284
Int. J. Web Based Communities, Vol. 6, No. 3, 2010
A cross-cultural analysis of Flickr users from Peru, Israel, Iran, Taiwan and the UK Amir Dotan and Panayiotis Zaphiris* Centre for HCI Design, School of Informatics, City University London, Northampton Square, EC1V 0HB London E-mail: sbbc739@soi.city.ac.uk E-mail: panayiotis.zaphiris@cut.ac.cy *Corresponding author Abstract: In this paper, we describe a study exploring cultural differences in user-generated content websites using Flickr, a popular social photo-sharing site, as a case study. The increased popularity of socially driven websites has created new challenges with regard to cross-cultural system design and localisation. This research aimed to identify some of these challenges and help bring cross-cultural usability studies up-to-date with the latest trends on the internet. It also sought to explore the usefulness of Geert Hofstede’s popular cultural model for studies of this sort. Findings reveal differences as well as similarities regarding tagging patterns and use of language to annotate content between the five chosen national cultures. These insights could inform the future localisation and internationalisation of user-generated content driven sites like Flickr and You Tube. Keywords: culture; localisation; cross-cultural design; CMC; Web 2.0; UGC; Peru; Israel; Iran; Taiwan; UK. Reference to this paper should be made as follows: Dotan, A. and Zaphiris, P. (2010) ‘A cross-cultural analysis of Flickr users from Peru, Israel, Iran, Taiwan and the UK’, Int. J. Web Based Communities, Vol. 6, No. 3, pp.284–302. Biographical notes: Amir Dotan is a Researcher at the Centre for HumanComputer Interaction Design, School of Informatics of City University London. Prior to joining City University, he worked as a Lecturer, teaching web development, interface design and HCI as part of an undergraduate multimedia arts degree programme. His research interests lie in HCI with an emphasis on interaction design, social and cultural aspects. Panayiotis Zaphiris is an Associate Professor in the Department of Multimedia and Graphic Arts of Cyprus University of Technology. Panayiotis has a PhD in Human-Computer Interaction from Wayne State University, USA. He also has an MSc in Systems Engineering and a BSc in Electrical Engineering both from University of Maryland, College Park, USA. He has worked for a number of years at the Centre for HCI Design of City University London where he reached the rank of Reader in HCI. His research areas are in the area of Human Computer Interaction, Social Computing and Inclusive Design with an emphasis on the design of interactive systems for people with disabilities and the elderly.
Copyright © 2010 Inderscience Enterprises Ltd.
A cross-cultural analysis of Flickr users
1
285
Introduction
Traditionally, cross-cultural usability studies examining online behaviour and user preferences focused on websites where users are seen as passive consumers of content. However, technological developments over the past five years, often referred to as ‘Web 2.0’, have significantly altered the web by transforming many internet users into active content sharers who use a plethora of social applications to form social networks and share content with groups of people and the world at large. We believe it is important to examine how cultural background might influence the use of social content-sharing applications in order to align cross-cultural usability studies with contemporary trends to develop suitable local and global designs. As popular sites like Flickr (www.flickr.com), YouTube (www.youtube.com) and Facebook (www.facebook.com) attract millions of people from all over the world, it becomes ever more important to explore how culture impacts tagging, public and private content sharing, motivation, social interactions, participation, interface customisation and other emerging HCI topics. Existing cross-cultural design literature, as well as recent and more Web 2.0 centric studies (e.g., Marcus, 2005; Marlow et al., 2006; Mathes, 2004), address some of these issues but there is still a great deal of scope for further research and insights, which we hope to contribute to with this research. The study we report in this paper investigated the impact of cultural background on the use of a user generated-content sites by focusing on Flickr, a photo sharing social network, as a case study. At the time of writing (January 2008), Flickr (2007) had over 20 million registered users and hosted more than 2 billion photos. In June 2009, the number of photos was reported to be 3.6 billion (Wolk, 2009). Like many other Web 2.0 sites, it is based on user generated content, social networking and user generated taxonomies i.e., tags, often referred to as Folksonomies. Geert Hofstede’s cultural model has been used extensively by HCI researchers in recent years in order to assess cultural differences. Given the richness of user generated content like Flickr, it was also decided to examine if the cultural scores developed by Hofstede could prove useful.
1.1 Aims and objective The study focused primarily on the utilisation of tags, use of language to annotate content i.e., English versus one’s native language, social interactions as well as motivation for uploading and sharing content. It aimed to extend existing knowledge of cross-cultural design and contribute contemporary perspective that would benefit practitioners and researchers focusing on user generated content driven websites. More specifically the objectives of the project were as follows: •
compare and contrast parameters between groups of users from five national cultures in order to determine the impact cultural orientation has on the use of Flickr, and user generated-content sites in general
•
apply Geert Hofstede’s cultural dimensions to those parameters and explore the degree to which they can explain differences and similarities between the chosen cultures
286
A. Dotan and P. Zaphiris
•
identify both Flickr-specific and general cross-cultural considerations that could inform the development of future and current participatory sites.
