Student: JonneMarie Bouman Date of completion: 29/06/2015 Supervisor: mw. drs. E.J.T. Weltevrede Second reader: mw. dr. C. Gerlitz University and program: University of Amsterdam MA New Media and Digital Culture Word Count: 20.610
Abstract This thesis uses different digital methods to do map concerns and conversation surrounding drones, privacy and surveillance when it comes to drones in U.S. public airspace. Within a framework of ANT, an issue mapping using different affiliated organizations’ websites will focus on whether the issue is recognized by these affiliated parties. Twitter’s mediumspecific features (the #hashtag and the retweet) will also be used to find out whether the issue is a topic of conversation on the platform and what this conversation is about. Doing research using ANT and digital methods will be critically addressed, as well as whether the use of “popular content” on twitter which is demarcated by hashtags and retweets within a dataset subsample gives representative and satisfying results. Keywords Drone, privacy, ActorNetwork Theory, issue mapping, Twitter, digital methods Acknowledgements First, I would like to thank mw. drs. E.J.T. Weltevrede for her excellent supervision from the start of this thesis until the very end. I would also like to thank mw. dr. C. Gerlitz for helping me get started, and being the second reader and also Erik Borra for his help in using the DMITCAT. Thanks also goes to the DMI Amsterdam, for making their tools available for usage. And last but not least, I would like to thank Suzanne Tromp, Evelien Christiaanse, Jakub Dutka, Richelle Werners, Yvette Ducaneaux, Peter, Joke and Madelon Bouman for their help and support during this process.
Table of contents 1. Introduction and Research Question 3 2. Issue Mapping: Latour’s ANT, Venturini’s Magma and “The Social” 6 3. Drones: Concerns in the US 10 3.1 Drones: Technical Features, Rules & Legislation 12 3.2 A reasonable expectation of privacy, surveillance, and drones 15 4. Conceptualization: Mapping privacy and surveillancerelated Concerns on Drones 15 4.1 Comprising a list of actors 15 4.2 Tools : The Issue Crawler 15 4.3 Issue Crawler Settings 15 5. Analyzing Issue Crawler Results 18 6. Results and visualizations IssueCrawler 21 6.1 All URLS 22 6.2 EPIC 25 6.3 Federal Aviation Administration 29 6.4 Electronic Freedom Frontier 31 6.5 Model Aircraft Clubs Hobbyists 34 7. Issue Crawler Findings 35 8. Investigating the conversation on drones, privacy and surveillance on Twitter 37 9. The DMITCAT “drones” dataset subsample 41 9.1 Results most frequently used hashtags and most frequently retweeted tweets 45 10. DMITCAT findings 64 11. Discussion, limitations and ideas for further research 65 12. Conclusion 66 Bibliography Appendices Appendix I: Complete URL list Appendix II: Crawls and starting points that did not comprise a network Appendix III: Model Aviation Hobbyists networked URLs queried for “privacy” and “surveillance”
1. Introduction In 2014 Elsvette Buenaventura could not sleep due to the flickering lights outside her 10th story window in California. It turned out to be a drone, hovering outside of her window 4 times a week, after which she stated: ‘We don’t know what he was looking for with his camera drone, all we felt was a violation of our privacy’(Marois n. pag.). Imagine a machine hovering outside your window, watching you and capturing all the information about your whereabouts, gender, skin color, body temperature and height. Does it give you the chills? Does the idea of being watched (surveillance) and all this information about you being captured (privacy) concern you? Well, I think you are not the only one concerned about the capabilities of drones.
Drones are vehicles which can be controlled from a distance, and are therefore unmanned (“Unmanned Aerial Vehicles”, also UAVs). They have been around for over a century and the first documented drone use took place in Austria in 1849, when they released about 200 pilotless balloons with bombs to attack Venice city. After this, the drone was mainly used for military purposes, until in 1889 the first aerial reconnaissance photos were taken by a drone (Schlag n. pag). Fast forward to now, a time in which the development and use of drones is expanding so quickly, drones are no longer just being used by governments, law enforcement agencies or companies using the drones for research or aerial imagery. Anyone really anyone can buy a drone for less than a hundred U.S. Dollars. Which makes it possible for anyone able to buy a drone to watch you from outside your window. Surveillance and privacy are closely related, people can feel their privacy is being violated, but who decides what privacy is and what reasonable surveillance is? In the case of surveillance by drones, the FAA (Federal Aviation Administration) decides whether actions that could be deemed to have traits of surveillance like photography, sound and videorecording is allowed, by giving permission to those who according to them are allowed to control a drone. At the same time, the U.S. government decides on whether someone’s privacy is being infringed upon by the Fourth Amendment’s statement that one should have a “reasonable expectation of privacy”. If people feel their reasonable expectation of privacy is being violated, they are known to ‘accept, negotiate, or resist’ the manner they are being surveilled upon (Lyon 278). This thesis sets out to investigate whether concerns on drones, surveillance and privacy can be found by tracing the actors in an issuenetwork and investigating in which ways these actors share concern on the subject. As the actors that will be mapped first exist of organizations’ websites, a subsample of a Twitter dataset on drones will also be researched upon, to see whether actors engage in a conversation on drones by using the mediumspecific features hashtags(#) and retweets(RT). To do map an issuenetwork to see which actors share concerns on the subject, starting points have to be chosen. I have decided to chose three parties which can be considered as main parties with
