicting it and trying to understand how different components of a tweet and the Twitter network affect it, but discussions on the interface design are scarce and is therefore another motivation behind this research. The aim of this study is examining number of retweet and age of tweet on twitter interface design whether it would affect users retweet or not.
2. Method
2.1. Participants
In this study, the participants were 89 undergraduate students from Sweden and Taiwan. Purposive sampling method was conducted to select valid samples who are required to have at least one-year experience of using Twitter. As a result, 18 out of 37 Taiwanese and 17 out of 52 Swedes were qualified in the experiment. The dropout rates of Taiwanese and Swede were 51.35% and 67.30% individually. All participants were recruited in the experiment voluntarily with no compensation.
2.2. Experimental design
Independent variables were: (1) the number of retweets and (2) the age of tweet; dependent variable was the intention of retweet inclination (i.e. retweetability). Simulated tweets were designed based on Twitter version 5.13.1 for Android as the testing material used in the experiment.
The design procedures of the experiment is shown as follows. First, set the number of retweets and the age of tweet. The former was set to 29times and 3266times; the latter was set to 2hours ago and March 11th. These settings were random with no purpose. Second, design the content of tweet. In this experiment, the content was chosen from top lists of twitter, and a recommendation of tourist attractions link of trip-advisor was adopted as the testing content
of simulated tweet. We arranged number of retweet and age of tweet with the permutation combination to four simulated tweet. Since the participants included Taiwanese and Swedish students, two versions of simulated tweet were made. One was Chinese version for Taiwanese students; the other was English version for Swedish students. Figure1 is one example of simulated tweet in two versions.
Table 1. Independent Variables in Experiment
#of Retweet
27times
3266times
Age of tweet
Taiwanese( 18)/Swedish(17) students
Taiwanese( 18)/Swedish(17) students
Taiwanese(18)/Swedish(17) students
Taiwanese( 18)/Swedish(17) students
2.3. Procedure
An online questionnaire was created to collect data and test the effect of the different variables. The link of online questionnaire was sent to Facebook time line and instant messaging individually. On the first page of questionnaire showed the introduction and basic information of the study including the aim of the research and the essential constraints of participants. In addition, participants were required to answer their inclination of participating in the experiment and the frequency of using Twitter. Participants were not told which variable was being tested at the moment but asked at each tweet to comment their choice. Likert scale in 5 grade from "very unlikely " to "very likely was used as evaluation. In the end of the questionnaire, participants were asked if they had noticed the difference between the tweets and what the difference was. Figure2 and figure3 show that first page of questionnaire and one example of page in two version. Finally, we used SPSS16.0 to analyse the relationship between variables.
Fig. 1. Left: One of example for English with tweet age 2hours ago and retweet# 27times. / Right: One of example for Chinese with tweet age March 11th and retweet# 3266times.
qualtrics
qualtrics
m-wreabwét«!»*»»«;» • &m«umtmmix • g^mirns«! • hiikii «SWS4? t»» • Ïiéi-Bliw • WW*« ■
■ an*jM№&B№»MM?
£S ■ • IMUM-mJtTwmertilBfiA ■ ■
(fc»ij;**ÎËÏfll!UTwiiteit|«t-*M±*l ?
nus research alms to examine me retweetability of a iweet You will be presented with tour different tweets all slighlly diflerenl. and your task is to rata how tikety it is that you would retwaet it as wall as adding a comment to why
Your answers are anonymous and your participation win be most appreciated!
Please note If you are not a student in Sweden with a Twitter account older than a year, then you are outside our target group
So, to gel started Are you a student in Sweden with a Twitter account older Dian a year? o Yes o No
№ftl8Twittei«r>*№»?
<■■ iMss-as-*
< -«J!-*
< -«U32-W;
C -«2-3»
How olten are you on T wider? c Never
o Less than Once a Month o Once a Month v 2-3 Times a Monji o Once a Week cr 2-3 Times a Week Daily
Fig. 2. First page of Online questionnaire Left: Chinese Version/ Right: English version.
Fig. 3. One of example for Online questionnaire. Left: Chinese version/Right: English Version.
SWEDISH
I Í 1 r\J 1 o 1
DAILY 2-3TI ME A WEEK ONCEA 2 - 3 T1 M E A ONCEA WEEK MONTH MONTH LESS THANONCE A MONTH NEVER
TAIWANESE
Í i 1 Í m o
DAILY 2-3TI ME A WEEK ONCE A 2-3TIME A ONCEA WEEK MONTH MONTH LESS THANONCE A MONTH NEVER
Fig. 4.Top: Frequency of Twitter usage amongst the Swedish participants. Bottom: Frequency of Twitter usage amongst the Taiwanese participants.
3. Result and discussion
3.1. Twitter usage
In the survey we asked the participants how frequently they use Twitter and we found that the Swedish participants in general user twitter more often than the Taiwanese participants. Fig4. Shows the result.
3.2. Descriptive statistics and Open ended question
The result of description statistic appears that participants gave low ratings about retweetability (Mean=2.00,SD=1.12). Results are shown on table2. Independently, analysing the retweetability rating of Taiwanese students and Swedish students. Although results were still appearing the low rating. However, Taiwanese students' rating of each question is higher than Swedish students. Further to analyse the answer of each open ended question in the end of each page. The results were shown that higher retweetability rating derived from Taiwanese students. For instance: "This content is very relaxed and positive so I think it can retweet to other one" (From Q3, rating: 4) and "I like traveling" (From Q2, rating:4).
