PS-15: Divides (Chair: Christoph Lutz)
The Empowered Refugee: The Smartphone as a Tool of Resistance on the Journey to Europe – Katja Kaufmann
Social media, participation, peer pressure, and the European refugee crisis: a force awakens? – Nils Gustafsson, Lund university, Sweden
My paper is about receiving/host nations. Sweden took in 160,000 refugees during the crisis in 2015. I wanted to look at this as it was a strange time to live in. A lot of people started coming in late summer and early autumn… Numbers were rising. At first response was quite enthusiastic and welcoming in host populations in Germany, Austria, Sweden. But as it became more difficult to cope with larger groups of people, there were changes and organising to address challenge.
And the organisation will remind you of Alexander (??) on the “logic of collective action” – where groups organise around shared ideas that can be joined, ideas, almost a brand, e.g. “refugees welcome”. And there were strange collaborations between government, NGOs, and then these ad hoc networks. But there was also a boom and bust aspect here… In Sweden there were statements about opening hearts, of not shutting borders… But people kept coming through autumn and winter… By December Denmark, Sweden, etc. did a 180 degree turn, closing borders. There were border controls between Denmark and Sweden for the first time in 60 years. And that shift had popular support. And I was intrigued about this. And this work is all part of a longer 3 year project on young people in Sweden and their political engagement – how they choose to engage, how they respond to each other. We draw on Bennett & Segerberg (2013), social participation, social psychology, and the notion of “latent participation” – where people are waiting to engage so just need asking to mobilise.
So, this is work in progress and I don’t know where it will go… But I’ll share what I have so far. And I tried to focus on recruitment – I am interested in when young people are recruited into action by their peers. I am interested in peer pressure here – friends encouraging behaviours, particularly important given that we develop values as young people that have lasting impacts. But also information sharing through young people’s networks…
So, as part of the larger project, we have a survey, so we added some specific questions about the refugee crisis to that. So we asked, “you remember the refugee crisis, did you discuss it with your friends?” – 93.5% had, and this was not surprising as it is a major issue. When we asked if they had discussed it on social media it was around 33.3% – much lower perhaps due to controversy of subject matter, but this number was also similar to those in the 16-25 year old age group.
We also asked whether they did “work” around the refugee crisis – volunteering or work for NGOs, traditional organisations. Around 13.8% had. We also asked about work with non-traditional organisations and 26% said that they had (and in 16-25% age group, it was 29.6%), which seems high – but we have nothing to compare this too.
Colleagues and I looked at Facebook refugee groups in Sweden – those that were open – and I looked at and scraped these (n=67) and I coded these as being either set up as groups by NGOs, churches, mosques, traditional organisations, or whether they were networks… Looking across autumn and winter of 2015 the posts to these groups looked consistent across traditional groups, but there was a major spike from the networks around the crisis.
We have also been conducting interviews in Malmo, with 16-19 and 19-25 year olds. They commented on media coverage, and the degree to which the media influences them, even with social media. Many commented on volunteering at the central station, receiving refugees. Some felt it was inspiring to share stories, but others talked about their peers doing it as part of peer pressure, and critical commenting about “bragging” in Facebook posts. Then as the mood changed, the young people talked about going to the central station being less inviting, on fewer Facebook posts… about feeling that “maybe it’s ok then”. One of our participants was from a refugee background and ;;;***
Q1) I think you should focus on where interest drops off – there is a real lack of research there. But on the discussion question, I wasn’t surprised that only 30% discussed the crisis there really.
A1) I wasn’t too surprised either here as people tend to be happier to let others engage in the discussion, and to stand back from posting on social media themselves on these sorts of issues.
Q2) I am from Finland, and we also helped in the crisis, but I am intrigued at the degree of public turnaround as it hasn’t shifted like that in Finland.
A2) Yeah, I don’t know… The middleground changed. Maybe something Swedish about it… But also perhaps to do with the numbers…
Q2) I wonder… There was already a strong anti-immigrant movement from 2008, I wonder if it didn’t shift in the same way.
A2) Yes, I think that probably is fair, but I think how the Finnish media treated the crisis would also have played a role here too.
An interrupted history of digital divides – Bianca Christin Reisdorf, Whisnu Triwibowo, Michael Nelson, William Dutton, Michigan State University, United States of America
I am going to switch gears a bit with some more theoretical work. We have been researching internet use and how it changes over time – from a period where there was very little knowledge of or use of the internet to the present day. And I’ll give some background than talk about survey data – but that is an issue of itself… I’ll be talking about quantitative survey data as it’s hard to find systematic collection of qualitative research instruments that I could use in my work.
So we have been asking about internet use for over 20 years… And right now I have data from Michigan, the UK, and the US… I have also just received further data from South Africa (this week!).
When we think about Digital Inequality the idea of the digital divide emerged in the late 1990s – there was government interest, data collection, academic work. This was largely about the haves vs. have-nots; on vs. off. And we saw a move to digital inequalities (Hargittai) in the early 2000s… Then it went quite aside from work from Neil Selwyn in the UK, from Helsper and Livingstone… But the discussion has moved onto skills…
Policy wise we have also seen a shift… Lots of policies around digital divide up to around 2002, then a real pause as there was an assumption that problems would be solved. Then, in the US at least, Obama refocused on that divide from 2009.
So, I have been looking at data from questionnaires from Michigan State of the State Survey (1997-2016); questionnaires from digital future survey in the US (2000, 2002, 2003, 2014); questionnaires from the Oxford Internet Surveys in the UK (2003, 2005, 2007, 2009, 2013); Hungarian World Internet Project (2009); South African World Internet Project (2012).
Across these data sets we have looked at questionnaires and frequency of use of particular questions here on use, on lack of use, etc. When internet penetration was less high there was a lot of explanation in questions, but we have shifted away from that, so that we assume that people understand that… And we’ve never returned to that. We’ve shifted to devices questions, but we don’t ask other than that. We asked about number of hours online… But that increasingly made less sense, we do that less as it is essentially “all day” – shifting to how frequently they go online though.