shared concerns on the subject of drones, privacy and surveillance, based on a history of legal battles between the parties. In the U.S., these three main parties that are concerned with drones, privacy and surveillance are the Federal Aviation Administration (hence ‘FAA’), the Electronic Freedom Frontier (hence ‘EFF’) and the Electronic Privacy Information Centre (hence ‘EPIC’). The FAA is the main organization in the U.S. that is responsible for providing safe aerospace within U.S. borders, while EPIC and the EFF are advocacy groups concerned with Privacy, rights and freedom of speech. It all started in October 2012 when the EFF decided to file suit against the FAA, demanding more transparency in who were allowed to fly drones at the time, and why. EFF won the case and got insight in the drone licensing process, information which had never been released before. According to the EFF’s webpage on the case: ‘This prompted significant public awareness and discussion about the privacy and surveillance issues with drones’. EPIC reacted to this ordeal by filing suit against the FAA in March 2015, for not protecting citizens’ privacy like they should according to EPIC’s website: ‘As a consequence of the FAA’s failure to establish drone privacy rules, millions of Americans now face the possibility of unchecked monitoring and harassment’. Even though the above number might sound like an exaggeration, in 2012 the FAA complied to the Modernization and Reform Act of 2102, in which the FAA is issued to start integrating drones in national airspace by the end of 2015. This opens up U.S. public airspace for commercial and domestic drones, while they were previously exclusively used by governmental institutions and the U.S. military. The expectation is that by the year of 2020, an amount of 30.000 drones will cloud U.S. public airspace, which has caused concern for the privacy of civilians. As of now, the use of drones in US public airspace is according to the FAA still limited to about 300 companies or private agencies using them, as well as to topographical locations like airports, military (training) ground and urban centers (Lowy, n.pag). At the same time, the FAA seems to have a hard time keeping legislation uptodate while the use of drones keeps expanding rapidly, as for example hobbyist drone pilots do not need permission to use their drones within public spaces. So, while the FAA is having a hard time keeping up, concerns have been aired about what the rapid growth of drones will do to civilians’ “reasonable expectation of privacy”. In this study, I will focus on these privacy and surveillance related concerns when it comes to drones in U.S. public airspace. The overall framework that will be used is that of ActorNetwork Theory (Hence: ‘ANT’) which was coined by French Philosopher Bruno Latour (2005). The methods that are used will be based on Digital Methods as explained in Digital Methods (2013) by Richard Rogers, combined with tools created by the Digital Methods Initiative (henceforth DMI) These methods will be used to produce an issue mapping by using EFF’s, EPIC’s and the FAA’s websites and looking for other
related websites to map the concerns, using the DMI Issue Crawler . The second method makes use of the 1 DMITCAT to do a Twitter Analysis on “popular content” by analyzing hashtag frequency and identical tweet frequency (retweets) within a subsample of the dataset on drones that has captured tweets over a period of approximately 2 and a half years. These complementary methods will be used to answer the central question in this study: To what extent do actors’ websites and tweets related to drones in U.S. public airspace acknowledge concerns of privacy and surveillance? Because two completely different (but as will be explained later complementary) methods will be used to answer the above question, I have decided to split the question up into two subquestions related to the two different methods: 1) Issue mapping: Do the actors within the issuenetwork acknowledge and share concerns on drones in public U.S. airspace, surveillance and privacy? 2) DMITCAT analysis: To what extent do actors engage in a conversation about concerns on drones in public U.S. airspace, surveillance and privacy by using hashtags and retweets? These two methods will be used because they will not only show a difference in governmental and nongovernmental or commercial organizations acknowledging and engaging with the issue and ‘the masses’ recognizing the issue, they are also in need of using different digital methods: one which focusses on issue mapping using the DMI Issue Crawler and the other which focuses on mapping the issue filtering certain Twitter specific features in the DMITCAT. The combination of these two methods was done to do a more indepth analysis of what the concerns entail. While the Issue Crawler map will show different actors’ shared concerns (whether they are of a surveillance or privacy nature or not), the Twitter analysis by hashtags and retweets can be used to show the public’s concerns through “popular content”. These methods will be elaborated on further on in this thesis. First a description of ActorNetwork theory is given, which constitutes the framework for this research. This first chapter will also explain why this theory was chosen as a framework and what connects it to issue mapping. After this, a short explanation of drones will be given by looking at technological features which make them capable of surveillance. This section will be followed up by a chapter on literature explaining why drones would call for concerns in the U.S. and will engage the literature to the methods, after which the methods can be explained and conceptualized. Both methods will be structured by explaining them, visualizing the results and elaborating on the findings. The last chapter of this thesis will conclude by answering the aforementioned research questions. 1 Information on the Issue Crawler as well as the tool itsself can be found here: https://wiki.digitalmethods.net/Dmi/ToolIssueCrawler.