Table 2. Descriptive Statistic: mean and stander deviation of retweet intention rating.
Question# N Mean( Std. Deviation)
Q1:27times/2HR 18 2.50(0.98)
Q2: 3266time/mar11th 18 2.66(1.32)
Taiwan Q3: 3266time/2HR 18 2.55(1.29)
Q4:27time/mar11th 18 2.11(1.07)
Total 72 2.45(1.17)
Q1:27times/2HR 17 1.64(1.05)
Q2: 3266time/mar11th 17 1.64(1.05)
Swedish Q3: 3266time/2HR 17 1.41(0.61)
Q4:27time/mar11th 17 1.41(0.61)
Total 68 1.52(0.85)
Q1:27times/2HR 35 2.08(1.09)
Q2: 3266time/mar11th 35 2.17(1.29)
All Q3: 3266time/2HR 35 2.00(1.16)
Q4:27time/mar11th 35 1.77(0.94)
Total 140 2.00(1.12)
In 140 answers of open-ended questions, most of the answers were negative answer. These answers can be divided into four categories: (1) The content of simulated tweet was regarded as ad, e.g. " Click bait", "it is ad" or "this content looks like an advertisement". (2) Showing no interest in the content of simulated tweet, e.g."Not interested" or "I'm not interested in advertising for companies". (3) The content of simulated tweet was no value to my follower or myself, e.g. "Don't care for the content" or " It doesn't give me or my followers anything. I might click on the link but I wouldn't retweet it". (4) Did not noticed the change of experiment, e.g. "Same image" , "The same as the rest" or "um, I just got this exact question..." . In addition, some participants mentioned that they wanted to check the URL content before deciding retweet or not. Most of the answers were the second or the third category. The fourth category echoed the problem that mentioned in section 2.1-the high dropout rate. Therefore incomplete questionnaire was checked again. Findings showed that the open ended answers of incomplete questionnaire was same as the fourth category.
3.3. Two-wayANOVA
Since the questionnaires was completed by Taiwanese students and Swedish students. Nationality would be considered as another factor. The four combination of number of retweet and age of tweet were considered as one factor with four levels. A two way ANOVA was conducted with SPSS16.0. Table3. Shows the result.
Table 3. Statistical output for the two-way ANOVA
Source Type III Sum of Squares df Mean Square F value P value
Nationality 23.904 1 23.904 16.761 .000
Question# 1.526 3 .509 .357 .784
Nationality * Question# .694 3 .231 .162 .922
Total 741.000 140
There was no statistical significant except nationality. According to the answer of open ended question, number of retweet and age of tweet were less important than the content of tweet, for most of the participants. In other word, content of tweet had more influence to retweetability than other two factors. Moreover the end question of questionnaire (Did you notice the differences of the tweets?)was analyzed. Only 16 of the participants (eight Taiwanese, eight Swedish) noticed the difference among the questionnaire. Reasonable assumption that the reason of no statistical significant not only the content of tweet had greater influence than number of retweet and age of tweet, but participants did not notice the change of experiment has to be taken into consideration. As for there are statistical significant in nationality. The main reason is Taiwanese students had higher retweetablilty rating than Swedish student. Although Swedish students are in general user twitter more often than the Taiwanese students.
3.4. Re-design
The above are a variety of reasons, although the number of retweet and age of tweet were not important factors. However, some participants said they want to check the URL content before decided whether retweet or not. According to the answer open-ended questions, our study recommend that twitter could add a URL preview rather was only a snapshot. Thereby providing more information to users, to increase retweetability. The design simulation shows on Fig5.
Fig. 5. Design simulation-the preview of link in tweet.
4. Conclusion
Through the research we found that age and number of previous retweets were not important factors when it comes to the intention of retweet inclination. What was important, was the author and content of the tweet, and if the user's followers would find it interesting. Retweeting seems to be more complex than just some of the data of a tweet as things like the user's values and Twitter personality might influence the choice to retweet.
Our design proposal is to based on the will to make it easier for the user to preview the content of the link, as the content of the link was stated to affect the decision to retweet or not.
Future research on intention of retweet inclination should look into other affecting factors. As the content of the tweet affected our results a lot, this is something that should be kept in consideration to try minimize its effect on future research.
Reference
[1].Murthy, D., Twitter: Social communication in the Twitter age. 2013: John Wiley & Sons.
[2].Twitter, I. 2014; Available from: https://about.twitter.com/company.
[3].Twitter, I. 2014 Available from: https://about.twitter.com/what-is-twitter/story-of-a-tweet.
[4].Boyd, D., S. Golder, and G. Lotan. Tweet, tweet, retweet: Conversational aspects of retweeting on twitter. in System Sciences (HICSS), 2010 43rd Hawaii International Conference on. 2010. IEEE.
[5].Toriumi, F., et al. Information sharing on twitter during the 2011 catastrophic earthquake. in Proceedings of the 22nd international conference on World Wide Web companion. 2013. International World Wide Web Conferences Steering Committee.
[6].Oh, O., K.H. Kwon, and H.R. Rao. An Exploration of Social Media in Extreme Events: Rumor Theory and Twitter during the Haiti Earthquake 2010. in ICIS. 2010.
[7].Suh, B., et al. Want to be retweeted? large scale analytics on factors impacting retweet in twitter network. in Social computing (socialcom), 2010 ieee second international conference on. 2010. IEEE.