2
Literature review
2.1 Theories of cultural differences Culture has been studied extensively, resulting in numerous theories and models by the likes of Geert Hofstede (1991) and Edward T. Hall (1976). As individuals are members of different groups and categories, they carry several layers of mental programming within themselves that correspond to different levels of culture (Hofstede, 1991). Such levels for example are national, regional, religious, gender and social class. All, to some extent, impact people’s perceptions and attitudes. There is considerable literature describing cultural differences and cultural orientation. Gould (2006) lists eight cultural theories that could inform interface design decisions (e.g., Condon and Yousef, 1975; Gudykunst, 1986). Geert Hofstede’s (1991) research into cultures dimensions and cultural distinctions is often cited by those who explore culturally informed design issues, as it offers a general theoretical basis for understanding cultural differences. After surveying 116,000 IBM employees worldwide, Geert Hofstede came up with four cultural dimensions he used to classify countries on a scale of 1 to 100 (a fifth dimension, long- vs. short-term orientation, was later added). Aaron Marcus (2005) offers the following summary of each dimension: •
Individualism vs. collectivism – The extent to which people are integrated into tight social networks.
•
Power distance – The extent to which less powerful people expect and accept that power is distributed unequally.
•
Masculinity vs. femininity – The relative desirability of material success versus quality of life and of assertive versus modest behaviour.
•
Uncertainty avoidance – The extent to which people tolerate ambiguity and risk or feel threatened by change.
Despite various criticisms, Hofstede’s model has been widely adopted by the HCI community as it provides researchers practical data used in quantitative research.
2.2 Cultural differences in system design Cultural diversity has been studied extensively in the context of HCI, propelled primarily by the global spread of the internet since the mid 1990s. Since information and communication technologies originated primarily in North America, they reflected the values and perceptions of people from that part of the world and took into account their mental models (Sacher et al., 2001). Research conducted since the mid 1990s has shown that HCI issues such as icon recognition, information architecture, communication styles, use of imagery, use of colours, user preferences and priorities are all to some extent
A cross-cultural analysis of Flickr users
287
culture specific (Barber and Badre, 1998; Hogan, 2006; Pfeil et al., 2006; Sacher, et al., 2001; Syarief et al., 2003; Yunker, 2003).
2.3 Flickr At the time of writing, academic research focusing on the use of Flickr is scarce and still in its infancy. This is not at all surprising considering that social networks exploded on the scene only a few years ago (Boyd and Ellison, 2007). Despite limited quantity, Flickr related literature addresses a variety of areas. These include social networking (e.g., Boyd and Ellison, 2007), Web 2.0 and New Media (e.g., Burgess, 2006), tagging (e.g., Ames and Naaman, 2007), browsing and search e.g. (Hogan, 2006; Lerman and Jones, 2007), amateur photography e.g. (Burgess, 2004) and photo sharing e.g. (Van-House, 2007). Flickr is a highly social environment and its function and essence stretches far beyond mere photo sharing. As a Web 2.0 application, it converges social networking, online communities and user-generated content and could be more accurately defines as a ‘photo sharing community’ (Burgess et al., 2006) or ‘social photo sharing’ (Lerman and Jones, 2007). Photo content, which traditionally used to be regarded as personal, is shared with a global audience (Koman, 2005). People viewing the content may be intimates, acquaintances or strangers (Van House, 2007). This sort of asynchronous interaction through which individuals stay close and informed about each other is referred to by Van House (2007) as Distant Closeness.
2.4 Tagging Tags are user-generated metadata used to annotate content in order to organise it and also to allow others to come across it, either intentionally or accidentally. Indexing content is nothing new, however, Web 2.0 application such as Del.icio.us and Flickr popularised the notion of ‘social tagging’ as tags facilitate the formation of ad-hoc communities and photo pools when users annotate using identical tags (Ames and Naaman, 2007; Mathes, 2004). Sometimes, tags are used as private codes among a group of users to facilitate semi-private communication (Rafferty and Hidderley, 2007). Marlow et al. (2006) chose Flickr as a case study and analysed tag usage within the system, by exploring 58 million tags and also looking at the number of photos a user uploaded, the number of distinct tags they used and the number of contacts they had. They reported that tagging patterns are substantially varied and that most users have very few distinct tags. The study also found a strong correlation between number of photos and number of distinct tags as well as a relationship between social affiliation (contacts) and tag vocabulary formation.
3
Methodology
Our project was both data-driven and theory-driven. Geert Hofstede’s model was adopted as a comparative framework and considered for most of the quantitative parts of the project. Unlike most past studies, qualitative analysis was also used to balance and complement the quantitative data. So for example, the tags analysis included not only the
288
A. Dotan and P. Zaphiris
total number of common tags shared by users from each culture, but also their content in terms of theme and possible relevance to Hofstede’s dimensions e.g., the frequency of country names was correlated with each culture’s power distance ranking.
3.1 Research approach The research was divided into the following three sections: •
User data – Collecting, comparing and correlating a range of Flickr parameters such as the average number of public photos per user. These average values produced a ‘cultural mean’ for each parameter, which were then correlated with Hofstede’s dimensions in order to explore whether the scores could help explain similarities and differences. This part of the project also included a language analysis in order to examine the extent to which users from each culture used English, their native language or a combination of the two to annotate content (tags, title and description).
•
Group data – Comparing common tags used by users in each culture and common Flickr groups users from each culture are members of. The data was gathered to explore frequency as well as content.
•
Open-ended questionnaire – Used to elicit qualitative data and explore users’ perceptions, motivations and reasons for using Flickr.
Finally, Flickr group discussions were used to further investigate issues that emerged during the data analysis, such as how users from Iran perceive the Flickr Pro account option given the fact they exhibited the lowest number of users with a Pro account out of all the users in the study.
