Ethics and interdisciplinarity in computational social science

Transcription

1 Methodological Innovations Online (2012) 7(1) Ethics and interdisciplinarity in computational social science Fabio Giglietto and Luca Rossi Department of Communication Studies, University of Urbino Carlo Bo Abstract During the last few years a growing amount of content produced by Internet users has become publicly available online. These data come from a variety of places, including popular social web services like Facebook and Twitter, consumer services like Amazon or weblogs. The research opportunities opened up by this socio-technological innovation are, as shown by the growing literature on the topic, huge. At the same time new challenges for social scientists arise. In this paper we will focus on two of the main challenges posed to the growth of the so-called computational social science: interdisciplinarity and ethics. While the searchability and persistence of this information make it ideal for sociological research, a quantitative approach is still challenging because of the size and complexity of the data. Collecting, storing and analyzing these data often require technical skills beyond the traditional curricula of social scientists. These projects require, in fact, collaboration with computer scientists. Nevertheless developing a common interdisciplinary project is often challenging because of the different backgrounds of the researchers. At the same time the availability of this content poses a challenge concerning privacy and research ethics. Due to the amount of data and the fact that the real identity of the author is often hidden behind a nickname, it is often impossible to ask the subject involved to consent to the use of their data. On the other hand, especially in the first wave of web 2.0, this information has been - intentionally or not - publicly shared by the users. While a technique of dis-embedding the identity of the user from the content analyzed is often the solution used to bypass this issue, an even more important privacy-related challenge for computational social science is emerging. Due to the wide adoption of social network sites such Facebook or Google+, where a user may decide to share his content with his/her group of friends only, the amount of public data will change and decrease in the future. We will discuss this issue by enumerating a number of possible future scenarios. Keywords: Social media, privacy, computational social sciences, research ethics Introduction Social Network Sites (SNS) like Facebook and Twitter are increasingly used not only for maintaining relationships, but also for sharing news, media content and information both within intimate groups of contacts and among widely distributed publics. Because of this convergence of social networking with media consumption and public communication, these social media platforms represent a significant shift away from Correspondence: Fabio Giglietto, Università degli Studi di Urbino Urbino, Italy. fabio.giglietto@uniurb.it ISSN: online DOI: /mio

2 broadcast media models and give a new and largely unexplored role to the users in communication processes. Twitter and Facebook, in particular, have become instrumental information sources in contexts ranging from global phenomena like natural disasters and political upheavals (Earle 2010) to everyday chatting among friends (boyd 1 et al 2010). At the same time these techno-social spaces introduce a new form of large-scale interpersonal communication in more mundane contexts like cultural and sporting events (Wu et al 2011). The digital nature of these data, together with their abundance unprecedented in social science history, leads, as we will see later, to the emergence of rigorous data-driven approaches as the only viable way to understanding the dynamics of communication in these contexts. The first research steps taken into this field have focused on qualitative approaches in order to understand a range of cultural and social practices within these platforms (Wu et al. 2010; Honeycutt and Herring 2009; Marwick and boyd 2010; boyd et al 2010) often by employing limited data sampling methods rather than studying more comprehensive datasets representing public online conversations (Wu et al., 2010). Additionally, much existing work has been limited to studies of single platforms (boyd et al 2010; Marwick and boyd 2010) with an in depth approach derived from previous online ethnographic research (Kozinets 2012). While these methods have many positive aspects in understanding hidden social practices and in placing new communication technologies within the users life context, they often lack the ability to address the complex and interconnected dynamics of the wider media system on a larger scale. As a consequence of this, during recent years, we have seen a growing gap between the kind of sociologically relevant data that were available and the kind of research methods that were - and largely still are in use. The reasons behind this gap are basically rooted in the paradigm shift that these new data and approaches require. Specifically, dealing with large amounts of digital data, requires the adoption of new perspectives and methodologies taken from research fields often far from the social sciences area: computer sciences and statistical sciences. Nevertheless once the