Now the State of the State Survey in Michigan is different from the other data here – all the others are World Internet Project surveys but SOSS is not looking at the same areas as not interent researchers neccassarily. In Hungary (2009 data) similar patterns of question use emerged, but particular focus on mobile use. But the South African questionnaire was very different – they ask how many people in the household is using the internet – we ask about the individual but not others in the house, or others coming to the house. South Africa has around 40% penetration of internet connection (at least in 2012 when we have data here), that is a very different context. There they ask for lack of access and use, and the reasons for that. We ask about use/non-use rather than reasons.
So there is this gap in the literature, there is a need for quantitative and qualitative methods here. We also need to understand that we need to consider other factors here, particularly technology itself being a moving target – in South Africa they ask about internet use and also Facebook – people don’t always identify Facebook as internet use. Indeed so many devices are connected – maybe we need
Q1) I have a question about the questionnaires – do any ask about costs? I was in Peru and lack of connections, but phones often offer free WhatsApp and free Pokemon Go.
A1) Only the South African one asks that… It’s a great question though…
Q2) You can get Pew questionnaires and also Ofcom questionnaires from their website. And you can contact the World Internet Project directly… And there is an issue with people not knowing if they are on the internet or not – increasingly you ask a battery of questions… and then filtering on that – e.g. if you use email you get counted as an internet user.
A2) I have done that… Trying to locate those questionnaires isn’t always proving that straightforward.
Q3) In terms of instruments – maybe there is a need to developmore nuanced questionnaires there.
Levelling the socio-economic playing field with the Internet? A case study in how (not) to help disadvantaged young people thrive online – Huw Crighton Davies, Rebecca Eynon, Sarah Wilkin, Oxford Internet Institute, United Kingdom
This is about a scheme called the “Home Access Scheme” and I’m going to talk about why we could not make it work. The origins here was a city council’s initiative – they came to us. DCLG (2016) data showed 20-30% of the population were below the poverty line, and we new around 7-8% locally had no internet access (known through survey responses). And the players here were researchers, local government, schools, and also an (unnamed) ISP.
The aim of the scheme was to raise attainment in GCSEs, to build confidence, and to improve employability skills. The Schools had a responsibility to identify students in need at school, to procure laptops, memory sticks and software, provide regular, structured in-school pastoral skills and opportunities – not just in computing class. The ISP was to provide set up help, technical support, free internet connections for 2 years.
This scheme has been running two years, so where are we? Well we’ve had successes: preventing arguments and conflict; helped with schoolwork, job hunting; saved money; and improved access to essential services – this is partly as cost cutting by local authorities have moved transactions online like bidding for council housing, repeat prescription etc. There was also some intergenerational bonding as families shared interests. Families commented on the success and opportunities.
We did 25 interiews, 84 1-1 sessions in schools, 3 group workshops, 17 ethnographic visits, plus many more informal meet ups. So we have lots of data about these families, their context, their lives. But…
Only three families had consistent internet access throughout. Only 8 families are still in the programme. It fell apart… Why?
Some schools were so nervous about use that they filtered and locked down their laptops. One school used the scheme money to buy teacher laptops, gave students old laptops instead. Technical support was low priority. Lead teachers left/delegated/didn’t answer emails. Very narrow use of digital technology. No in-house skills training. Very little cross-curriculum integration. Lack of ICT classes after year 11. And no matter how often we asked about it we got no data from schools.
The ISP didn’t set up collections, didn’t support the families, didn’t do what they had agreed to. They tried to bill families and one was threatened with debt collectors!
So, how did this happen? Well maybe these are neoliberalist currents? I use that term cautiously but… We can offer an emergent definition of neoliberalism from this experience.
There is a neoliberalist disfigurement of schools: teachers under intense pressue to meet auditable targets; the scheme’s students subject to a range of targets used to problematise a school’s performance – exclusions, attendance, C grades; the scheme shuffled down priorities; ICT not deemed academic enough under Govian school changes; and learning is stribbed back to narrow range of subjects and focus towards these targets.
There were effects of neoliberalism on the city council: targets and “more for less” culture; scheme disincentivised; erosion of authority of democratic institutional councils – schools beyond authority controls, and high turn over of staff.
There were neoliberalist practices at the ISP: commodifying philanthropy; couldn’t not treat families as customers. And there were dysfunctional mini-markets: they subcontracted delivery and set up; they subcontracted support; they charged for support and charged for internet even if they couldn’t help…
Q1) Is the problem digital divides but divides… Any attempt to overcome class separation and marketisation is working against the attempts to fix this issue here.
A1) We have a paper coming and yes, there were big issues here for policy and a need to be holistic… We found parents unable to attend parents evening due to shift work, and nothing in the school processes to accommodate this. And the measure of poverty for children is “free school meals” but many do not want to apply as it is stigmatising, and many don’t qualify even on very low incomes… That leads to children and parents being labelled disengaged or problematic
Q2) Isn’t the whole basis of this work neoliberal though?]
A2) I agree. We didn’t set the terms of this work..
Q1/comment) RSE and access
A1 – Huw) Other companies the same
Q2) Did the refugees in your work Katja have access to Sim cards and internet?
A2 – Katja) It was a challenge. Most downloaded maps and resources… And actually they preferred Apple to Android as the GPS is more accurate without an internet connection – that makes a big difference in the Aegean sea for instance. So refugees shared sim cards, used power banks for the energy.
Q3) I had a sort of reflection on Nils’ paper and where to take this next… It occurs to me that you have quite a few different arguements… You have this survey data, the interviews, and then a different sort of participation from the Facebook groups… I have students in Berlin here looking at the boom and bust – and I wondered about that Facebook group work being worth connecting up to that type of work – it seems quite separate to the youth participation section.
A3 – Nils) I wasn’t planning on talking about that, but yes.
Comment) I think there is a really interesting aspect of these campaigns and how they become part of social media and the everyday life online… The way they are becoming engaged… And the latent participation there…
Q3) I can totally see that, though challenging to cover in one article.
Q4) I think it might be interesting to talk to the people who created the surveys to understand motivations…
A4) Absolutely, that is one of the reasons I am so keen to hear about other surveys.
Q5) You said you were struggling to find qualitative data?