3.2 Procedure The data collection process was systematic and iterative to ensure the validity of the results with regards to the size of the sample used.
3.2.1 Identifying and choosing comparison parameters Flickr offers researchers numerous parameters, primarily quantitative, that could be used to compare users. The site was carefully studied in order to capture all observable options. Once the parameters were identified, the task at hand was identifying those relevant to the study. The selection process was governed by two main considerations: •
Cultural relevance – The extent to which the parameter could be associated with any of Geert Hofstede’s dimensions. The use of language to annotate content for example was seen as more relevant to individualism and power distance than the time the photo was taken or the type of camera used to capture it.
•
Technical/privacy constraints – Not all relevant data could be accessed, as some parameters like privacy settings were private and could not be accessed externally without each user’s login details. For that reason, data such as number of contacts specified as ‘friend’ or ‘family’ could not be incorporated although highly relevant to the study and its objectives.
A cross-cultural analysis of Flickr users
289
After considering these factors, the following nine parameters were chosen for the user data analysis: 1
type of account (pro or regular)
2
number of testimonials written about the user
3
number of contacts
4
number of public photos
5
number of photos set as ‘favourite’
6
number of unique tags across all photos
7
average number of tags per photo
8
number of public groups
9
number of sets.
The language used to annotate content was also captured in order to explore language preferences that could help to understand the rationale for uploading and sharing content, i.e., annotations written in English may suggest the content is aimed at a global audience more than a close group of family members and friends.
3.2.2 Choice of cultures The choice of cultures used in the study was governed primarily by Geert Hofstede’s individualism index as it was seen to be the most relevant to the use of Flickr. The initial hypothesis was that users from a culture with high individualism index would be more likely to use Flickr to express and expose themselves as opposed to users from a culture that is more group-oriented. Furthermore, it was also hypothesised that users from a culture with low individualism index would be more likely to use Flickr to share content with a close social circle, most probably family members and friends. The aim was to choose cultures with high, low and medium individualism value in order to explore to what extent this difference could be observed on Flickr by correlating that value with Flickr data such as the average number of groups a user is a member of. Uncertainty Avoidance was another dimension that was referred to in relation to the use of tags, groups and sets to organise content. The selected cultures and their corresponding scores across all Hofstede’s dimensions are provided in Table 1. Table1
The chosen cultures and their ranking according to Geert Hofstede Uncertainty avoidance
Power distance
Individualism
Masculinity
Peru
16
64
42
87
Taiwan
17
58
45
69
Israel
54
13
47
81
UK
89
35
66
35
Iran
41
58
43
59
290
A. Dotan and P. Zaphiris
3.2.3 User selection At the time of writing, Flickr did not support a ‘search member by location’ option so each user had to be hand picked and reviewed manually using a special sampling tool, developed for the study. Users were selected based on the location and description specified in their profile and had to have a minimum of 100 public photos. By default it was assumed that a user was a member of a national culture, unless anything was written to suggest otherwise (e.g., being a student in another country etc.). The language used in the profile was also considered a reasonable indication. The fact that Flickr is banned in Iran (Garbia, 2007) was discovered once work had already commenced and it was not possible to replace it with another culture. This was a crucial contextual consideration, which was considered during the analysis stage. Such form of government-imposed censorship could be seen as manifestation of Iran’s large power distance and as such, was deemed relevant to the study. From a methodological and cross-cultural perspective, there was no justification to expect all the users to have the same level of access to Flickr. Technological or political related restrictions are major issues that shape people’s abilities, perceptions and experiences.
3.2.4 Data collection – user sampling The data collection was an iterative process that consisted of continues sampling until saturation of results was reached. A special sampling tool was developed in order to collect publicly available user data and display random content and its metadata (40 photos per user). A separate monitoring interface displayed the results in real time showing the mean average and standard deviation across all the parameters. This was a necessary feedback in order to assess the impact of new users on the emerging results. Eventually, the entire sample used in the research was 250 users (50 from each culture). It was decided that a random sample of 40 photos per user would be statistically sufficient (i.e., a total of 10,000 photos were studied in this project).
3.3 Data mining tool A sampling tool was developed using Flickr’s API, which offers limited access to Flickr’s database. The tool was semi-automatic in the sense that it presented and calculated data automatically, but also supported manual data entries that were necessary for the language analysis, as a script could not determine the language used to annotate content. The sampling tool displayed a random sample of 40 photos based on a given user name, and provided the option to save the collection details to a database to be reviewed later. Each photo was presented along with its associated metadata (e.g., title, description, tags, groups and sets). The bottom part of the interface included text fields used for manual data entries. Despite its effectiveness, the tool had two main limitations: •
Access to private data – Due to API privacy restrictions, the sampling tool could only access publicly available data.
A cross-cultural analysis of Flickr users
291
•
Limited access to public photos – Another limitation imposed by the API, was that only the first 500 public photos could be displayed. Users with more than 500 photos accounted for 30% of the entire sample used.