2. Issue Mapping: Latour’s ANT, Venturini’s Magma and “The Social”
Bruno Latour’s Reassembling the Social (2005) has often been referred to as “The Guide to Issue Mapping”(Rogers, Querubin, Kil 7). The ActorNetwork Theory (hence: ‘ANT’) is a theory on tracing associations that was coined by Bruno Latour in Reassembling the Social (2007). This guide to Issue Mapping will be used in combination with certain digital methods and digital methods tools in mapping actors’ concerns on drones, privacy and surveillance. Latour has coined a new meaning of what “the social” is. While the social has been seen as a presupposed structure by people calling something “social” like “society”, Latour proposes a different uptake: in stead of seeing “the social” as a set structure and focussing on the nodes within that structure, Latour states the researcher has to ‘trace associations’, which are traces left by the human or nonhuman actors within a network. This tracing of associations can be related to digital methods in the sense that mediumspecific features structuralize associations which can easily be retrieved, visualized and analyzed using digital methods tools. When looking at actors in digital methods, what would they be? Are the users of platforms seen as actors? Are mediumspecific features actors? Is the researcher an actor? One important statement that differentiates Latour from other social researchers is nonhuman actors are actors too, meaning animals, as well as ‘lifeless objects’ and technological actors such as platforms, accounts and digital tools also have agency. In the case of mapping an issue around affiliated actors’ websites concerning drones, the websites can be seen as the actors, as they are the ones leaving traces (in the shape of mediumspecific features) that can be analyzed. When using digital methods to Issue Crawler Twitter, the mediumspecific features of the platform which can be considered traceable movements are the actors being researched as they are the ones carrying meaning. They are actors because they act and possess agency, making them able to substantiate why they act the way they do. Latour differentiates between two types of actors which transport meaning in different ways: mediaries (transport information without transforming it) and intermediaries (transport and transform information). To answer ‘To what extent do actors’ websites and tweets related to drones in U.S. public airspace acknowledge concerns of privacy and surveillance?’, associations will be traced by tracing websites (mediaries), retweets (intermediaries), and hashtags (intermediaries). Websites are usually meant to provide users of information, they can be seen as mediaries . Retweets are intermediaries because the information is being transformed from the moment it is being retweeted, by giving it more attention and moving it around within a network. The same counts for hashtags, which are intermediaries for placing tweets in a topical category or bigger conversation by tagging a tweet. These mediumspecific features have made data on the web countable and has made the tracing of associations a quantitative as well as a
qualitative method. Though, what every researcher has to keep in mind when tracing associations is not interrupting the natural flow of the actors. One could argue the researcher is an intermediary actor in tracing associations, as the researcher inflicts on the data by filtering and thus transforming it in a specific way. Keeping a distance from bias is hard because of this carefully selected filtering process, as the researcher as an actor has agency to make certain decisions. I have decided to accept this faith, as long as my methodological traces stay traceable. Someone who would agree with this notion of the researcher being an actor is Tomasso Venturini. In his article Diving in Magma: How to Explore controversies with ActorNetwork Theory (2010) Venturini uses the metaphor of magma to explain the different states a controversy goes through: liquid, to rock, and back, and all these states of “magma” coexist. The researcher has to leave the actors uninterrupted but has to see up close which actors are involved and what their ideas are. Thus, the researcher has to dive in: Diving in magma! (264). According to Venturini, cartographers should be as curious and open to surprise as possible, but are unable to be unbiased: ‘You shall observe from as many viewpoints as possible’ (260) he states, to be the least biased. At the same time conclusions that are drawn from a mapping are always biased in a way. Venturini encourages methodological promiscuity by stating: ‘You shall not restrain your observation to any single theory or methodology’(260). This combination of theory and different methodologies also has to do with including as many actors as possible within a debate to give the least biased networked issueresults according to Venturini. This might be problematic combined with Digital Methods and especially in using the DMI Issue Crawler as the main thing about the tool seems to be Less is More, meaning the less starting points or “seeds” are being used, the less the network drifts from the issue. This statement will be challenged by using the Issue Crawler to do an exhaustive crawl, to find out by visualizing and analyzing the results whether less actually is more: the cartography of controversies. Venturini explains the cartography of controversies (also called “shared uncertainties”) as ‘a set of techniques to explore and visualize issues’(258), which as will be seen later on in this thesis can be done using digital methods. This shared uncertainty has to do with actors agreeing on disagreeing. The actors involved in the controversy can not ignore one another, but have to work out a compromise to live together in harmony. It is not the researcher’s job to solve the issue that might be apparent, the researcher is there to “just observe”, a notion by Latour which is critiqued by Venturini for being too vague. He explains the common notion of “just” has complicated controversy mapping for any researcher using ANT, as the notion of “just” sounds like it’s easy (259). While I agree on Venturini criticizing Latour for making controversy mapping sound easy, I do not think this is what Latour meant by “Just observing”. I think he meant the researcher should be unbiased, which as discussed earlier is inherently as well as
technically impossible as mediumspecific features have to be filtered and chosen by me for particular reasons to follow the actors’ tracks as discussed earlier. In ‘just observing’ the controversy that has formed around drones, surveillance and privacy, digital methods tools can be used to see which traces are left behind by which actors and what this states on their stance at a certain point in time without interrupting the actors’ natural flow. It is a given in science that the more people being questioned, empirical research being done and thus the more results a researcher has, the more balanced out the final conclusion will be. As, according to Venturini, the key to drawing an effective map is by drawing many maps, combining them and seeing whether they can be assembled to show the most effective result, using different methods to trace associations Twitter dataset analysis and Issue Crawler mapping and thereby combining different maps, their complementary features can show more associations than if just one was used (260). They might even show actors or concerns I as a researcher had not thought of before, which is something Venturini warns the researcher for, namely: disagreeing minorities bring a controversy into being (798). A researcher doing controversy mapping should have an endless amount of respect for the actors that are being researched. So, the issuefication of a subject takes place because situations or things that have been taken for granted for a long time are suddenly being questioned, in which all the actors should be questioned and the researcher should never exclude any actors. An example Venturini gives in the debate on climate change is: ‘Who, before global warming, ever thought that inuit communities or polar bears may have opinions on industrial strategies? Today we know that they have and that they should be listened to’ (262). This also means that tracing associations might show actors in the debate the researcher has not thought of himself, offering new associations that can be traced to Issue Crawler a network. While slowly pulling more actors into the debate by tracing associations as described above, what always has to be done is order and unfold complexity as stated by Venturini in his sequel “Building on Faults”(2010: 797). When mapping drones as a social controversy among the lines of the above theories, several questions must be answered prior to starting the methodological research: What makes drones suitable for surveillance and how might they inflict on privacy, thus: How are drones being issuefied? Using digital methods for cartography as will be done in this thesis is called neocartography. One article that inspired the method for this research was “Landscaping climate change: a mapping technique for understanding science and technology debates on the World Wide Web”(2000) by Richard Rogers and Noortje Marres, who explain neocartography and the purpose of it. Doing neocartography means adding a layer of online information to an onground debate using digital methods tools, this can be done by mapping mediumspecific features like mentions, tags, or hyperlinked websites. This type of
mapping is used for this thesis and is meant to as stated by Rogers and Marres ‘[...] provide a semblance of given socioepistemic networks on the web’ (1). A map of hyperlinks between organizations that may (or may not) leave traceable associations will be constructed using different digital methods tools. By linking to another organization's’ website affiliated with the debate, or by actually not linking to it, it might show which actors share the same concerns on drones, privacy and surveillance. The mapping of hyperlinked websites in this thesis will be done by using the DMI’s Issue Crawler which shows the linkage between organizations’ websites. The question that I will answer using these findings is: Do the actors within the issuenetwork acknowledge and share concerns on drones in public U.S. airspace, surveillance and privacy? Also, mediumspecific features of microblogging platform Twitter will be used to see what is being discussed about the subject on that platform. The most frequently used hashtags and identical retweets have been used to compose a dataset of the most popular content on the dates there was tweeted the most on the subject of drones. In this way the conversational aspect of Twitter can be put to use to find out what the actors (Twitterusers) discuss. The question I will answer using this method is: What does the most popular content within the drones dataset subsample say about possible concerns on privacy, surveillance and drones?