A5 – Katja) You can usually download quantitative instruments, but that is harder for qualitative instruments including questions and interview guides…
XP-02: Carnival of Privacy and Security Delights – Jason Edward Archer, Nathanael Edward Bassett, Peter Snyder, University of Illinois at Chicago, United States of America
Nathanial: We have prepared three interventions for you today and this is going to be kind of a gallery exploring space. And we are experimenting with wearables…
Fitbits on a Hamster Wheel and Other Oddities, oh my!
Nathanial: I have been wearing a FitBit this week… but these aren’t new ideas… People used to have beads for counting, there are self-training books for wrestling published in the 16th Century. Pedometers were conceived of in Leonardo di Vinci’s drawings… These devices are old, and tie into ideas of posture, and mastering control of physical selves… And we see the pedometer being connected with regimes of fitness – like the Manpo-Meter (“10,000 steps meter) (1965). This narrative takes us to the 1970s running boom and the idea of recreational discipline. And now the world of smart devices… Wearables are taking us to biometric analysis as a mental model (Neff – preprint).
So, these are ways to track, but what happens with insurance companies, with those monitoring you. At Oriel Roberts university students have to track their fitness as part of their role as students. What does that mean? I encourage you all to check out “unfitbit” – interventions to undermine tracking. Or we could, rather than going to the gym with a FitBit, give it to Terry Crews – he’s going anyway! – and he could earn money… Are fitness slaves in our future?
So, use my FitBit – it’s on my account
And so, that’s the first part of our session…
?: Now, you might like to hear about the challenges of running this session… We had to think about how to make things uncomfortable… But then how do you get people to take part… We considered a man-in-the-middle site that was ethically far too problematic! And no-one was comfortable participating in that way… Certainly raising the privacy and security issue… But as we talk of data as a proxy for us… As internet researchers a lot of us are more aware of privacy and security issues than the general population, particularly around metadata. But this would have been one day… I was curious if people might have faked your data for that one day capture…
Nathanial: And the other issue is why we are so much more comfortable sharing information with FitBit, and other sharing platforms, faceless entities versus people you meet at a conference… And we didn’t think about a gender aspect here… We are three white guys here and we are less sensitive to that being publicised rather than privatised. Men talk about how much they can benchpress… but personal metadata can make you feel under scrutiny
Me: I wouldn’t want to share my data and personal data collection tools…
Borrowing laptop vs borrowing phone…
?: In the US there have been a few cases where FitBits have been submitted as evidence in court… But that data is easier to fake… In one case a woman claimed to have been raped, and they used her FitBit to suggest that
Nathanial: You talked about not being comfortable handing someone your phone… It is really this blackbox… Is it a wearable? It has all that stuff, but you wear it on your body…
??: On cellphones there is FOMO – Fear Of Missing Out… What you might mix…
Me: Device as security
Comment: Ableism embedded in devices… I am a cancer survivor and I first used step counts as part of a research project on chemotherapy and activity… When I see a low step day on my phone now… I can feel this stress of those triggers on someone going through that stress…
Nathanial: FitBit’s vibrate when you have/have not done a number of steps… Trying to put you in an ideological state apparatus…
Jh: That nudge… That can be good for able bodied… But if you can’t move that is a very different experience… How does that add to their stress load.
Again looking at the condition of virtuality – Hayles 2006(?)
Vision is constructed… Thinking of higher resolution… From small phone to big phone… Lower resolution to higher resolution TV… We have spectacles, quizzing glasses and monocles… And there is the strange idea of training ourselves to see better (William Horation Bates, 1920s)… And emotional state interfering with how you do something… Rgeb we have optomitry and x-rays as a concept of seeing what could not be seen before… And you have special goggles and helmets… LIke the idea of the Image Accumulator in Videodrome (1985?), or the idea of the Memory recorder and playback device in Brainstorm (1983). We see embodied work stations – Da Vinci Surgery Robot (2000) – divorcing what is seen, from what is in front of them…
There are also playful ideas: binocular football; the Decelerator Helmet; Meta-perceptional Helmet (Cleary and Donnelly 2014); and most recently Google Glass – what is there and also extra layers… Finally we have Oculus Rift and VR devices – seeing something else entirely… We can divorce what we see from what we are perceiving… We want to swap people’s vision…
1. Raise awareness about the complexity of electronic privacy and security issues.
2. Identify potential gaps in the research agenda through playful interventions, subversions, and moments of the absurd.
3. Be weird, have fun!
“Cell phones are tracking devices that make phonecalls” (Applebaum, 2012)
I am interested in IMSI catcher which masquerades as a wireless base station, prompting phones to communicate with it. They are used by police, law inforcement, etc. They can be small and handheld, or they can be drone mounted. And they can track people, people in crowds, etc. There is always a different way to use it – you can scan for people in crowds. So if you know someone is there you can scan for it in a different way. So, these tools are simple and disruptive and problematic, especially in activism contexts.
But these tools are also capable of caturing transmitted content, and all the data in your phone. These devices are problematic and have raised all sorts of issues about their use, who and how you use them. I’d like to think of this a different way… Is there a right to protest? And to protest anonymously? We do have anti-masking laws in some places – that suggests no right to anonymous protest. But that’s still a different privacy right – covering my face is different from participating at all…
Protests are generally about a minority persuading a majoruty about some sort of change. There is no legal rights to protest anonymously, but there are lots of protected anoymous spaces. So, in the 19th century there was big debate on whether or not the voting ballot should be anonymous – democracy is really the C19th killer app. So there is a lovely quote here about the “The Australian system” by Bernheim (1889) and the introduction of anonymous voting. It wasn’t brought in to preserve privacy. At the time politicians brought votes – buying a keg of beer or whatever – and anonymity was there to stop that, not to preserve individual privacy. But Jill LePore (2008) writes about how our forebears considered casting a “secret ballot” to be “cowardly, underhanded and dispicable”.
So, back to these devices… There can be an idea that “if you have nothing to fear, you have nothing to hide”, but many of us understand that it is not true. And this type of device silences uncomfortable discourse.
Mathias Klang, University of Massachusetts Boston
Q1) How do you think that these devices fit into the move to allow law inforcement to block/”switch off” the camera on protestors/individuals’ phones?