4
Results and discussion
The data analysed in this study was obtained from 10,000 photos, which were uploaded to Flickr by 250 users from five cultures (50 users from each culture). The metadata created by users to describe their photos was analysed both quantitatively (e.g., average number of photos per user in each culture) and qualitatively (e.g., the frequency of certain content). Correlations were calculated between Hofstede’s scores and the various parameters listed in the methodology section. For this study, it was assumed that that correlations with absolute values greater than 0.6 are sufficient to indicate the tendency of the relation between the parameter and Hofstede’s values. Table 2 shows the correlations between Flickr parameters and Hofstede’s dimensions. A ‘cultural average’ was produced for each parameter by calculating the mean average. For example, the average number of public photos a user uploaded or the average number of unique tags used. Most of the correlations with were very weak. Three out of 32 correlations were deemed sufficient. Table 2
Correlations between Flickr parameters and Geert Hofstede’s scores
Pro account
Uncertainty avoidance
Power distance
Individualism
Masculinity
–0.58
–0.27
0.45
0.74
Public photos
0.11
0.10
–0.32
0.04
Public contacts
0.08
0.17
–0.12
–0.49
Public groups
–0.71
–0.27
0.8
0.59
Tags
0.89
–0.16
–0.45
–0.5
Sets
–0.34
–0.15
–0.16
0.5
Users with a testimonial
–0.32
0.23
0.14
–0.14
Favourites
0.11
–0.35
0.20
–0.25
Note: Only three (in italics) were deemed significant
These results are not surprising given contextual considerations and seeing how data such as number of contacts and groups was too general. Standard deviations were considerable across most parameters; however this was anticipated due to the heterogeneous nature of each sample group. Still, the patterns that have been observed and described in this section are consistent; suggesting that despite in-group diversity there is basis to conclude that noticeable culturally related differences exist on Flickr. More significant correlations were found between various Flickr parameters like the number of public photos and number of contacts and favourites. Such dependencies lend themselves to easier interpretation, which is both data and context driven. For example, a strong positive correlation between the percentage of users that have a Pro account and
292
A. Dotan and P. Zaphiris
number of public photos was predictable given that having a Pro account allows users to display more than 200 photos, which is otherwise the limit. Despite the weak correlations with Hofstede’s dimensions, there are clear differences, as well as similarities between the five cultures. This leads us to infer that the model cannot be seen as a reliable means to prove or disprove the existence of cultural differences in this context, when relying purely on quantitative data. Correlations with more qualitative data proved more meaningful as will be described in the following sections.
Account type The account type analysis revealed that among the five cultures, the UK had the highest number of users with a Pro account (84%), followed by Taiwan (76%). By its very definition, a Pro (professional) account could be seen as targeting frequent users who wish to enjoy unlimited storage capacity as well as the freedom to create an unlimited number of sets (collections). Initially, the findings could be seen to indicate that overall, users from the UK and Taiwan use Flickr to greater extent than users in the other groups as they are willing to pay for extended options (at the time of writing a Pro account cost $24.95 a year). It should be pointed out that having a Pro Flickr account does not automatically suggest that a user is a professional or hobbyist photographer, as the questionnaire showed that some users view Flickr primarily as an online storage facility. Since Iran had the lowest number of users that had a Pro account, the issue was raised and discussed in one of the Iranian Flickr groups. Users commented that purchasing a Pro account was regarded as expensive and that due to payment restrictions, users in Iran who managed to bypass the state sanctioned ban on Flickr, experience considerable difficulties purchasing a Pro account. Flickr uses Yahoo!-Pay and Paypal online payment services to allow people to buy a Pro account, however both do not accept payments from Iran (this was confirmed by Paypal). Users also commented that international banks which provide credit cards such as MasterCard and VisaCard do not support Iranian banks making it difficult to use them instead of Paypal and Yahoo!-Pay. At the time of writing, such claim could not be fully verified.
Public photos The public photos analysis revealed that users from Taiwan have on average the largest number of publicly available photos (mean = 961, SD = 795). Users from Iran show a significantly small amount (mean = 198, SD = 128), while users from Peru (mean = 513, SD = 896), Israel (mean = 485, SD = 629) and the UK (mean = 198, SD = 128) are located in the middle with relatively similar values. These findings could be linked to the number of users with a Pro account in each group, as users with a regular account are limited to 200 public photos. Iran has the lowest number of users with a pro account (30%), which explains the low number of public photos. When considering these findings, it is worth referring to the questionnaire, which covered 45 participants, with almost equal number of users from each culture. It highlighted a variety of reasons why people use Flickr such as socialising, storing content online, receiving feedback and photography related advice, keeping friends and family
A cross-cultural analysis of Flickr users
293
informed about their lives and so on. It is clear that Flickr fulfils different needs, which may be linked to the amount of content uploaded to the site. Someone using Flickr primarily as online storage is more likely to consider quantity over quality unlike a user, who uploads content for others to comment on and provide professional advice. When asked ‘What are the main reasons you use Flickr?’ the following response was typical among users from the UK: “I use flickr to display my photos, connect with other photographers and to learn how to be a better photographer. I also use it to socialise and meet up with people”. Israelis referred more to social aspects highlighting sharing content with friends and strangers and some mentioned being able to put their life on display for others to see: “Although I am a very private person, Flickr is a way to expose myself in front of others. It’s a way to explore new areas of my personality as well as that of others. It is also a way to document a period and get motivated to take photographs”. One important issue is how selective users are about the type and amount of content they upload, taking into account technical constraints imposed by the type of Flickr account they have. The quantitative and qualitative data suggests that users from the UK, Iran and Israel are relatively more selective and place emphasis on quality and meaning rather than quantity. Figure 1 Number of users with a Flickr Pro account/GDP per capita (correlation: 0.91) (see online version for colours)
Users with a Pro account
90% UK
80%
Taiwan
70% 60%
Israel
50% 40%
Peru
30%
Iran
20% 10% 0% $0
$5,000
$10,000
$15,000
$20,000
$25,000
$30,000
$35,000
GDP per capita 2006 etc. (Source: CIA - The World Factbook)
Contacts The contacts analysis shows that users from Iran have a considerable larger number of contacts compared to the other four groups (mean = 326, SD = 521). This finding, along with others (testimonials, groups, number of comments per photo) could suggests that users from Iran are more socially active on Flickr, however such assumption is highly speculative in this context since Flickr contacts are different than usual ‘friends’ and ‘buddies’ list found on other social networks. Users may add another user as a contact simply because they want to be notified when that user uploads new content. One hypothesis that emerged throughout the study was that because they have to bypass a state sanctioned ban, Iranian users maintain a close-knit community, in part through mutual contact connections. Unlike users from the four other cultures, Iranians on Flickr take a risk by using a site the government blocked. Such experience may act as
294
A. Dotan and P. Zaphiris
‘social glue’, providing an additional common ground beyond having the same cultural background.