3. Drones: Concerns in the U.S. To make the issue mapping more comprehensible for those unfamiliar with the subject of drones, some background information has to be given on what “drones” actually are and what the technical capabilities are which make them sensitive subjects for criticism when it comes to the subject of privacy and surveillance. Drones were formerly mostly used by military and governmental agencies, but have recently become available to the public, posing threats to privacy of everyone in the USA through the hightech surveillance capabilities drones have (Cavoukian 2012). The move from drones of this type of respectively secretive surveillance to the anyone being able to use a drone in the public realm has raised questions among citizens, governments and privacy rights organizations. 3.1 Drones: Technical Features, Rules & Legislation As Cavoukian put it: ‘The word ‘surveillance’ comes from the French word for ‘watching over. ‘Sur’ means ‘from above’ and ‘veiller’ means ‘to watch. In this context, it is easy to appreciate that drone technology represents the “cold technological embodiment of observation”’(Cavoukain 41). Drones are Unmanned Aerial Vehicles (UAVs) and therefore the term ‘drone’ is interchangeably used with the term ‘UAV’. The FAA differentiates between three different types of droneflying: Public Operations (governmental), Civil Operations (nongovernmental), and Model Aircraft (Hobby or Recreation only). Civil operations, which according to the website are: ‘Any operation that does not meet the statutory criteria for a public aircraft operation is considered a civil aircraft operation and must be conducted in accordance with all FAA regulations applicable to the operation.’ Public Operations for this matter are defined on a flightbyflight basis, meaning the considerations are based on ownership of the aircraft, the operator, purpose of the flight and persons being on board of the aircraft inflight. When wanting to legally fly a drone and be able to use for example the imagery, a specific Certificate of Waiver or Authorization (COA) is issued by the FAA to give permission to public agencies and organizations if they wish to use an aircraft. As of now, about 350 public agencies have gotten FAA’s permission to fly drones for different purposes.To do so, a defined part of airspace can be used and might include specific safety rules, depending on the type of operation. These rules and regulations do not apply to users that fit in the category “hobbyists”, who can technologically do what the other users can, but are hobbyist because no money is involved. According to the FAA websites, these rules and regulations are set up to ensure UAVs do not bring danger to already existing aviation operations, meaning they are not allowed to be used in populated areas and the aircraft has to be observed by someone in another aircraft or from someone on ground. The
FAA states the following about contemporary public uses:’[...] include law enforcement, firefighting, border patrol, disaster relief, search and rescue, military training, and other government operational missions.’ But, then what makes them so different from other previously used aircrafts like helicopters and airplanes? The difference drones are always unmanned and are sometimes imperceptible to the ones being watched, while for example a helicopter is hard to ignore. They are usually lighter than other aircrafts because of their reduced size and weight, which is a result of the cockpit being on ground and allows them to fly closer to private spaces or even inside private spaces or as Cavoukain calls them “urban canyons”(6). It does not stop there, drones can even be equipped with a type of drone software that can do ‘soft biometric recognition’ which makes them able to recognize a person’s weight, height, gender and skin color. (BrackenRoche et al. 1819). This would not matter a great deal, except for most drones being equipped with camera technologies, which makes them able to record and transmit photo images to the ground control station. And because of the many technological developments in the field, drones have become cheaper and are therefore more available to the public, which has raised special concerns to privacy advocates as the FAA expects to have 30.000 drones up in public U.S. air by 2020. The availability and usability to the public are also made easier by the possibility to control the drone using an iPhone, iPad, or iPod Touch by generating their own WiFi network, and therefore allow connection to users’ devices but also become capable of hacking these networks giving them the nickname “flying laptops” (Cavoukian 19). ‘Miniature surveillance drones, unseen digital recognition systems, and surreptitious geolocational monitoring are readily available, making longterm surveillance relatively ‘easy and cheap’(Cavoukian 2324). What should be kept in mind is these drones are not specifically used with the purpose of surveillance, there are many types of uses which can differ from creative use (art, photography, video imagery) to disaster control (tornado research, fire control) to law enforcement practices (mapping and predicting criminal activities) (Cavoukian 2012). So, even though drones have the capabilities to do great things apart from their technological features making them suitable for surveillance, as is explained earlier using the lawsuits of the EFF and EPIC versus FAA for examples there seem to be concerns for privacy. These concerns will be further elaborated on in the next subchapter, which will focus on why the lack of clarity in U.S. law provides for concerns for privacy now drones are becoming these highly available surveillance tools.