initial difficulties are resolved and a truly interdisciplinary research process emerges, the benefits in terms of what can be studied from a sociological perspective become evident. This article will describe the complex set of practices behind the emergence of a computational data-driven sociological research by showing a research example where this methodology has been fruitfully used and by mentioning several other projects where it has also been used. In addition to that, the article will discuss the ethical issues surrounding the use of large quantities of unfiltered user-generated content or information by focusing on a specific example taken from our research experience. Background and description of approach The digital data-driven approach, combining together a large quantity of digital data and sociological analysis, is an emerging approach that has been identified by some as a new form of escience, comprising both ehumanities and esocial Sciences (Dutton et al 2010), Web Science (Manovich 2008; Hendler et al 2008) and Big Social Data (Manovich 2011). The origins of such perspectives can be traced back in time (Edling 2009) but nowadays are strongly influenced by the increasing availability of large and near real-time digital datasets for example on interactions in social media spaces (Lazer et al 2009). The approach could also be placed in the context of Culturomics (Michel et al 2011). Contemporary communications and social processes leave - both intentionally and unintentionally - a growing number of digital traces: personal communication shared in social network sites, family relationships declared on Facebook or political thoughts and opinions posted on Twitter are just the top of the iceberg of the dataset available to digital social researchers. Persistence and searchability (boyd 2008) of digital data made this information available to all those who had the proper technological skills to deal with the amount of data. The process that leads towards innovation in research methodology starts with acknowledgment of the existence and the value of such data. 26

3 The concept of persistence, introduced by boyd, not only stresses the technological persistence of contemporary communication which is a condition well-known to any new media scholar but also underlines how nowadays this characteristic pervades every aspect of our communication. It is getting harder by the day to make a distinction between communicative content produced to stay and that which is produced for short-term purposes. Every message in our digital experience is stored and saved (made persistent) even if it is just an online chat between friends. Those who are familiar with the popular Gmail chat service are well aware of this phenomenon and of its consequences: every chat message is automatically saved as if it were a mail message and can be searched for and retrieved in the future. This applies not only to the Google chat service but to the majority of digital communication sites. Online chatting within this perspective no longer has the volatile nature of offline chatting but instead approaches the permanent and indexed status of written text. At the same time, mobile devices, such as the Apple iphone or Android based phones, may host applications aimed to log, store and share data about walked distances, visited places or fitness activities. In conjunction with ad hoc Bluetooth/WiFi-enabled monitoring devices, these mobile phones may also log heart rate, blood pressure, weight and several other physical parameters. Once logged and stored, these data become permanent. When shared (several applications offer the opportunity to share these data with friends in order to show progress and get advice and support from other users), this information often becomes a starting point for conversations among friends. An initial consequence of this shift is that every kind of communication both formal and informal, whether official document or trivial chat now has the characteristics of permanence and searchability which before the digital era were only available for certain kind of documents. Searchability, which is in fact made possible by persistence, is the second key characteristic of digital communication that has to be explained. As a matter of fact, while writing, as technology allowed a large quantity of communication to become available as data (and the quantity of available data increased with the alphabetization rate of societies), only a tiny part of these data were actually made searchable (Giglietto 2009; Giglietto and Rossi 2009). That was the communication perceived worthy by society (literature, law codex, technical information) and we had special places and professions to conserve and manage this information. We had a clearly identifiable distinction between what has been called a cured semantic (Leydesdorff 2000) which was permanent and searchable, and what we could define as trivial semantic which, even when it was permanent, had a private dimension, such as private diaries or correspondence between lovers or friends. Within this scenario, social sciences largely chose to work with searchable data while only minor research was based on the privately-produced communication and in those cases it necessarily followed a qualitative approach due to the specific unsearchability of that kind of data. Contemporary