Public groups The public groups analysis showed that users from the UK (mean=94, SD=70) and Iran (mean = 89, SD = 85) are members of more Flickr groups compared to the other cultures. As noted during the literature review and reinforced by the questionnaire, groups on Flickr serve various purposes, not all of which are social. Like contacts, they serve both social and functional purposes. They offer users a place where they can discuss issues with others on specific topics but at the same time serve as promotional mechanism used to potentially expose content to a wider audience. In that sense, being a member of a large amount of groups may actually suggest less social and more self-promotion driven motives. As will be described shortly, compared to the other four cultures, users from Iran have more Flickr groups in common. This could be further evidence of heightened social activity on Flickr among users from Iran.
Testimonials The testimonials analysis showed that users from the UK have fewer testimonials written about them compared to users from the other cultures. However, the UK comes second with 22% after Iran (with 46%) when examining how many users have at least one testimonial written about them. A testimonial could be written for various purposes and most of the testimonials observed during the research that were understood included recommendation and very positive feedback (e.g., “I have known X for several months on Flickr. He has a special gift for art, truth and friendship. I am blessed to know him. He has helped me to expand my view of the world and develop real friendship and connections within our human family. Thank you, X.”). The high number of testimonials users from Iran have written about them could suggest a more significant social experience on Flickr.
4.1 Group data The group data results refer to data more qualitatively by looking at the content (as well as volume) of frequently used tags in each culture, common Flickr groups among users from each culture and the language used to annotate content, i.e., English language versus the user’s native language. Users from the UK and Iran tag have the highest number of tags in common, while users from Peru, Taiwan and Israel share very few tags (see Table 3). When analysing the content of the frequently used tag one can see that a high percentage of users from Peru and Iran used the tag ‘Peru’ and ‘Iran’ respectively compared to users from Israel and UK. Users from Iran had significantly more Flickr groups in common than any other culture, and almost all were Iranian by definition, as opposed to a general group, e.g., Black & White Photography (see Table 5). As far as language used to annotate content, users from Iran and Israel used English almost exclusively. Results regarding common tags used and common groups users from each culture are member of are consistent with the user data presented so far. The high number of tags
A cross-cultural analysis of Flickr users
295
used by users from the UK and Iran, compared to the other cultures, reinforces the notion that on average users from those countries utilise Flickr to showcase their work to a wide and potentially global audience, as opposed to a relative small group of people. Aiming to gain recognition, feedback and exposure by a global and unknown audience could suggest that the content uploaded to Flickr by the majority of users from the UK and Iran fits the description of amateur photography rather than domestic photography, which is less likely to attract professional constructive feedback from other users. Table 3
Common tags used by at least 50% of the users from each culture
% 80%–84%
Peru
Iran
Israel
iran
lima
taipei
tree, sunset
taiwan
church, water
flower, sky, tree
flowers, beach, bridge, london, flower
light, green, girl, tehran
blue, sky, garden , river, red, clouds, reflection, night, england, light
nature, blue, persia, sunset
55%–59%
50%–54%
Total:
UK
peru
65%–69% 60%–64%
Taiwan
flower
9
12
48
telaviv
48
red, night, israel
47
Common tags among users Tag frequency was explored in order to assess collective patterns in terms of themes and culturally relevant references. These included mostly the name of the country, capital city and anything that could have been identified to have a potential political, religious and historical connotation. Like with any content analysis, it was very difficult, if not impossible, to draw conclusions as to the meaning of the tags used. Therefore the analysis focused on commonalities, number of common tags in each group and the language used. The difference between the five cultures was very noticeable and users from the UK and Iran had far more tags in common, i.e., tags used by 30% or more users in the sample. The tags were generic and could be considered strategic in the sense that they are assigned to a photo not simply to describe it, but also to ensure other users discover it while searching the site. This might serve as further evidence regarding the motivation to use Flickr and share content. With regard to qualitative content analysis, it was not possible nor was it realistic to attempt and interpret the subjective meaning of tags. However, the name of a country and capital city was seen as worth exploring given the potential cultural context. As seen in the table above, more than 80% of the users from Iran and Peru used the tags ‘Iran’ and ‘Peru’ respectively. Also, it was more common for users from Iran and Peru to include their capital city’s name as a tag. One hypothesis might be that the high frequency suggests stronger national orientation among users from Iran and Peru compared to the other groups. There is an interesting strong correlation between each culture’s power distance score and the frequency of both the country and capital city’s name.