A1) Well people can resist these surveillance efforts, and you will see subversive moves. People can cover cameras, conceal devices etc. But with these devices it may be that the phone becomes unusable, requiring protestors to disable phones or leave phones at home… And phones are really popular and well used for coordinating protests
Bryce Newell, Tilburg Institute for Law, Technology, and Society
I have been working on research in Washington Stat, working with law enforcement on license plate recognition systems and public disclosure law. And looking at what you can tell. So, here is a map of license plate data from Seattle, showing vehicle activity. In Minneapolis similar data being released led to mapping of the governer’s registered vehicles..
The second area is about law enforcement and body cameras. Several years ago peaceful protestors at UC Davis were pepper sprayed. Even in the cropped version of that image you can see a vast number of phones out, recording the event. And indeed there are a range of police surveillance apps that allow you to capture police encounters without that being visible on the phone, including: ACLU Police Tape, Stop and Frisk Watch; OpenWatch; CopRecorder2. And some of these apps upload the recording to the cloud right away to ensure capture. And there have certainly been a number of incidents from Rodney King to Oscar Grant (BART), Eric Garner, Ian Tomlinson, Michael Brown. Of these only the Michael Brown case featured law enforcement with bodycams. There has been a huge call for more cameras on law enforcement… During a training meeting some officers told me “Where’s the direct-to-YouTube button?” and “If citizens can do it, why can’t we also benefit from the ability to record in public places?”. There is a real awareness of control and of citizen videos. I also heard a lot of there being “a witch hunt about to begin…”.
So, I’m in the middle of focused coding on police attitudes to body cameras. Police are concerned that citizen video is edited, out of context, distorting. And they are concerned that it doesn’t show wider contexts – when recording starts, perspective, the wider scene, the fact that provocation occurs before filming usually. But there is also the issue of control, and immediate physical interaction, framing, disclosure, visibility – around their own safety, around how visible they are on the web. They don’t know why it is being recorded, where it will go…
There have been a number of regulatory responses to this challenge: (1) restrict collection – not many, usually budgetary and rarely on privacy; (2) restrict access – going back to the Minneapolis case, within two weeks of the map of governer vehicles being published in the paper they had an exemption to public disclosure law which is now permanent for this sort of data. In the North Carolina protests recently the call was “release the tapes” – and they released only some – then the cry was “release all the tapes”… But on 1st October law changed to again restrict access to this type of data.
But different state provide different access. Some provide access. In Oakland, California, data was released on how many license plates had been scanned. In Seattle data on scans can, because the data for many scans of one licence plates over 90 days is quite specific, you can almost figure out the householder. But granularity varies.
Now, we do see body cameras of sobriety tests, foot chases, and a half hour long interview with prostitute that discloses a lot of data. Washington shares a lot of video to YouTube. We see that in Rotterdam, Netherlands police doing this too.
But one patrol office told me that he would never give his information to an officer with a camera. Another noted that police choose when to start recording with little guidance on when and how to do this.
And we see a “collatoreal visibility” issue for police around these technologies.
Q1) Is there any process where police have to disclose that they are filming with a body cam?
A1) Interesting question… Initially they didn’t know. We used to have two party consent process – as for tapings – to ensure consent/implied consent. But the State attorney general described this as outside of that privacy regulation, saying that a conversation with a police officer is a public conversation. But police are starting to have policies that officers should disclose that they have cameras – partly as they hope and sometimes it may reduce violence to police.
Data Privacy in commercial users of municipal location data – Meg Young, University of Washington
My work looks at how companies use Seattle’s location data. I wanted to look at how data privacy is enacted by Seattle municipal government? And I am drawing on the work of Annemarie Mol and John Law (2004), an ethnographer working on health, that focuses on the lived experience. My data is drawing on ethnographic as as well as focus groups, interviews with municipal government and local civic technology communities. I really wanted to present the role of commercial actors in data privacy in city government.
We know that cities collect location data to provide services, and so share it for third parties to do so. In Washinton we have a state freedom of information (FOI) law, which states “The people of this state do not yield their sovereignty to the government…”, making data requestable.
In Seattle the traffic data is collected by a company called Acyclica. The city is growing and the infrastructure is struggling, so they are gathering data to deal with this, to shape traffic signals. This is a large scale longitudinal data collection process. Acyclica are doing that with wi-fi sensors sniff MAC addresses, the location traces sent to Acyclica (MAC salted). The data is aggregated and sent to the city – they don’t see the detailed creepy tracking, but the company does. And this is where the FOI law comes in. The raw data is on the company side here. If the raw data was a public record, it would be requestable. The company becomes a shield for collecting sensitive data – it is proprietizing.
So you can collect data, have service needs met, but without it becoming public to you and I. But analysing the contract the terms do not preclude the resale of data – though a Seattle Dept. of Transport (DOT) worker notes that right now people trust companies more than government. Now I did ask about this data collection – not approved elsewhere – and was told that having wifi settings on in public making you open to data collection – as it is in public space.
My next example is the data from parking meters/pay stations. This shows only the start, end, no credit card #s etc. The DOT is happy to make this available via public records requests. But you can track each individual, and they are using this data to model parking needs.
The third example is the Open Data Portal for Seattle. They pay Socrata to host that public-facing data portal. They also sell access to cleaned, aggregated data to companies through a separate API called the Open Data Network. The Seattle Open Data Manager didn’t see this situation as different from any other reseller. But there is little thought about third party data users – they rarely come up in converations – who may combine this data with other data sets for data analysis.
So, in summary, municipal government data is no less by and for commercial actors as it is the public. Proprietary protections around data are a strategy for protecting sensitive data. Government transfers data to third party
Q1) Seattle has a wifi for all programme
A1) Promisingly this data isn’t being held side by side… But the routers that we connect to collect so much data… Seeing an Oracle database of the websites fokls
Q2) What are you policy recommendations based on your work?
A2) We would recommend licensing data with some restrictions on use, so that if the data is used inappropriately their use could be cut off…
Q2) So activists could be blocked by that recommendation?
A2) That is a tension… Activists are keen for no licensing here for that reason… It is challenging, particularly when data brokers can do problematic profiling…
Q2) But that restricts activists from questioning the state as well.