digital communication is radically changing this scenario by allowing us to conduct research on a large quantity of data that has mainly been published directly by the authors (User Generated Content). Therefore all the ongoing diffusive and pervasive digital communication, which characterizes our media scene, is now available as a potential source of data for sociological analysis. These new datasets, available to social researchers, often cover aspects of social life that are generally difficult to investigate with quantitative approaches and are rather understood owing to the use of in-depth qualitative methods: friendship, creativity, civic engagement or political participation are all areas for which, thanks to the analysis of large quantities of online digital communications, new and interesting research horizons are opening up. The prerequisite of the methodological innovation we are describing is, primarily, the perception of a gap between these new unexplored research potentials and the methodological practices that are usually available as standard tools for social researchers. 27

4 This is precisely where a strong co-operation between social sciences and computer sciences is needed. Nevertheless, this kind of co-operation requires huge changes both in the perspective of the social scientists and those of the computer scientists: on the one hand social scientists have to understand the restrictions and opportunities of the digital nature of the research data but, on the other hand, computer scientists have to understand that there are far more complex subjects and social dynamics behind raw digital data that can only partially be explainable formally. On a theoretical level we could divide the process of researching with digital user-generated data into three different phases, each with specific issues: 1) Data identification 2) Data acquisition and cleaning 3) Data interpretation Data identification As we already mentioned, the digital data available to researchers nowadays are extremely heterogeneous: from blog entries to Twitter status updates, from geolocalized check-in to pictures posted on Flickr, the range of digital spaces that can be investigated in the search for socially relevant answers is vast and of a great complexity. Every piece of digital data is, in fact, inevitably bound to the social practices behind that online environment. The social value of a picture shared and saved in a digital album on Flickr is very different from the value of a picture quickly shared for a laugh on Twitter. At the same time, a topical comment on current events posted on Facebook will have a different perspective to a blog entry made on a thematic weblog. Understanding these different social dynamics and different social values is a necessary task that the researcher has to accomplish in order to start a deeper data-driven analysis, because any subsequent results will of necessity have to be framed within the socio-technical description of the online space where the data were originally generated. Among the various possibilities for researchers, the analysis of content posted on social network sites (like Facebook or MySpace) or microblogging sites (like Twitter of Friendfeed) seems to be the most common choice so far for researchers working with this kind of approach. Data acquisition and cleaning When researchers are dealing with messages posted on microblogging sites by a large number of users the problems raised are both technical and analytical. On the one side, it is necessary to find a reliable way to retrieve, store and then query the data. This can be usually done using the application programming interface (API) that allows third party software tools to acquire a large quantity of data in a structured manner. Once acquired, the data will most probably need to be cleaned from the increasing amount of spam and irrelevant messages that are commonly present on every digital dataset of user-generated content. This cleaning phase, when it involves large quantity of data, can only be done automatically, therefore requires a detailed understanding of the kind of data required for research. These preparatory phases are the most complex and therefore an effective co-operation between scientists of different backgrounds is of paramount importance. Incorrect data selection or cleaning could lead to the loss of large quantities of data and of months spent on data acquisition. What has to be fully understood is that computational support will not end with the first phases of the research. In fact, due to the quantity of data involved, only a computer-based data retrieval technique can enable the social researcher to focus on the most relevant data. 28