296
A. Dotan and P. Zaphiris
Cultures with tall hierarchical structure and considerable emotional distance between members of different levels are often nationalistic. This is also supported by the fact they are more likely to be a collectivist culture as Hofstede points out. The link between a culture’s power distance value and frequency of tags could indicate that Hofstede’s scores are indeed visible to certain extent on Flickr, and perhaps in other user content driven websites. As the following two figures illustrate, there is a strong correlation between a culture’s power distance index and the frequency of the capital city and country name in tags by users. If the frequency of the capital city and country name is seen as indicator of nationalistic orientation, that this could suggest that Hofstede’s power distance score can in fact explain these patterns. Table 4 shows the tags that are shared by users in at least three cultures. These findings provide an interesting universal perspective and emphasise similarities that transcend cultural background. Figure 2 Frequency of capital city in tags/power distance (correlation: 0.85) (see online version for colours)
Frequency of capital city
80% 70%
Iran
Peru
60% UK
50%
Taiwan
40% Israel
30% 20% 10% 0% 0
10
20
30
40
50
60
70
Power Distance
Figure 3 Frequency of country name/power distance (correlation: 0.77) (see online version for colours) Frequency of country name
90% Iran
80% 70%
Peru Taiwan
60% Israel
50%
UK
40% 30% 20% 10% 0% 0
10
20
30
40
Power Distance
50
60
70
A cross-cultural analysis of Flickr users Table 4 Tag
297
Shared tags among all cultures Taiwan
Peru
UK
Beach
30%
38%
58%
Flower
50%
Night
44%
56%
Red Sea
30%
Green 30%
Iran
Israel 42%
64%
32%
50%
56%
52%
52%
48%
54%
48%
38%
40%
48%
56%
46%
Sunset
34%
66%
50%
34%
Light
34%
50%
56%
46%
Sky
36%
54%
60%
38%
bw
42%
44%
40%
tree
66%
60%
30%
People
38%
42%
30%
Portrait
42%
46%
46%
Art
44%
34%
30.00%
Language used to tag photos When examining the common tags across the five cultures it is interesting to note that all tags except four are written in English. Only users from Peru use tags in Spanish. These are: de (‘The’), Playa (‘Beach’), Mar (‘Sea’) and Miraflores, which is a seaside district in Lima, the capital city of Peru. This is consistent with the qualitative data gathered in the questionnaire in which the vast majority of respondents from each culture noted they tend to use English to annotate their photos unless it was perceived as having a strong cultural context, which was difficult or impractical to translate to English. When annotating photos, users consider their global contacts and the wider Flickr community, and opt to use English, which is less likely to exclude people. One respondent in the questionnaire referred to English as being more ‘friendly’ than using one’s native language. Tags written in a non-English language are presumably aimed at a relative small group of people such as friends, family members and people from the same country. The use of language, which was also explored with regards to the title and description assigned to a photo may be affected by a number of factors. Besides implying the potential target audience and whether or not it is seen as global or local, using one’s native language could also be affected by English proficiency. Spanish and Chinese are spoken by a large global population unlike Hebrew and Persian, which are spoken in Israel and Iran, respectively. Users from those countries therefore may be more inclined to annotate in English rather than their native language in order to ensure wider exposure. Users from Iran and Israel had very few photos that were tagged exclusively in Persian and Hebrew (Iran = 1.4% of each user’s photo sample, Israel = 5.2% of each user’s photo sample). 55% of the photos in each photo sample for the same users were tagged exclusively in English compared to 7.2% (Peru) and 39% (Taiwan).
298
A. Dotan and P. Zaphiris
Common Flickr groups Similar to the common tags analysis, the common Flickr groups analysis examined frequent Flickr groups among users in each culture. Again, the aim was to explore primarily in-group similarities with regard to Flickr groups users are members of. Groups were analysed according to their apparent subject, the use of language in the group’s name and the percentage of users from each culture who were members of particular groups. Table 5
Common Flickr groups among all cultures
Peru
Taiwan
UK
Iran
Israel
Peru (56%)
Flickr Taiwan (English) (76%)
UK (86%)
Iranian (90%)
Israel (56%)
Peru Photos (56%)
Taiwan (68%)
FlickrCentral (42%)
Iran (68%)
Tel Aviv (44%)
Peru Images (44%)
I ♥ Taiwan (34%)
England (40%)
Iranian Photographers @ Flickr (50%)
=== The Israel Project === (42%)
Lima Fotos! (34%)
Night Images (30%)
Black and White (32%)
Iranian Beauty (50%)
Tel Aviv Stories (40%)
B&W (24%)
FlickrCentral (26%)
London (30%)
Persia and Persian monuments (50%)
flickr.co.il Israeli flickr community (34%)
Latinoamericanos! (22%)
Mobile01 (21%)
The British Countryside (28%)
*ISFAHAN* (48%)
my israel (32%)
Soda Stereo (22%)
Taipei (24%)
Sunsets and sunrises around the world (28%)
Iranian Photo Contest (42%)
B&W (30%)
South America public group (20%)
FLOWERS (20%)
UK Landscapes, Sea and Country. (26%)
Iranian Portrait (40%)
TA girls (30%)
Black and White (18%)
Chinese (18%)
Britain in Pictures (Buildings) (26%)
MY WINNERS!Trophy INVITED photos ONLY/Comment on 2 of your faves (40%)
- מפגש פליקר והעיר התמלאה !בפלאש (26%)
The results offer great deal of insight and reveal interesting findings and clear themes. For example, the majority of groups shared by Iranians have a national orientation as the word ‘Iranian’ appears quite frequently and in context that is unique to this culture compared to the other four (e.g., Iranian Women, Iranian Beauty, etc.) On the other hand, most groups common to users from the UK revolve around pictorialist themes of
A cross-cultural analysis of Flickr users
299
landscapes and objects, reinforcing the notion that the majority of them are hobbyists and not domestic photographers likely to use Flickr to share content with friends and family members. As explained in the methodology section, since some users were sampled from the groups UK, Iranian, Israel, Taiwan and Peru, these groups should be ignored. The percentage indicates the number of users in each culture that are members of each Flickr group. When analysing the results, one of the most noticeable findings is that beyond the fourth most common group, the percentage drops significantly below 30% as the subjects become more specific and the chances of users in each sample having specialised groups in common diminishes. This is true in all cultures except Iran where the lowest percentage is 34% compared to 16%, 14%, 22% and 20% in Peru, Taiwan, UK and Israel respectively. The percentages across all 15 groups common by users from Iran are relatively high with a small margin between each one. This raises two possibilities: •
As a collective, users from Iran have more similar interests on Flickr compared to users in the other cultures in this study. In that respect, they are more homogenous.