3.2 A reasonable expectation of privacy, surveillance, and drones This subchapter connects the subjective claim of a reasonable expectation of privacy to surveillance and the panoptic theories which explain why people fear for or have concerns for drones. The correlation between surveillance and privacy has been around for quite a while and is applicable in many situations. One of the simplest example being the following: someone sees their next door neighbour peeping through the curtains (surveillance), this someone might feel watched because they have an expectation of privacy, according to Steven L. Nock in The Costs of Privacy (1993). As explained below, this expectation of privacy has to be “reasonable” according to the Fourth Amendment. According to different scholars’ uptakes on the concerns for surveillance and privacy in connection to droneuse in the U.S., most of the concerns are a result of the FAA not feeling specifically responsible for civilian privacy rights as they are responsible for creating and maintaining safe U.S. airspace. This means there is no particular legislative organization focussing on the subject, apart from governments who only seem worried about government intrusion on civilian privacy, while the drone has transgressed from an exclusive governmental use to a more public use (Thompson 2). According to Thompson in “Drones in Domestic Surveillance Operations”(2013) , the Fourth Amendment protecting citizens from unwarranted searches and seizures does not cover the field of droneuse in the U.S. as it only is about the governmental surveillance of citizens and does not take into account citizens spying on or observing other citizens. What is interesting though, is the fourth amendment being based on
“reasonableness”, meaning complaints about privacy infringement should only be taken seriously according to the subjective claim of “a reasonable expectation of privacy” (Thompson 2). Which immediately calls for the question: when is the expectation of privacy reasonable? According to lawyer Richard M. Thompson Drones in Domestic Surveillance Operations (2013), one has a reasonable expectation of privacy within their own homes, but this is still related to the law enforcement use of drones. What really seems to be the issue is civilians being able to use drones, and not having to follow a set of rules, as the FAA has not yet required them to get permission. In this case, the reasonable expectation of privacy within a home, could be deemed unreasonable when it comes to civilians using drones as, according to Caren Myers Morrison, the fear for drones seems to be quite the abstract but not completely crazy one of continuous mass surveillance, by which she compares this feeling of constantly being watched to panoptic theories as have been explained by Foucault (3). Someone who agrees on this comparison is Ryan Calo, who states that the “reasonable expectation of privacy” in US law makes it so citizens are not able to enjoy their right to privacy whenever they are visible, which they seem to be all the times as it is now also happening (by almost invisible machines) from the skies
(31). People now constantly feel observed, turning visibility into a trap, because as soon as a U.S. citizen leaves the frontdoor, his expectation of privacy is no longer reasonable (33).
This claim of drones making people feel observed is confirmed by Calo who says the following on drones and privacy: ‘[…] the unmanned aircraft systems threaten to perfect the art of surveillance’(30). People thus might not know if they are being observed at a certain time which might create a panoptic feeling as explained by Foucault in Discipline and Punish (1995), based on Jeremy Bentham’s design of the infamous Panopticon prison.
The Panopticon was a design by Jeremy Bentham that was eventually never realised. It was supposed to be a round prison with a cylinder in the middle. On the outer walls of the circle would be the cells of arrested criminals and the guards would be in the middle cylinder. The guards could see anyone at anytime, while the prisoners could not see them and never knew whether they were being watched. In this way, control and consciousness were inverted: visibility became a trap (208). According to Foucault: ‘The body now serves as an instrument or intermediary: if one intervenes upon it to imprison it, or to make it work, it is in order to deprive the individual of a liberty that is regarded both as a right and as property’(11). This liberty Foucault speaks about can be interpreted as being free in the literal sense of the word (being able to move wherever and whenever), but also meaning the right to privacy. Bentham thought of a new way to exercise power over others without actually being in touch or contact with the observed bodies (136). While surveillance or privacyinfringement are not the main goal of every drone in U.S. airspace, the expectancy of the air buzzing with thousands of drones by the time of 2030, contributes to this of feeling entrapped, because one never knows when (s)he is being watched. A disciplinary mechanism as described by Michel Foucault in Discipline and Punish (1995), in which prisoners adjust their behaviour because they have the feeling they are constantly being watched, could be compared to the situation of fear for drones being used to observe normal citizens (197). The drones being able to capture and store information on any person by using the technical traits that have been described in the previous chapter, gives people a feeling of inverted panopticism (Rosen, 2013). Foucault says the following about this type of imprisonment: ‘He is seen, but he does not see; he is the object of information, never a subject in communication’(208). Another thing that could connect the panoptic ideas to drone use and surveillance, is the ways of surveillance being available to anyone. Just like the prison inspector could let his family or random strangers be the overseers for a day and in that way would make the surveillance machina available to anyone, so available has droneuse become to anyone, where it at first only would be available to governmental institutions and military organizations. This above section of literature can especially be related to the results that are being retrieved by the DMITCAT on drones because of its conversationlike structure: do users show a sense of feeling
observed? Or do they not implicitly state what they are worried about concerning their privacy and the possibility they are under surveillance. In the case of this thesis, it would be interesting to see if the websites, retweets and hashtags name specific concerns which are related to the drone’s technological features, or whether the results seem to focus on this panoptic feeling of being observed, as
BrackenRoche et al. state one of the main concerns: ‘What can it see and why is it watching?’(4). Chris Schlag seems to summarize the debate in The New Privacy Battle: How the Expanding Use of Drones Continues to Erode Our Concept of Privacy and Privacy Rights(2013): ‘The largest privacy concern arising out of drone use is the drone’s ability to operate as a powerful, inconspicuous, and autonomous surveillance tool’(Schlag, 15). This quote summarizes the common concerns that seem most prominent when it comes to literature on the subject: being watched without seeing the one watching. It seems to be an overall worry, as technologies become more and more able to observe without being visible to the public, we become less aware of the fact we are being surveilled and worse of all: have nothing to say or do about it but protest (and probably lose the battle).