5 29 Data Interpretation So far, the research phases have only dealt with the preliminary tasks necessary to obtain a reliable, clean dataset for sociological analysis and interpretation. One of the major strengths of a digital data-driven approach is that once the dataset has been acquired, a large variety of sociological analyses are still possible. As we will see later in the examination of a specific case study, the availability of a large and well-structured digital dataset allows the researcher to perform both qualitative discourse analysis on the collected texts and statistical analysis of users activity, or to identify time trends in users actions. Within this context, digital data-driven approaches are mostly aimed at providing researchers with large quantities of rich digital data and not at forcing researchers towards a specific approach. Nevertheless, in the case of hundreds of thousands or even more textual entities, qualitative approaches require adaptation with collaborative practices among a large group of researchers supported by tools available using the latest software packages. These general statements acquire a new meaning when it comes to research practices. The following paragraphs address these issues in the context of one recent research project which focused on the analysis of the online conversations generated by the death of Italian TV host Mike Bongiorno on 9th September Case Study Michael Nicholas Salvatore Bongiorno, best known as Mike Bongiorno, was considered to be the most famous TV host in Italy, with an incredibly long and successful career dating back to His death on 8th September 2009, had a huge impact on the Italian media and thousands of people attended his State funeral. A brief background to Mike Bongiorno s life is required in order to understand many of the results that we are going to present later in the article as well as the scale of reaction to his death. Mike Bongiorno was born in New York in 1924 to an Italian expatriate family and moved back to Italy after the divorce of his parents, at the age of eleven. After a long series of perils during the Second World War (he collaborated with the Italian resistance against the Nazis occupation and was imprisoned in Nazi concentration camps from 1944 to 1945), he started working as an announcer on the Italian national radio channel in At that time Italy was still in the process of conceiving national broadcast television, which officially started only two years later, in 1954, with Mike Bongiorno as one of the hosts of the prime time programs. A few years later, Mike Bongiorno introduced in Italy an adapted version of the US quiz show The $64,000 Question under the name Lascia o Raddoppia which enjoyed incredible success and had a huge impact on Italian popular culture of the time. For the next 50 years Mike Bongiorno worked continuously as a TV host, both in the public state-owned channels and later in many privately-owned channels. His name has always been linked to the quiz-show genre, which he introduced in Italy and, due to his astonishingly long career, he has always been described as the father of Italian television. His role was so important in the construction of a modern Italian pop culture that calling him a star would be reductive. To that effect, Umberto Eco wrote an essay titled The Phenomenology of Mike Bongiorno in 1961, in which he used the popular TV host to explain how television was changing modern Italian society (Eco 1993). The study of online communication about the death of Mike Bongiorno was therefore aimed at two different goals: on the one hand, the research aimed to test the new data-acquisition and analysis methodology and, on the other hand, it wanted to observe what kind of social phenomena the breaking news of the death of a wellknown public figure generates online 2. To perform our analysis we monitored the activity of Friendfeed, a well-known microblogging service, for two weeks, from September 6, 2009, 00:00 AM to September 19, 2009, 12:00 PM. The service was monitored at a rate of about 1 to 2 updates every second (depending on network traffic) through an ad hoc software solution querying the Friendfeed API. First we used a language

6 identification tool to retrieve all Italian posts, as due to the local relevance of the object of the study we had to obtain a subset of the data based on a national criterion. Then, after a manual analysis of the most relevant events taking place during the monitoring period, we chose some keywords to be used to filter the entries concerning those events. While some keywords unambiguously identified a specific topic, some others could refer to multiple distinct topics. As an example, although Mike is not a typical Italian name and it usually refers to Mike Bongiorno, we found a few entries related to other persons with the same name, and we therefore had to manually clear the resulting dataset. At this point, we had a set of posts about Mike Bongiorno and could analyse all the conversations and messages. Figure 1.Friendfeed messages on Mike Bongiorno produced between Sep. 6 to Sep An initial observation of the collected data, made possible only through analysis of the aggregated data, stresses the time trend of the online conversation as shown in figure 1. Conversations about the specific topic reach a clearly visible peak on the day of the death and a minor peak on the day of the funeral. These two peaks represent the public interest in a specific topic that can, in this way, be clearly mapped and visualized. After a preliminary description of the data, we started a more detailed observation of the conversations and discovered that more than the 50% of the comments were generated by only 7 entries while 364 entries generated no comments at all. This difference between a small section of the entries that was heavily commented on and shared and a larger section of entries that was basically ignored has already been observed several times in online social activities and it has been commonly related to the more general power law distribution or the distribution. In addition to this extremely skewed distribution of activities, it is worthy of note that the first message reporting news of the death received a huge attention, collecting more than 130 comments in a few hours. By the end of the day the death of Mike Bongiorno, discussed in several threads, had generated 585 comments, 276 made during the first hour. These data, made possible only through computer-based computation and filtering techniques, suggested that several different social practices were taking place at the same time originating from the same event (the death