•
The number of users from Iran on Flickr is smaller than users from the other cultures in the study; therefore they are more likely to cluster together in common groups.
The fact that a considerable number of users from Iran in the sample are members of the same Flickr groups links to previous findings suggesting high social activity, primarily the percentage of users that have a testimonial written about them and number of contact and favourites.
5
Conclusions
This study sought to explore cultural differences in user content driven websites by focusing on Flickr as a case study, as well as explore the extent to which Geert Hofstede’s cultural model could be used to explain any differences and similarities The data analysis revealed noticeable and consistent trends and patterns; however most were too contextualised and sometimes general to be interpreted using Geert Hofstede’s model. The research covered a wide range of quantifiable parameters as well as qualitative data. Interesting patterns, correlations, similarities and differences emerged which strongly suggest that cultural background has an observable impact on the use of Flickr and potentially similar user content driven websites. Correlations with Hofstede’s scores which were based purely on quantitative date were weak, suggesting either the model is not very useful for exploring cultural difference in this context, or that cultural differences are less visible given the data at hand. Some previous studies also suggest that the model cannot account for cultural differences regarding the use and implementation of websites. In that regard, the findings presented here are consistent. The study focused on overall group differences and aimed to answer such questions as: “Would users from Peru be more likely to have fewer contacts than users from the UK?” for example. The view taken is statistical and does not try to account for the
300
A. Dotan and P. Zaphiris
behaviour of specific individuals. Flickr users are unique individuals that are affected by the culture they were born into and all are also affected by a greater and ‘Flickr culture’ that every user, to a certain extent could be seen to be a member of. The study showed that there are clear user preferences regarding annotation patterns and choice of language. Users from Iran and Israel opted to use English to annotate their content, presumably in order to reach a wider global audience. This was less of a concern for users from Peru and Taiwan whose native languages are more widely spoken in the world. Cultural background was also evident when examining the number of common tags shared by users in each culture. Users from Peru and Taiwan had few tags in common, which could suggest that they are less interested in sharing the content they upload with a global, perhaps unfamiliar audience. Such hypothesis could be supported by extensive use of Spanish and Chinese to annotate content compared with Hebrew and Persian. All this could suggest that the motivation to share content online with strangers around the world rather than familiar people (friends and family) has some cultural bias. As Peru and Taiwan had the largest power distance and lowest individualism value, one might argue that a link does exist between Hofstede’s scores and the patters we have observed on Flickr.
6
Future research
This study has shown that user content driven websites like Flickr offer researchers a vastly rich social environment with an abundance of options to explore cultural difference on the internet and in Web 2.0 sites in particular. From a cross-cultural perspective, there is considerable scope to explore the following aspects on Flickr and other user content driven websites: •
Private versus public – This study did not address the privacy settings found on Flickr and it strongly felt that referring to them in a comparative study could prove extremely valuable as it may expose culturally-related privacy preferences. Focusing on private content could provide data specific enough to yield more meaningful results when correlated with Hofstede’s dimension. One possible hypothesis could be that users from cultures with low individualism value may be more likely to have private photos, contacts and groups.
•
Visual content analysis – The type of content that users upload to Flickr is extremely diverse and very generally could be classified as either domestic or non-domestic as some users see themselves as amateur photographers and others take more casual photos. It would be interesting to explore this from a cross-cultural angle and explore as well specific themes that users from different countries upload, though such an approach would be less HCI-related.
•
Extending current study – Given the sampling tool and monitoring interface are in place, this study could be extended in the future by either adding more users from the current five cultures or introducing new ones. The latter would provide a richer overview regarding the data that has been collected so far.