4. Conceptualization: Mapping privacy and surveillancerelated Concerns on Drones As explained above, there are multiple reasons for the concerns of privacy and surveillance within the realm of drones. My task now, is to find out whether these concerns are being discussed or recognized on different platforms. The Issue Crawler will be used with starting points related to drones, privacy and surveillance in the U.S. to find out whether they recognize the concerns and acknowledge one another in these concerns.The DMITCAT then will be used to see what is being discussed or what is “popular content” on Twitter concerning the subject, by looking at a dataset subsample of hashtags and retweets on drones, privacy and surveillance. Two different questions can are asked, one for both methods: 1. Issue mapping: Do the actors within the issuenetwork acknowledge and share concerns on drones in public U.S. airspace, surveillance and privacy? 2. DMITCAT analysis: To what extent do the actors engage in a conversation about concerns on drones in public U.S. airspace, surveillance and privacy by using hashtags and retweets? These methods and the tools that will used are explained further below, first the Issue Crawler methods, results and findings will be explained, after which the same will be done for the DMItcat analysis. 4.1 Tools : The Issue Crawler The Issue Crawler is a tool developed by The Govcom.org Foundation and collaborators. The tool can 2 locate and visualize networks on the web and can be used to crawl for issue networks comprised of actors in the form of organizations, private institutions, companies and private actors that can be affiliated with a subject. The tool can be used to answer questions like: What is the network around this particular issue at a particular time? Why are certain actors not in the network? Which actors share concerns on a certain subject? To find prominent actors in the debate, an in depthliterature search was done on main parties concerned with the use of drones in US public airspace after which link lists from their websites could be used by transmitting them to URLs using the Harvester : a tool that extracts only the URLs from a 3 particular webpage. The biggest USbased parties concerning the use of drones are the EFF (Electronic Freedom Frontier), EPIC (Electronic Privacy Information Center) and FFA (Federal Aviation Administration). EFF’s and EPIC’s main issue with drones being the lack (in transparency) of legislation 2 The issue Crawler Tool and instructions can be found here: https://www.issuecrawler.net/index.php 3 The DMI Harvester Tool and instructions can be found here: https://tools.digitalmethods.net/beta/harvestUrls/
when it comes to the use of drones, and the FFA being responsible for this lack of clarity, as is explained in the previous sections of this thesis. Websites were found through different manual snowballing methods, starting from the FAA, EFF or EPIC websites. A short description of their websites will be given to explain why these websites were suitable for inclusion in the tracing of concerns. The FAA’s website does not have a specific section of their website dedicated to “issues” and mainly focusses on the legislative section of drones in U.S. public airspace. However, they do have a section of their website which is dedicated to “UAVs” and links to privacy matters in which they assure the reader they do not only take economical advantage of UAVs in consideration, they promise to also concerns raised for privacy, civil rights, and civil liberties. The EFF website has a section dedicated to “privacy” which then links to issues and the EFF page on privacy issues concerning “drones and UAVs”. EFF’s focus is on defending rights in the digital world and their main concern when it comes to drones is police and law enforcement spying on normal citizens, while EPIC pays attention to commercial, military and lawmakers’ drones. NGO EPIC’s website has a section dedicated to “policy issues” which links to a page on “drones and UAVs”. Knowing these NGO’s recognizing the issue and blaming the FAA for a lack of clarity on the issue, it is rather interesting to see whether these websites link to one another. To do so, the Issue Crawler will be used. The actors that are key in relation to how they position themselves towards privacy and drones in this thesis are the citizens in danger of being watched or feeling observed and violated in their reasonable expectation of privacy, these actors are represented by organizations concerned with privacy, surveillance and drones like the EFF and EPIC. 4.2 Comprising a list of actors The way to find these actors and place them within an issue network, is by finding a list of websites comprising these actors, after which this comprised list of URLs can be ran through the Issue Crawler. Stepwise the following has been done: 1) Find actors through an indepth literature research and snowballing method taking the EFF, FAA and EPIC websites as starting points. 2) When a list of websites concerning drones and privacy or surveillance could be found, the DMI Tool The Harvester is used to extract the URLs from this list. 3) Anchor texts which linked to websites intext were extracted using the DMI Link Ripper 4
4) Saving the URLs in a list, stating the date of retrieval of the URL, the organization or type of organization the website belongs to (list can be found in appendix I). This type of organization being one of the following parties in the US: Academics, airframe manufacturers, users that got a Certificate of Authorization from the FAA (COA), components manufacturers, dsitributors, sections of the EFF website related to the use of drones , electrical companies using drones, Sections of the FAA website related to the use of drones, US governmental websites as those related to border protection and the Senate, governmental websites relating to institutions like the CIA and law enforcement, hardware manufacturers, information providers, infrastructure inspection companies, journals on drones, model aircraft clubs (hobbyists), NGO’s concerned with the use of drones, service facturers, solution providers, agricultural companies using drones, real estate companies using drones, companies using the imagery provided for drones for different kinds of mapping (of the landscape), companies using drones for research, and companies using drones for video aerial services (media companies and such). In total 247 URLs were copied and saved, distributed over the above mentioned actorgroups which can be found in appendix I. The saved URLlist including all the URLs which was composed as explained above is ran through the Issue Crawler to get an overall feel for the data. All the different actorsgroups’ URLs are also separately ran through the Issue Crawler to see which actors recognize the concerns. The URL list and groups can be found in Appendix I. 4.3 Issue Crawler Settings Particular settings have to be set to do an issuenetwork mapping using the Issue Crawler, the information on the govcom website states Issue Crawler’s default settings are already in place to do an issue mapping 5 network, the crawl depth, iteration and starting points will be shown in the visualisations in chapter 6 of this thesis. An explanation of their meaning is as follows: Crawl depth: The depth per crawl can be set on a scale from 1 to 2. Figuratively speaking, the starting points that have been used for each crawl are considered to be depth 0, the pages found for URL links on a page of depth 5 Explanations, methods and guidelines for using the Issue Crawler can be found here: http://www.govcom.org/
N are considered to be depth N+1. The depth for this research was set to 2, as it is the default setting for constructing issue networks. Iteration: The amount of repetition of steps to find linked URLs. The iteration for this paper is set to 1, as it is the default setting for constructing issue networks. Starting points: The URLs that are crawled to locate a network. The URLs that could be used as starting points were found according to the method as described above and can be found in Appendix I. Bruns states in “Methodologies for Mapping the Political Blogosphere: an Exploration using the IssueCrawler Research Tool”(2007): ‘Indeed, researchers may need to conduct a number of exploratory crawls to better understand the implications of their choices’(n.pag), meaning the choice in starting points or “seeds” that have been used by the researcher. As the entire URL list was put through the crawler with the purpose to get a feel for the overall dataset, smaller groups composed of actors within an actorgroup as explained above were put through the Issue Crawler as follows: Group 1: The entire dataset including all the URLs in appendix I This network was ran through the Issue Crawler for the sake of exhaustiveness and getting a feel of the overall dataset. The URLs that were visualized in the network were then extracted and queried for the word “drone” or “uav” using the Google Scraper and categorizing the websites by frequency of the 6 mentioning of “drone” or “uav”. By doing this, the visualization of the network becomes more qualitative and will show whether drones are especially discussed on for example .gov or .org websites. The following datasets were ran through the Issue Crawler, after which the resulting networks’ URLs were queried for “drones” using the Google Scraper if the results seemed unclear. This way a more indepth analysis of the Issue Crawler networks can be done. Group 2: A Dataset for EPIC websites concerning drones, privacy or surveillance Group 3: for FAA websites concerning drones, privacy or surveillance Group 4: A dataset for EFF websites concerning drones, privacy or surveillance Group 5: A dataset for hobbyist interested in (flying) drones When using the Issue Crawler, govcom.org gives a few guiding points, which were also put to use in the crawls in this thesis. Big media sites should be avoided, to follow up this tip, social networking sites like Twitter and Facebook as well as newspapers’ websites have been left out of the initial starting points.