Response – Sandra Braman
I think that these presentations highlight many of the issues that raise questions about values we hold as key as humans. And I want to start from an aggressive position, thinking about how and why you might effectively be an activist in this sort of environment. And I want to say that any concerns about algorithmically driven processes should be evaluated in the same way as we would social process. So, for instance we need think about how the press and media interrogate data and politicians
? “Decoding the social” (coming soon) is looking at social data and analysis of social data in the context of big data. She argues that social life is too big and complex than predicatable data. Everything that people who use big data “do” to understand patterns, are things that activists can do too. We can be just as sophisticated as corporations.
The two things I am thinking about are how to mask the local, and how to use the local… When I talk of masking the local I look back to work I did several years back on local broadcasting. There is mammoth literature on TV as locale, and production and how that is separate, misrepresenting, and the assumptions versus the actual information provided vs actual decision making. My perception is that social activism is that there is some brilliant activity taking place – brilliance at moments, specific apps often. And I think that if you look at the essays that Julian Assange before he founded WikiLeaks, particularly n weak links and how those work… He uses sophisticated social theory in a political manner.
But anonymity is practicably impossible… What can we learn from local broadcast? You can use phones in organised ways – there was training for phone cameras for the Battle of Seattle for instance. You can fight with indistinguishable actions – all doing the same things. Encryption is cat and mouse… Often we have activists presenting themselves as mice, although we did see an app discussed at the plenary on apps to alert you to protest and risk. And I have written before on tactical memory.
In terms of using the local… If you know you will be sensed all the time, there are things you can do as an activist to use that. It is useful to think about how we can conceive of ourselves as activists as part of the network. And I was inspired by US libel laws – if a journalist has transmission/recording devices but are a neutral observer, you are not “repeating” the libel and can share that footage. That goes back to 1970s law, but that can be useful to us.
We are at risk of being censored, but that means that you have choices about what to share, being deliberate in giving signals. We have witnessing, which can be taken as a serious commitment. That can happen with people with phones, you can train witnessing. There are many moments were leakage can be an opportunity – maybe not with volume or content of Snowden, but we can do that. There are also ways to learn and shape learning. But we can also be routers, and be critically engaged in that – what we share, the acceptable error rate. National Security are concerned about where in the stream they should target the misinformation – activists can adopt that too. The server functions – see my strategic memory piece. We certainly have community-based wifi, MESH networks, and that is useful politically and socially. We have responsibilities to build the public that is appropriate, and the networking infrastructure that enables those freedom. We can use more computational power to resolve issues. Information can be an enabler as well as influencing your own activism. Thank you to Anne and her group in Amsterdam for triggering thinking here, but big data we should be engaging critically. If you can’t make decisions in some way, there’s no point to doing it.
I think there needs to be more robustness in managing and working with data. If you go far then you need a very high level of methodological trust. Information has to stand up in court, to respect activist contributions to data. Use as your standard, what would be acceptable in court. And in a Panspectrum (not Panopticon) environment, when data is collected all the times, you absolutely have to ask the right questions.
Q1) I was really interested in that idea of witnessing as being part of being a modern digital citizens… Is there more on protections or on that which you can say
A1 – Sandra) We’ve seen all protections for whistle blowing in government disappear under Bush (II)… We still have protections for private sector whistle blowers. But there would be an interesting research project in there…
Q2) I wondered about that idea of cat and mouse use of technology… Isn’t that potentially making access a matter of securitisation…?
A2) I don’t think that “securitisation” makes you a military force… One thing I forgot to say was about network relations… If a system is interacting with another system – the principle of requisite variety – they have to be as complex as the system you are dealing with. You have to be at least as sophisticated as the other guy…
Q3) For Bryce and Meg, there are so many tensions over when data should be public and when it should be private, and tensions there… And police desires to show the good things they do. Also Meg, this idea of privatising data to ensure privacy of data – it’s problematic for us to collect data, but now a third party can do that.
A3 – Bryce) One thing I didn’t explain well enough is that video online comes from police, and from activists – it depends on the video here. Some videos are accessed via public records requests and published to YouTube channel – in fact in Washington you can make requests for free and you can do it anonymously. Police department does public video. Whilst they did a pilot in 2014 they had a hackathon to consider how to deal with redaction issues… detect faces, blur them, etc.. And proactive posting of – only some – video. The narrative of sharing everything, but that isn’t the case. The rhetoric has been about being open, by privacy rights and the new police chief. A lot of it was administrative cost concerns… In the hackathon they asked if posting in a blurred form, it would do away with blanket requests to focus requests. At that time they dealt with all requests for email. They were receiving so many emails and under state law they had to give up all the data and for free. But state law varies, in Charlotte they gave up less data. In some states there is a a differnet approach with press conferences, narratives around the footage as they release parts of videos…
A3 – Meg) The city has worked on how to release data… They have a privacy screening process. They try to provide data in a way that is embedded. They still have a hard core central value that any public record is requestable. Collection limitation is an important and essential part of what cities should be doing… In a way private companies collecting data results in large data sets that will end up insecure in those data sets… Going back to what Bryce was saying, the bodycam initiative was really controversial… There was so much footage and unclear what should be public and when… And the faultlines have been pretty deep. We have the Coalition for Open Government advocates for full access, the ACLU worried that these become surveillance cameras… This was really contentious… They passed a version of a compromise but the bottom line is that the PRA is still a core value for the state.
A3 – Bryce) Much of the ACLU, nationally certainly, was to support bodycams, but individuals and local ACLUs change and vary… They were very pro, then backing off, then local variance… It’s a very different picture hence that variance.
Q4) For Matthias, you talked about anti-masking laws. Are there cases where people have been brought in for jamming signals under that law.
A4 – Matthias) Right now the American cases is looking for keywords – manufacturers of devices, the ways data is discussed. I haven’t seen cases like that, but perhaps it is too new… I am a Swedish lawyer and that jamming would be illegal in protest…
A4 – Sandra) Would that be under antimasking or under jamming law.
A4 – Matthias) It would be under hacking laws…
Q4) If you counter with information… But not if switching phone off…
A4 – Matthias) That’s still allowed right now.
Q5) Do you do work comparing US and UK bodycamera?
A5 – Bryce) I don’t but I have come across the Rotterdam footage. One of my colleagues has looked at this… The impetus for adoption in the Netherlands has been different. In the US it is transparancy, in the Netherlands it was protection of public servants as the narrative. A number of co-authors have just published recently on the use of cameras and how they may increase assault on officers… Seeing some counter-intuitive results… But the why question is interesting.