7 of a well-known person). Why did some of the messages generate a huge number of comments and interactions while others received few or no reaction at all? Entries with zero comments represent the large majority of entries (N = 364). What are all those entries about? Many of them seem to have an informative function, but then why did people keep posting about the death of Mike Bongiorno when the news was already spreading within the Friendfeednetwork? If we proceed with a qualitative analysis of those entries, it is possible to highlight two main social uses. The first one is the explicit news propagation and the second one is what we could define as the mourning ritual of the networked public. On microblogging sites, news propagation occurs through the choice of a single user to re-share a specific content within his/her personal network of followers. Many of the entries with no comments at all have this specific function: to convey the news. They are not supposed to start any conversation but just to contribute to the diffusion of some news that is perceived as important. While this propagation activity seems to work for the messages posted just after the event, there are many messages posted when the original news had already widely spread. These messages do not seem to be aimed at communicating en event, but at expressing a final homage for the death: Bye Mike! We're missing you! or Bye Grandpa Mike! Bye Mike, you've been a milestone of our TV. Those entries are not intended to convey the news; they generally aim to place the event in the users own stream of online activities and to show their feelings towards it. It seems to be a way to express, in the networked semi public space, a feeling towards a public event. It does not really ask for comments. Besides the main findings of this research which have been extensively presented elsewhere (Magnani, Montesi and Rossi 2010), what must be stressed from the methodological point of view is on the one hand the research procedure that follows the schema we have identified (data identification; data acquisition and cleaning; data interpretation) and, on the other hand, that this led to some findings which would have been achievable without the computational analysis of a large quantity of data. Reflections The approach exemplified above by the case of Mike Bongiorno, has been employed in several other research projects developed by our group of colleagues at the Department of Communication at the University of Urbino Carlo Bo. We have explored different kinds of social networks and content sharing sites: Flickr!, YouTube, Blogs, Twitter, FriendFeed and lately the public contents available on Facebook pages. The horizon of topics addressed by these projects is also vast and varied: we compared European Union and United States identities by analysing Flickr! Photos (Giglietto and Rossi 2006), observed the vanishing of public/private boundary in YouTube videos of pregnancy (Fonio et al 2007), retrieved and analysed blog conversations concerning the youth generational identity in Italy (Rossi 2010), re-constructed and visualized Twitter conversations on the rescue of a group of Chilean miners trapped underground for 69 days in 2010 (Rossi et al 2011) and explored the idea of using Facebook Likes to forecast electoral results in the administrative elections in Italy (Giglietto 2012). We learned a lot through these experiences and we are now able to easily identify the right social media for the topic we decide to address. At the same time, we have both developed and identified software tools which facilitate retrieval of contents, as well as support the quantitative and qualitative process of content analysis. While presenting these projects, especially during the first years, we sometimes encountered scepticism expressed by our fellow colleagues in Italy and abroad. The most frequent issues concerned the statistical representativeness of social media users among the whole population and the 31