A cross-cultural analysis of Flickr users
301
References Ames, M. and Naaman, M. (2007) ‘Why we tag: motivations for annotation in mobile and online media’, in Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 28 April–03 May 2007, San Jose, California, USA, pp.971–980. Barber, W. and Badre, A. (1998) ‘Culturability: the merging of culture and usability’, in Proceedings of the Fourth Conference on Human Factors and the Web, 5 June 1998, Basking Ridge, NJ, available at http://zing.ncsl.nist.gov/hfweb/att4/proceedings/barber/ (accessed 10 March 2007). Boyd, D.M., and Ellison, N.B. (2007) ‘Social network sites: definition, history and scholarship’, Journal of Computer-Mediated Communication, Vol. 13, No. 1, article 11, http://jcmc.indiana.edu/vol13/issue1/boyd.ellison.html. Burgess, J., Foth, M. and Klaebe, H. (2006) ‘Everyday creativity as civic engagement: a cultural citizenship view of new media’, in Proceedings Communications Policy & Research Forum, Sydney, available at http://creativitymachine.net/downloads/publications/Burgess FothKlaebe_CPRF_EverydayCreativity.pdf, (accessed 15 November 2007). Burgess, J. (2006) ‘Vernacular creativity, cultural participation and new media literacy: photography and the Flickr network’, AOIR7.0 Internet Convergences, Brisbane, Australia, 27–30 September 2006, available at: http://eprints.qut.edu.au/archive/00007828/ (accessed 15 November 2007). Condon, J.C., and Yousef, F.S. (1975) An Introduction to Intercultural Communication, Macmillan, New York. Flickr Public Relations, personal correspondence, 5 December 2007. Garbia, B.S. (2007) ‘Iran: Flickr users vs. The State’, available at http://advocacy.globalvoicesonline.org/2007/02/14/iran-flickr-users-vs-the-state/ (accessed 10 December 2007). Gould, W.E. (2005) ‘Synthesizing the literature on cultural values’, in N. Aykin (Ed.): Usability and Internationalization of Information Technology, p.92, Lawrence Erlbaum Associates, Mahway, NJ. Gudykunst, W.B. (Ed.) (1986) Intergroup Communication, Edward Arnold, London. Hall, E.T. (1976) Beyond Culture, Doubleday, New York. Hofstede, G. (1991) Cultures and Organizations – Software of the Mind, McGraw-Hill, London. Hogan, M. (2006) ‘Tag, you’re ‘it’: preserving the photographic personal archive through Flickr.com’, available at http://www.docam.ca/Pedagogie/Seminaire_2006/Mel_Hogan.pdf (accessed 13 August 2007). Koman, R. (2005) ‘Stewart Butterfield on Flickr’, O’Reilly network, available at http://www.oreillynet.com/pub/a/network/2005/02/04/sb_flckr.html (accessed 16 September 2007). Marcus, A. (2005) ‘User interface design and culture’, in N. Aykin (Ed.): Usability and Internationalization of Information Technolog, pp.54–47, Lawrence Erlbaum Associates, Mahway, NJ. Marlow, C., Naaman, M., Boyd, D. and Davis, M. (2006) ‘Ht06, tagging paper, taxonomy, flickr, academic article, to read’, in HYPERTEXT ’06: Proceedings of the seventeenth conference on Hypertext and hypermedia, New York, NY, USA, 2006, ACM Press, pp.31–40. Mathes, A. (2004) ‘Folksonomies – cooperative classification and communication through shared metadata, unpublished paper’, available at http://www.adammathes.com/academic/computermediatedcommunication/folksonomies.html (accessed 2 December 2007). Lerman, K. and Jones, L. (2007) ‘Social browsing on flickr’, in Proceeding of International Conference on Weblogs and Social Media (ICWSM-07), 2007. Pfeil, U., Zaphiris, P. and Ang, C.S. (2006) ‘Cultural differences in collaborative authoring of Wikipedia’, Journal of Computer-Mediated Communication, Vol. 12, No. 1, article 5. http://jcmc.indiana.edu/vol12/issue1/pfeil.html.
302
A. Dotan and P. Zaphiris
Rafferty, P and Hidderley, R. (2007) ‘Flickr and democratic indexing: dialogic approaches to indexing’, Aslib Proceedings, Vol. 59, Nos. 4/5, pp.397–410, available at: http://www.emeraldinsight.com (accessed 18 October 2007). Sacher, H., Tng, T. and Loudon, G. (2001) ‘Beyond translation: approaches to interactive products for Chinese consumers’, International Journal of Human-Computer Interaction, Vol. 13, No. 1, pp.41–51, available at: http://leaonline.com/doi/pdf/10.1207/S15327590IJHC1301_3 (accessed: 20 August 2007). Syarief, A., Giard, J.R., Detrie, T. and McBeath, M.K. (2003) ‘An initial cross-cultural survey of user perception on web icon design for travel websites’, the 6th Asian Design International Conference, 14–17 October 2003, Ibaraki, Japan, online, available at: http://www.idemployee.id.tue.nl/g.w.m.rauterberg/conferences/CD_doNotOpen/ADC/final_pa per/013.pdf (accessed 10 March 2007). Van House, N.A. (2007) ‘Flickr and public image-sharing: distant closeness and photo exhibition’, in CHI ‘07 Extended Abstracts on Human Factors in Computing Systems, San Jose, CA, USA, 28 April – 03 May 2007, CHI ‘07, ACM, New York, NY. Wolk, D. (2009) Future of Open Source: Collaborative Culture, wired magazine, available at: http://www.wired.com/dualperspectives/article/news/2009/06/dp_opensource_wired0616 (accessed 6 July 2009). Yunker, J. (2003) Beyond Borders – Web Globalization Strategies, New Riders Publishing, Berkley.