Comment) Is there any aspect of cameras being used in higher risk areas that makes that more likely perhaps?
A5 – Sandra) It’s the YouTube on-air question – everyone imagines themselves on air.
Q6) Two speakers quoted individuals accused of serious sexual assault… And I was wondering how we account for the fact that activists are not homogenous here… Particularly when tech activists are often white males, they can be problematic…
A6) Techies don’t tend to be the most politically correct people – to generalise a great deal…
A6 – Sandra) I think they are separate issues, if I didn’t engage with people whose behaviour is problematic it would be hard to do any job at all. Those things have to be fought, but as a woman you should also challenge and call those white male activists on their actions.
Q7 – me) I was wondering about the retention of data. In Europe there is a lot of use of CCTV and the model there is record everything, and retain any incident. In the US CCTV is not in widespread use I think and the bodycam model is record incidents in progress only… So I was wondering about that choice in practice and about the retention of those videos and the data after capture.
A7 – Bryce) The ACLU has looked at retention of data. It is a state based issue. In Washington there are mandatory minimu periods… They are interesting as due to findings in conduct they are under requirements to keep everything for as long as possible so auditors from DOJ can access and audit. Bellingham and Spokane, officers can flag items, and supervisors can… And that is what dictates retention schedule. There are issues there of course. Default when I was there was 2 years. If it is publicly available and hits YouTube then that will be far more long lasting, can pop up again… Perpetual memory there… So actual retention schedule won’t matter.
A7 – Sandra) A small follow up – you may have answered with that metadata… Do they treat bodycam data like other types of police data, or is it a separate class of data?
A7 – Bryce) Generally it is being thought of as data collection… And there is no difference from public disclosure, but they are really worried about public access. And how they share that with prosecutors… They could share on DVD… And wanted to use share function of software… But they didn’t want emails to be publicly disclosable with that link… So being thought about as like email.
Q8 – Sandra) On behalf of colleagues working on visual evidence in course.
Comment – Micheal) There is work on video and how it can be perceived as “truth” without awareness of potential for manipulation.
A8 – Bryce) One of the interesting things in Bellingham was release of that video I showed of a suspect running away… The footage was following a police pick up for suspected drug dealing but the footage showed evasion of arrest and the whole encounter… And in that case, whether or not he was guilty of the drug charge, that video told a story of the encounter. In preparing for the court case the police shared the video with his defence team and almost immediately they entered a guilty plea in response to that… And I think we will see more of that kind of invisible use of footage that never goes to court.
And with that this session ends…
PA-31:Caught in a feedback loop? Algorithmic personalization and digital traces (Chair: Katrin Weller)
1Hans Bredow Institute for Media Research; 2; 3Alexander von Humboldt Institute for Internet and Society; 4GESIS Leibniz Institute for the Social Sciences
?? – Marco T Bastos, University of California, Davis and Cornelius Puschmann, Alexander von Humboldt Institute for Internet and Society
Marco: This is a long-running project that Cornelius and I have been working on. At the time we started, in 2012, it wasn’t clear what impact social media might have on the filtering of news, but they are now huge mediators of news and news content in Western countries.
Since then there is some challenge and conflict between journalists, news editors and audiences and that raises the issue of how to monitor and understand that through digital trace data. We want to think about which topics are emphasized by news editors, and which are most shared by social media, etc.
So we will talk about taking two weeks of content from the NYT and The Guardian across a range of social media sites – that’s work I’ve been doing. And Cornelius has tracked 1.5/4 years worth of content from four German newspapers (Suddeutsche Zeitung, Die Zeit, FAZ, Die Welt).
With the Guardian we accessed data from the API which tells you which articles were published in print, and which have not – that is baseline data for the emphasis editors place on different types of content.
So, I’ll talk about my data from the NY Times and the Guardian, from 2013, though we now have 2014 and 2015 data too. This data from two weeks is about 16k+ articles. The Guardian runs around 800 articles per day, the NYT does around 1000. And we could track the items on Twitter, Facebook, Google+, Delicious, Pinterest and Stumbleupon. We do that by grabbing the unique identifyer for the news article, then use the social media endpoints of social platforms to find sharing. But we had a challenge with Twitter – in 2014 they killed the end point we and others had been using to track sharing of URLs. The other sites are active, but relatively irrelevant in the sharing of news items! And there are considerable differences across the ecosystems, some of these social networks are not immediately identifiable as social networks – will Delicious or Pinterest impact popularity?
This data allows us to contrast the differences in topics identified by news editors and social media users.
So, looking at the NYT there is a lot of world news, local news, opinion. But looking at the range of articles Twitter maps relatively well (higher sharing of national news, opinion and technology news), but Facebook is really different – there is huge sharing of opinion, as people share what lies with their interests etc. We see outliers in every section – some articles skew the data here.
If we look at everything that appeared in print, we can look at a horrible diagram that shows all shares… When you look here you see how big Pinterest is, but in fashion in lifestyle areas. The sharing there doesn’t reflect ratio of articles published really though. Google+ has sharing in science and technology in the Guardian, in environment, jobs, local news, opinion and technology in the NYT.
Interestingly news and sports, which are real staples of newspapers but barely feature here. Economics are even worse. Now the articles are english-speaking but they are available globally… But what about differences in Germany… Over to Cornelius…
Cornelius: So Marcos’ work is ahead of mine – he’s already published some of this work. But I have been applying his approach to German newspapers. I’ve been looking at usage metrics and how that relationship between audiences and publishers, and how that relationship changes over time.
So, I’ve looked at Facebook engagement with articles in four German newspapers. I have compared comments, likes and shares and how contribution varies… Opinion is important for newspapers but not necessarily where the action is. And I don’t think people share stories in some areas less – in economics they like and comment, but they don’t share. So interesting to think about the social perception of sharability.
So, a graph here of Die Zeit here shows articles published and the articles shared on Facebook… You see a real change in 2014 to greater numbers (in both). I have also looked at type of articles and print vs. web versions.