8 ethical issues related to the use of content without asking the permission of the authors/subjects involved. The first remark often also raised for Internet surveys, as pointed out by Marcel Das in his paper (Das 2012) was usually countered firstly by clearly declaring the non-representative nature of the analysis carried out and, secondly, by highlighting the constantly increasing number of people involved in the use of social media. Even though the proportion of the population using social media is constantly increasing, Twitter or Facebook users can t be considered representative of the entire population. This initial remark is now becoming less common, while new, less ingenuous questions concerning sampling strategies and transparency in the process of data retrieval are often raised (boyd and Crawford 2011). On the other hand, the ethical issues are trickier and give rise to important considerations on the status of usergenerated content itself. The issue of ethics in Internet research has a long tradition. The topic was first raised as a critique to the very first computer-mediated communication (CMC) studies, such as the ethnographic work published in 1995 by Sherry Turkle. Bassett and O'Riordan (2002) pointed out the limits of the traditional human subjects research ethics (first the rights of the human subject and then the aims of the researcher) when applied to Internet studies by arguing about the textual and not only the spatial nature of the Internet. Even at that early stage of CMC s development, the need for a distinction of the ethics relating to the various internet services ( , newsgroups, thematic forums and personal public websites) was already clear. The way traceability as some authors call what we refer to as searchability in this paper affects the ethics of an ethnographer doing e-research is addressed by Beaulieu and Estalella. Starting from a critical discussion of their own fieldwork, the authors highlight the broad impact of e-research on the entire research eco-system beyond the traditional issues of privacy and copyright (Beaulieu and Estalella 2011). Internet research ethics is also an issue often discussed in the mailing list of the Association of Internet researchers 3. Recently, a working group of this association posted a draft of AOIR Guidelines for Internet Research Ethics for review by fellow association members. The document clearly frames the issue in the context of guidelines for researchers as opposed to policy and advocates an approach to ethics which considers the specific needs of each case. Different types of data collected, contexts, venues and interactions bring the researcher to different questions about ethical practice. When it comes to the specific case of user-generated content retrieved from a social network site, prevailing opinion seems to be a thoughtful approach brilliantly summed up by the sentence just because it is accessible doesn t make it ethical (boyd and Crawford 2011). While most researchers agree on applying ethical policies mutated by the tradition of participatory observation, most of the discussions recently has focused on the distinction between public and published. Consensus on the definition of these terms and their relationship to one another is still lacking when it comes to usergenerated content. While these contents are deliberately published by authors in a public space, the standard ethical approach would require researchers to inform subjects involved in the research and ask for their permission. Unfortunately, asking for such permission in large scale web projects is technically unfeasible. It is possible to strip personal information from data and make it anonymous, but most of the times this process is still considered questionable, since it is almost impossible to automatically strip all the personal references that may appear in a blog post, photo, tweet or video. The idea of analysing user-generated content was first conceived as an extension of the traditional study focused on content analysis of mass media. It is interesting to note that such an ethical issue was never raised when the content being analysed is produced by journalists. It seems logical to analyse content that was deliberately made public by someone without asking permission or informing the author when the content is produced by a journalist and printed in a newspaper. What changes if the author is not a professional journalist, but an amateur using the Internet to publish their owncontent? 32

9 This question brings us to the core of the ethical issue. Since the amateur Internet publishing phenomenon is relatively young, we are still lacking a clear and shared understanding of the status of these contents. Sometimes it is not clear whether the amateur author is aware of the audience and of how networked publics could easily scale. In such a chaotic scenario, we should probably leave the researchers the right to experiment without the strict application as sometimes practised by the Institutional Review Boards and other research ethics committees of old rules to a changed world. Sometimes the nature of the social network analyzed, plays a role in the ethical choice of the researcher. Facebook and Twitter are both aimed at supporting the user in the process of sharing their content, but while the first is understood by most users as a protected networked platform where content is shared among a restricted group of other users only (Friends), the second, on the other hand is public by default. The privacy policies of Facebook make it hard if not impossible to retrieve content from restricted user profiles. Accessing these data may be possible by asking and being accepted as a Friend by the subject we desire to observe. In the Friend request the researcher should inform the subject of the aim and scope of the research project. While most Facebook profiles are restricted in some way to a certain audiences, some profiles are deliberately or accidentally left open. Ad hoc search engines, such as may easily retrieve these