Thus, before starting a crawl duplicates and big (social) media websites were removed from the starting points. While media sites like newspapers or journals would cause an issuedrift because of their abundance on information on practically every subject and thus were left out, social media websites were also left out because they would probably link to most websites, meaning large social media nodes usually pop up in networks as many sites have a ‘follow us on twitter’ or ‘like us on Facebook’ link. 5. Analyzing Issue Crawler Results When analysing the issue network that is a result of the measures taken above, one can not ignore “the politics of webspace” as described in Digital Methods (2013) by Richard Rogers. One important statement he makes is ‘Making a link to another site, not making a link, or removing a link may be viewed, sociologically or politically as acts of association, nonassociation, or disassociation,
respectively’(44). The links show whether organizations, companies or private entities are to be affiliated with one another. As Noortje Marres states in Why Map Issues?(2015): ‘Hyperlinks do not offer ‘neutral’ delineating data sets, they are instruments for the organizations of networked information, and as such they participate in the (devaluation) of digital content’(13). Websites do not link to one another without a certain purpose and this purpose can be retrieved using ANT to trace associations and digital methods to visualize and Issue Crawler them. Rogers differentiates between different organizational structures of websites, for example claiming governmental websites usually link to other governmental websites, and NGO’s usually providing “public interest” links, calls for action and requests for funding (44). To do research by looking at the way websites link (or do not) to one another, the relationships or the lack thereof can be seen easily. When it comes to these statements and this thesis, it would be interesting to see which websites link to one another on the subject of drones, privacy and surveillance to find out whether they have shared concerns. The Issue Crawler can be used to follow outlinks (the websites a certain website links to). The type of analysis that will be done here is a “colink analysis”, which means only the websites who have 2 links to the seeds (starting points) will be used in the results. When mapping the network that will flow out of the Issue Crawler what has to be kept in mind according to Rogers is ‘[...] the analyst's’ focus is on the real’(53). By which he means the web is no longer a different space but could say something about reality, onground reality, that is. At the same time Rogers refers to Venturini who in “Diving in Magma”(2010) has stated the World is not the Web, or the other way around. An onground issue may
not show up in the crawled results, so when the researcher is looking for a network, it does not necessarily mean an issue network is the outcome, even though it is in the real world.
Doing network analysis as described above by Rogers in Digital Methods (2013), can be put to work in an Issue Crawler network analysis in which interlinkage or the lack thereof might say something about the position of certain organizations and whether the linked websites have shared concerns on the subject. An epistemic network as can be done using the Issue Crawler only visualizes a debate if actors acknowledge one another by linking to them: ‘Broadly speaking, hyperlinking by one organization to another, and reciprocal hyperlinking, may be said to represent a single or common acknowledgement of meaningful acknowledgement in the debate’. (Rogers, Marres 6). But what should be kept in mind is two websites who are presupposed to have a position in the debate not linking, does not mean they both do not have a role in the debate (Rogers, Marres 13). The above method thus does not promise an expected outcome: the outcome of an issue network that can easily be analyzed. Every outcome is an outcome though, and thus, the networks will be analyzed accordingly. A network as can be comprised by the Issue Crawler might detect substantive dynamics of controversy online (12). The way this comes forward (as you will see in the next chapter) is for example Twitter (which was purposely left out of the seeds) still often being a main node in the network, which probably is a consequence of mediatechnological dynamics like ‘Follow us in Twitter’buttons as explained earlier (Marres 13). To analyse the results that have come out of the IssueCrawler as hyperlinked networks, a few questions will be used to observe the linkage between these players in the debate and answer the main question: What type of websites are included, what is their main purpose and how do they link to other organizations? In what way is the network structured, is it dense? What does this linkage say about the type of community the website belongs to, is it a closed community? Which websites are most linked to (the biggest node) and which websites are more on the outskirts of the network? Which organization links to which organization? How are governmental (.gov), nongovernmental (.org) and corporations’ (.com) websites linked to one another? What does this linkage say about their place in the debate? And finally the main question: Do the actors within the issuenetwork acknowledge and share concerns on drones in public U.S. airspace, surveillance and privacy?
6. Results and visualizations IssueCrawler The methods as have been explained in chapter five have been put to work carefully to see whether shared concerns can be found in an issue network surrounding drones and privacy, which is constructed through connections between different actors. Something that is good to remember in analyzing the results, is NGO’s being more prone to link to other NGO’s while governmental websites (like websites of a specific department) tend to only link to other webpages or websites within their own governmental body (Marres, Richards 2000). Clusters of these two types of websites are expected to show. All the above mentioned actor groups were independently ran through the Issue Crawler, but not all the actorgroups as differentiated above constructed networks according to the Issue Crawler results. Therefore, only the networks that could be made by the Crawler are visualized in the following subchapters, including a legend of what the nodes’ colors mean, the date on which the crawl was done, and the crawl’s starting points and settings. The groups’ starting points that could not be formed into an issue network are shown in tables in Appendix II. They were left in the Appendix because they might be of interest for further research.