So, some observations: niche social networks (e.g. Pinterest) are more relevant to news sharing than expected. Reliance on FB at Die Zeit grew suddenly in 2014. Social nors of liking, sharing and discussing differ significantly across news desks. Some sections (e.g. sports) see a mismatch of importance and use versus liking and sharing.
In the future we want to look at temporal shifts in social media feedback and newspapers coverage. Monitoring
Q1) Have you accounted for the possibility of bots sharing content?
A1 – Marcus) No, we haven’t But we are looking across the board but we cannot account for that with the data we have.
Q2) How did you define or find out that an article was shared from the URLs
A2) Tricky… We wrote a script for parsing shortened URLs to check that.
A2 – Cornelius) Read Marco’s excellent documentation.
Q3) What do you make of how readers are engaging, what they like more, what they share more… and what influences that?
A3 – Cornelius) I think it is hard to judge. There are some indications, and have some idea of some functions that are marketed by the platforms being used in different ways… But wouldn’t want to speculate.
Twitter Friend Reportoires: Inferring sources of information management from digital traces – Jan-Hinri Schmidt; Lisa Merton, Wiebke Loosen, Uwe, Kartin?
Our starting point was to think about shifting the focus of Twitter Research. Many studies are on Twitter – explicitly or implicitly – as a broadcast paradigm, but we want to conceive of it as an information tool, and the concept of “Twitter Friend Reportoires” – using “Friend” in the Twitter terminology – someone I follow. We ware looking for patterns in composition of friend sets.
So we take a user, take their friends list, and compare to list of accounts identified previously. So our index has 7,528 Twitter account of media outlets (20.8%) of organisations (political parties, companies, civil society organisations (53.4%) and of individuals (politicians, celebrities and journalists, 25.8%) – all in Germany. We take our sample, compare with a relational table, and then to our master index. And if the account isn’t found in the master index, we can’t say anything about them yet.
To demonstrate the answers we can find with this approach…. We have looked at five different samples:
- Audience_TS – sample following PSB TV News
- Audience_SZ – sample following quality daily newspapers
- MdB – members of federal parliament
- BPK – political journalists registerd for the bundespressekonferenz
- Random – random sample of German Twitter users (via Axel Bruns)
We can look at the friends here, and we can categorise the account catagories. In our random sample 77.8% are not identifiable, 22.2% are in our index (around 13% are individual accounts). That is lower than the percentages of friends in our index for all other audiences – for MdB and BPK a high percentage of their friends are in our index. Across the groups there is less following of organisational accounts (in our index) – with the exception of the MdB and political parties. If we look at the media accounts we can see that with the two audience samples they have more following of media accounts than others, including MdB and BPK… When it comes to individual public figures in our indexes, celebrities are prominent for audiences, much less so for MdB and BPK, but MdB follow other politicians, and journalists tend to follow other politicians. And journalists do follow politicians, and politicians – to a less extent – follow journalists.
In terms of patterns of preference we can suggest a model of a fictional user to understand preference between our three categories (organisational account, media account, individual account). And we can use that profile example and compare with our own data, to see how others behaviours fit that typology. So, in our random sample over 30% (37,9%) didn’t follow any organisational accounts. Amongst MdB and BPK there is a real preference for individual accounts.
So, this is what we are measuring right now… I am still not quite happy yet. It is complex to explain, but hard to also show the detail behind that… We have 20 categories in our master index but only three are shown here… Some frequently asked questions that I will ask and answer based on previous talks…
- Around 40% identified accounts is not very must is it?
Yes and no! We have increased this over time. But initially we did not include international accounts, if we did that we’d increase share, especially with celebrities, also international media outlets. However, there is always a trade off, there will also be a long tail… And we are interested in specific categorisations and in public speakers as sources on Twitter.
- What does friending mean on Twitter anyway?
Good question! More qualitative research is needed to understand that – but there is some work on journalists (only). Maybe people friend people for information management reasons, reciprocity norms, public signal of connection, etc. And also how important are algorithmic recommendations in building your set of friends?
Q1 – me) I’m glad you raised the issue of recommendation algorithms – the celebrity issue you identified is something Twitter really pushes as a platform now. I was wondering though if you have been looking at how long the people you are looking at have been on Twitter – as behavioural norms
A1) It would be possible to collect it, but we don’t now. We do, for journalists and politicians we do gather list of friends of each month to get longitudinal idea of changes. Over a year, there haven’t been many changes yet…
Q2) Really interesting talk, could you go further with the reportoire? Could there be a discrepancy between the reportoire and their use in terms of retweeting, replying etc.
A2) We haven’t so far… Could see which types of tweets accounts are favouriting or retweeting – but we are not there yet.
Q3) A problem here…
A3) I am not completely happy to establish preference based on indexes… But not sure how else to do this, so maybe you can help me with it.
Analysing digital traces: The epistemological dimension of algorithms and (big) internet data – Katharine Kinder-Kuranda and Katrin Weller
Katherine: We are interested in the epistemiological aspects of algorithms, so how we research these. So, our research subjects are researchers themselves.
So we are seeing real focus on algorithms in Internet Research, and we need to understand the (hidden) influence of algorithms on all kinds of research, including researchers themselves. So we have researchers interested in algorithms… And in platforms, users and data… But all of these aspects are totally intertwined.
So lets take a Twitter profile… A user of Twitter gets recommendations of who to follow in a given moment of time, and they see newsfeeds at a given moment of time. That user has context that as a researcher I cannot see or interpret the impact of that context on the user’s choice of e.g. who they then follow.
So, algorithms observe, count, sort and rank information on the basis of a variety of different data sources – they are highly heterogeneous and transient. Online data can be user-generated content or activity, traces or location data from various internet platforms. That promises new possibilities, but also raises significant challenge, including because of its heterogeneity.
Social media data has uncertain origins, about users and their motivations; often uncertain provenance of the data. The “users that we see are not users” but highly structured profiles and the result of careful image-management. And we see renewed discussion of methods and epistemology, particularly within the social sciences, for instance suggestions include “messiness” (Knupf 2014), and ? (Kitchen 2012).
So, what does this mean for algorithms? Algorithms operate on an uncertain basis and present real challenges for internet research. So I’m going to now talk about work that Katrin and I did in a qualitative study of social media researchers (Kinder-Kurlanda and Weller 2014). We conducted interviews at conferences – highly varied – speaking to those working with data obtained from social media. There were 40 interviews in total and we focused on research data management.