contents, but research on this data should most probably be considered unethical, since it is unclear whether the user wishes to publically share these. Besides user profiles, Facebook also hosts, so-called, pages or business pages. All the contents published on Facebook pages (posts by page administrators, posts by fans, comments to both) are public by default. Contents are indexed by search engines and it is possible to retrieve and store these contents using tools such as DiscoverText ( Even if it is impossible to prove, it is reasonable to believe that users who posted or commented on Facebook pages are aware of the fact that they are exposing their contents to a public arena. Acting in a public arena may cause your actions to be observed by someone. Research on this kind of contents should therefore be considered a slightly more ethical method than the one described above. If we decide to focus on the clear will of the author to share contents publically in order to decide if (and to what degree) it is ethical to research on these contents without asking permission and without informing, we should consider Twitter rather than Facebook. By its very nature, Twitter is aimed at sharing news and information with the world. The goal of the user sharing a tweet is most often to reach the largest possible audience (even if exceptions exist, see Marwick and boyd 2010). Since most research projects focus on public events, it is reasonable to assume a clear will or awareness on the part of the users that they are sharing contents publically. Often researchers retrieve contents from Twitter by querying the system with a specific hashtag. A hashtag is a word prefixed by the # symbol chosen on a peer-to-peer basis by the user community to identify tweets related to a specific topic. The hashtag is deliberately added to a tweet by the users and may therefore be considered proof of the will to make their tweets part of a bigger public conversation. Retrieving and analysing these contents should not therefore raise ethical concerns. 33

10 34 Conclusion In our experience, the process of methodological innovation was triggered by a technological and social innovation. At the same time the idea of using user-generated content for sociological research may be considered an extension of traditional study based on the content analysis of data produced by mass media. Although the first exploratory works in this field date back to 2006 and the methods described in these pages have been quickly adopted by Internet marketers, they are still sometimes considered questionable by the community of social scientists. Whilst concerns about the representativeness of the Internet population as compared with the whole population are becoming less and less common, the ones relating to the ethical issues still require some consideration. We propose to consider practices such as using hashtags as a clue to users willingness to be part of a public conversation. Retrieving and analysing this data should therefore raise less ethical concerns. A close collaboration between social and computer scientists is still highly recommended even though, some research software packages released during the last few months now offer ready-made tools aimed at helping social scientists dealing with online data. The collaboration is still crucial because social scientists are not adequately trained in dealing with large databases. Skills such as advanced statistics or social network analysis should be part of the curriculum of future social scientists. At the same time, dealing with large databases requires investment often beyond the average budget of research projects in the social sciences. The phenomenon of user-generated content is still relatively new. Content-sharing platforms, as well as user behaviors, are constantly changing. A regular review of these technical and social developments is a key factor for the success of research projects based on this method. For example, during the last few years, blogs have been almost abandoned by users who joined social network sites such as MySpace and Facebook. Being aware of these rapid changes is essential to correctly design a research project based on this method. The process of methodological innovation should therefore be considered especially in these rapidly-changing times as an on-going process. References Bassett, E. H. and O Riordan, K. (2002) Ethics of internet research: contesting the human subjects model. Ethics and Information Technology, 4(1): Computer Ethics: Philosophical Enquiries (CEPE). Beaulieu, A. and Estalella, A. (2011) Rethinking Research Ethics for Mediated Settings. Information, Communication and Society, Routledge. boyd, d. (2008) Taken Out of Context: American Teen Sociality in Networked Publics. PhD Dissertation.University of California-Berkeley, School of Information. boyd, d., Golder, S. andlotan, G. (2010) Tweet, Tweet, Retweet: Conversational Aspects of Retweeting on Twitter. In HICSS-43. IEEE: Kauai, HI, USA. boyd, d.,and Hargittai E. (2010) Facebook Privacy Settings: Who Cares?, First Monday15 (8). boyd, d. and Crawford, K. (2011) Six Provocations for Big Data. SSRN elibrary.ssrn. doi: /ssrn Das, M. (2012) Innovation in online data collection for scientific research: the Dutch MESS project.methodological Innovations Online, 7(1) 7-24