6.1 All URLs Fig. 1: A network of all the URLs in appendix I in the IssueCrawler and its legend. Settings: Iterations 1, depth 2. May 14th, 2015. As explained before, it might be necessary to run multiple crawls before finding the wanted or expected results. This has been done to structure the above network, in which a process of analysis has lead to letting certain URLs out of the analysis because of their content creating socalled “issuedrift”. When for example using an animal rights party that only dedicates a small section of their website to drones (for example through one blogpost), the focus would be led away from the actual issue. Different iteration and depth settings were also tested to find out which networks would prove the most interesting for research. Bruns argues the settings of smaller iterations and depth might lead to a more “immediate issue network”: ‘This points to an important fact to keep in mind when analysing IssueCrawler data; crawls which are set to use a high number of iterations or high level of crawl depth may proceed beyond what can be regarded as the immediate issue network under investigation, if the neighbourhood of the initial seeds does not immediately constitute a strong network in itself’ (Bruns, n.pag). After this selection process, the above network came out. This is a network of an exhaustive list, because the idea of this network is getting an overall view of the dataset.
When looking at the above crawl, there is especially a strong concentrated network when it comes to governmental sites linking to one another, while the expectation was they would also strongly link to the FAA as it is also a governmental organization which is especially connected to the issue because of its authority in American public airspace. The FAA also being a section of the US DOT (Department Of Transport) would raise the expectation the DOT website would at least be included in the visualized network, it is not. When looking at further clusters, the network is biased towards governmental organizations not especially having anything to do with the issue of drones. The expectation would be the outlinks from companies’ websites using drones would eventually also link to the FAA, as the FAA’s website emphasis is on the judicial claims on droneuse which would be of interest to companies giving drones for rent, selling them or otherwise providing customers with drones or because the FAA is closely related to the U.S. Department of Transport, which is governmental. The governmental websites that seem to be linked above seem to be highly interlinked not because of their shared concerns on drones necessarily, but they are probably interlinked due to other shared interests. This will be checked by doing a Google Scrape using the websites which are shown in the above network and querying them for “drone” or “uav”. The frequency of one of these words in the websites Issue Crawler says something about their involvement in the subject. The Google Scraper tag cloud can be found below in fig. 2 and will be analyzed there.
The websites that do link to the FAA are section508.gov, weather.gov, grants.gov, nws.noaa.gov, and cdc.gov. These are all dotgov websites. Section 508 is a website which focusses on making IT more comprehensible and easy to use for people with disabilities, so has little to nothing to do with drones. When searching for “drones” in the website’s search function, it displays no results. Weather.gov is a website of the National Weather Service, which when looking for “drones” shows results that include the word “drone” in a location’s name as it gives people the opportunity to look for the weather forcast in their area. It also shows the website uses drones to example find the endpoint of a storm, but the website does not have a specific section dedicated to drones. Cdc.gov then, is the website for “Centers for Disease Control and Prevention”, which also does not have a specific section on drones, but links to the FAA on account of injured or sick people needing to be able to take aerial transport to medical help. These results show the necessity of further research by using the Google Scraper. A result that also stands out in this visualisation is the nodes of EPIC and EFF completely disappearing, making a clear (overall) lack of websites on concerns with drones, privacy, or surveillance being very apparent. The small cluster of NGO’s (green nodes) in the top of the visualisation are related to model aviation organizations. These websites all seem to link to the use of drones, but do not have a specific section on issues, privacy or surveillance concerning droneuse.
So, the network seems to mainly exist out of governmental websites, which do not recognize the issue. To find out whether the URLs in the network actually concern drones at all, the URLs in the network were queried for “drone” OR “UAV” in the Lippmannian Device, coming up with the following results: Fig 2: Google scraper result: issuenetwork websites scraped for “drone” or “UAV” to see the occurrence of these words. 22062015. By scraping the URLs that are displayed in the network for “drone” or “UAV” using the Google Scraper (fig. 2) an overview of the types of websites/actors mentioning drones could be extracted. As can be seen above, most of the websites discussing drones are governmental or commercial websites, while there are only three prominently visible NGO websites discussing the subject: modelaircraft.org, asme.org, asce.org, acm.org and nfpa.org. After further inspection, these are websites dedicated to civil engineering and model aviation, computing science and the national fire protection association. None of the websites have a specific page dedicated to drones or issues on privacy but mention the words “drone” or “UAV” in articles related to other causes. The governmental websites that mention “drone” or “UAV” the most are indeed websites which after further manual inspection mention the words, but do not recognize the issue, meaning they do not have a specific page on issues related to drones, privacy or surveillance. The FAA websites is an exception in this, which is not surprising as it is the regulatory organization for U.S. airspace. The issue network may have been influenced by the big amount of starting points as as stated earlier maybe less actually is more. On the other side, the network has helped me get an overview of the data that are included in my URLlist. This network does not particularly seem to acknowledge privacy or
surveillance issues in relation to drones, by applying the “less is more”method by checking smaller portions of the URLlist I hope to find a more visible issuenetwork of shared concerns among actors. 6.2 EPIC The websites that were used for this initial crawl were parts of the EPIC website on droneuse and the FAA lawsuit. These were chosen because EPIC recognizes droneuse as an issue in relation to privacy and surveillance. It will be interesting to see whether EPIC links to other types of (for example) governmental organizations or NGO’s. Again the Google Scraper will be used to find out whether drones are discussed in all the URL’s as displayed in the network. Fig. 3: Pages from the EPIC website concerned with drones, privacy and the FAA (lawsuit). May 21st, 2015. Settings: iterations 1, depth 2. Foia.rocks is the black node in the bottom of the map, which is a website that was launched by EPIC to promote open government and The Freedom of Information Act. The NGO’s (green nodes) in the website are highly interlinked, they all focus on freedom, privacy and rights. Most of these NGO’s claim to be a portal between citizens and governmental organizations, and link to the FTC (Federal Trade Commission)