We found that researchers found very individual ways to address epistemological challenges in order to realise the potential of this data for research. And there were three real concerns here: accessibility, methodology, research ethics.
- Data access and quality of research
Here there were challenges of data access, restrictions on privacy of social media data, technical skills; adjusting research questions due to data availability; struggle for data access often consumes much effort. Researchers talks about difficulty in finding publicatio outlets, recognition, jobs in the disciplinary “mainstream” – it is getting better but a big issue. There was also comment on this being a computer science dominated fields – which had highly formalised review processes, few high ranking conferences, and this enforces highly strategic planning of resources and research topics. So researchers attempts to acieve validity and good research quality are constrained. So, this is really challenging for researchers.
2. New Methodologies for “big data”
Methodologies in this research often defy traditional ways of achieveing research validity – through ensuring reproducability, sharing of data sets (ethically not possible). There is a need to find patterns in large data sets by analysis of keywords, or automated analysis. It is hard for others to understand process and validate it. Data sets cannot be shared…
3. Research ethics
There is a lack of users informed consent to studies based on online data (Hutton and Henderson 2015). There are ethical complexity. Data cannot really be anonymised…
So, how do algorithms influence our research data and what does this mean for researchers who want to learn something about the users? Algoritms influence what content users interact with, for example: How to study user networks without knowing the algorithms behind follower/friend suggestions? How to study populations?
To get back to the question of observing algorithms? Well the problem is that various actors in the most diverse situations react out of different interests to the results of algorithic calculations, and may even try to influence algorithms. You see that with tactics around trending hashtags as part of protest for instance. The results of algorithmic analyses presented to internet users with information on how algorithms take part.
In terms of next steps. researchers need to be aware that online environments are influenced by algorithms and so are the users and the data they leave behind. It may mean capturing the “look and feel” of the platform as part of research.
Q1) One thing I wasn’t sure about… Is your sense when you were interviewing researchers that they were unaware of algorithmic shaping… Or was it about not being sure how to capture that?
A1) Algorithms wasn’t the terminology when we started our work… They talked about big data… the framing and terminology is shifting… So we are adding the algorithms now… But we did find varying levels of understanding of platform function – some were very aware of platform dynamics, but some felt that if they have a Twitter dataset that’s a representation of the real world.
Q1) I would think that if we think about recognising how algorithms and platform function come in as an object… Presumably some working on interfaces were aware but others looking at, e.g. friendship group, took data and weren’t thinking about platform function, but that is something they should be thinking about…
Q2) What do you mean by the term “algorithm” now, and how that term is different from previously…
A2) I’m sure there is a messyness of this term. I do believe that looking at programmes, wouldn’t solve that problem. You have the algorithm in itself, gaining attention… From researchers and industry… So you have programmers tweaking algorithms here… as part of different structures and pressures and contexts… But algorithms are part of a lot of peoples’ everyday practice… It makes sense to focus on those.
Q3) You started at the beginning with an illustration of the researcher in the middle, then moved onto the agency of the user… And the changes to the analytical capacities working with this type of data… But how much is the awareness amongst researchers of how the data, the tools they work with, and how they are inscribed into the research…
A3) Thank you for making that distinction here. The problem in a way is that we saw what we might expect – highly varied awareness… This was determined by disciplinary background – whether STS researchers in sociology, or whether a computer scientist, say. We didn’t find too many disciplinary trends, but we looked across many disciplines…. But there were huge ranges of approach and attitude here – our data was too broad.
Q1 – Cornelius) I think that we should say that if you are wondering about “feedback” here, it’s about thinking about metrics and how they then feedback into practice, if there is a feedback loop… From very different perspectives… I would like to return to that – maybe next year when research has progressed. More qualitative understanding is needed. But a challenge is that stakeholder groups vary greatly… What if one finding doesn’t hold for other groups…
Q2) I am from the Wikimedia Foundation… I’m someone who does data analysis a lot. I am curious if in looking at these problems you have looked at recommender systems research which has been researching this space for 10 years, work on messy data and cleaning messy data… There are so many tiny differences that can really make a difference. I work on predictive algorithms, but that’s a new bit of turbulence in a turbulent sea… How much of this do you want to bring this space…
A2 – Katrin) These communities have not come together yet. I know people who work in socio-technical studies who do study interface changes… There is another community that is aware that this exists… And is not aware so closely… But see it as tiny bits of the same puzzle… And can be harder to understand for historical data… And getting an idea of what factors influence your data set. In our data sets we have interviewees more like you, and some with people at sessions like this… There is some connection, but not all of those areas coming together…
A2 – Cornelius) I think that there is a clash between computational social science data work, and this stuff here… That predictable aspect screws with big claims about society… Maybe an awareness but not a keenness. In terms of older computer science research that we are not engaging in, but should be… But often there is a conflict of interests sometimes… I saw a presentation that showed changes to the interface, changing behaviour… But companies don’t want to disclose that manipulation…
Comment) We’ve gone through a period, disheartened to see it is still there, that researchers are so excited to trace human activities, that they treat hashtags as the political debate… This community helpfully problematises or contextualises this… But I think that these papers are raising the question of people orientating practices towards the platform, from machine learning… I find it hard to talk about that… And how behaviour feeds into machine learn… Our system tips to behaviour, and technology shifts and reacts to that which is hard.
Q3) I wanted to agree with that idea of the need to document. But I want to push at your implicit position that this is messy and difficult and hard to measure… But I think that applies to *any* methods… Standards of data removal, arise elsewhere, messiness occurs elsewhere… Some of those issues apply across all kinds of research…
A3 – Cornelius) Christian would have had an example on his algorithm audit work that might have been helpful there.
Comment) I wanted to comment on social media research versus traditional social science research… We don’t have much power over our data set – that’s quite different in comparison with those running surveys, undertaking interviews… and I have control of that tool… And I think that argument isn’t just about survey analysis, but other qualitative analysis… Your research design can fit your purposes…
Twitter recommend algorithms, celebrities and noise. Time on twitter. Overall follower/following counts? Does friend suggest influence?
Advertistors? and role in shaping content in news