11 Dutton, W.H.Jeffreys, P.W. and Goldin, I. (2010) World Wide Research: Reshaping the Sciences and Humanities, Cambridge, MA: MIT Press. Earle, P. (2010) Earthquake Twitter, Nature Geoscience, 3(4): Eco, U. (1993) Misreading. Boston, MA: Mariner Book. Edling, C. (2009) We Always Know More Than We Can Say: Mathematical Sociologists on Mathematical Sociology. In P. Hedstrom and B. Wittrock (Eds.), Frontiers of Sociology (pp ). Leiden: IDC Publisher. Fonio, C., Giglietto, F., Pedrioli, S., Rossi, L. and Pruno, R. (2007) Eyes on You. Narrating pregnancy in a networked space, presented at Toward a Social Science of Web2.0 conference, York, UK. Giglietto, F. and Rossi, L. (2006) Eyes on Europe, presented at IVSA Conference, Urbino Italy. Giglietto, F.and Rossi, L. (2009) Toward a bridge between sociocybernetics and internet studies. Journal of Sociocybernetics, 7(2): 1-3. Giglietto, F. (2009) Social semantics in a networked space. New perspectives for social science. In J. M. Aguado, B. Scott and E. Buchinger (Eds.), Technology and Social Complexity.Murcia: University of Murcia Press. Giglietto, F. (2012) If Likes Were Votes: An Empirical Study on the 2011 Italian Administrative Elections In International AAAI Conference on Weblogs and Social Media. Hendler, J., Shadbolt, N., Hall, W., Berners-Lee, T. and Weitzner, D. (2008) Web science, Communications of the ACM, 51(7): 60. Honeycutt, C. and Herring, S.C. (2009) Beyond Microblogging: Conversation and Collaboration via Twitter. In Proceedings of the Forty-Second Hawaii International Conference on System Sciences (HICSS-42), Los Alamitos, CA, USA: IEEE Computer Society: Kozinets R. V. (2012) Marketing Netnography: Prom/ot(ulgat)ing a New Research Method. Methodological Innovations Online, 7(1): Lazer, D., Pentland, A., Adamic, L., Aral, S., Barabasi, A., Brewer D., Christakis N., Contractor N., Fowler J., Gutmann M., Jebara T., King G., Macy M., Roy D. and Van Alstyne M. (2009) Social science. Computational social science.science (New York, N.Y.), 323(5915): Leydesdorff, L. (2000). A Sociological Theory of Communication, Parkland, FL: Universal Publishers. Magnani, M., Montesi, D. and Rossi, L. (2010) Friendfeed Breaking News: death of a public figure.in Second IEEE International Conference on Social Computing. LOS ALAMITOS, USA: IEEE computer Society, pp Manovich, L. (2008) Cultural Analytics: Analysis and Visualization of Large Cultural Data Sets. Screen, (February), pp.1-23.available at Manovich, L. (2011) Trending: The Promises and the Challenges of Big Social Data, Debates in the Digital Humanities, ed M.K.Gold. The University of Minnesota Press, Minneapolis, MN< [15 July 2011]. 35

12 Marwick, A.E. and boyd, d. (2010) I Tweet Honestly, I Tweet Passionately: Twitter Users, Context Collapse, and the Imagined Audience. New Media and Society. 13(1): Michel, J.-B., Shen, Y. K., Aiden, A. P., Veres, A., Gray, M. K., Pickett, J. P., Hoiberg, D., Clancy, D., Norvig, P., Orwant, J., Pinker, S., Nowak, M. A. and Aiden, E. L. (2011) Quantitative analysis of culture using millions of digitized books. Science (New York, N.Y.), 331(6014): doi: /science Rossi, L. (2010) Media and Generation: how user generated content reshape generational identity in the Mass Media System. SociologiadellaComunicazione, 40: Rossi, L., Magnani M., Iadarola B., (2011) #rescatemineros: global media event in the microblogging age, presented at AOIR Association of Internet Researcher Conference, Seattle, USA. Turkle, S. (1995) Life on the screen: Identity in the age of the Internet. New York: Simon and Schuster. Wu, S., Hofman, J. M., Watts, D. J. and Mason, W. A., (2010) Who Says What to Whom on Twitter Categories and Subject Descriptors. In ACM WWW 11. Biography Fabio Giglietto is PhD in Communication Studies from the Department of Communication Studies at the University "Carlo Bo" of Urbino where he also teaches Theory of Information in the Faculty of Social Science (B.A. Communication Sciences). Since 1997 is a member of LaRiCA (Research Laboratory on Advanced Communication). His main research interests is theory of information, communication and social systems with a specific focus on the relationship between social systems and new technologies. Since 2005 he is also member of RC51 on Sociocybernetics, a special interest group within International Sociological Association. Luca Rossi is a postdoctoral research fellow at the Department of Communication Studies, University of Urbino Carlo Bo, Italy. He graduated in Sociology and received a PhD in Communication Studies at the University of Urbino Carlo Bo. His main research interests are on social impact of digital technologies and new communicative practices. He did on User Generated Contents, Online Multiplayer Role Playing Games and Social Network Site. His current researches are focused on new methodological innovation for Sociological research in Social Network Sites. He is currently part of the Special Interest Group in Social Network Analysis (SIGSNA) a joint research project of Department of Communication Studies of the University of Urbino Carlo Bo and Department of Computer Science of the University of Bologna Please note that danah m. boyd, for personal and political reasons explained at spells her name using lowercase letters. 2 The results of this research are discussed in detail in Magnani, Montesi and Rossi (2010). 3 Archive available at