1
“IT IS OUT THERE, ON THE INTERNET”
An Exploration of Open-Source Investigation as a New Research Methodology
in International Development Studies
A Robustness Analysis
Tim Valentijn Verlaan 08 August 2019 University of Amsterdam
Supervisor: Robbin Jan van Duijne Second Reader: drs. Courtney Vegelin
MSc Thesis
International Development Studies Graduate School of Social Sciences
University of Amsterdam
2
Abstract
A recent meta-analysis of research methods in International Development studies showed that 98% of UK research projects involve empirical data collection through fieldwork in a developing country. Nonetheless, within the field of International Development Studies, there are growing debates about the negative consequences of this fieldwork tradition. These debates included questions of researcher
positionality, research community exploitation, white saviour complex, research fatigue among specific
populations/communities, under-researched areas due to conflict and inaccessibility, the implications of long-haul flights for fieldwork on climate change, and safety risks associated with conducting fieldwork in developing countries. While fieldwork remains crucial to understand the context and culture of a research location, the field of International Development Studies would benefit from an additional research methodology that is less intrusive, less resource-dependent and does not rely on
the physical proximity of the researcher to the research location.
A research methodology that meets these criteria is ‘open-source investigation’, often referred to as OSINT. This study looks at the possibilities of employing OSINT as a research method in International Development Studies. It does so by analysing the robustness of the methodology as applied in three case studies. These case studies are different across multiple dimensions, however, the common denominator between these cases is that they all revolve around human rights violations in conflict areas in developing countries. The main research question of this study is “How robust is Open-Source
Investigation (OSINT) and can it be used as a viable methodology for Development Studies in researching human rights violations in a conflict context?”. This study finds open-source investigation
to have the potential of being a robust research method. The main limitations to the robustness of OSINT are technological in nature or derived from a lack of clear research protocol. While OSINT seems diametrically opposed to the ontological and epistemological position of research in Development studies, there are many unique possibilities it has to offer. The author believes that if the youngest generation of researchers is made aware of OSINT’s potential in Development studies, it can have a bright future as an alternative research methodology. This thesis comes with an ‘Introductory Guide to OSINT’.
Keywords: open-source investigation, OSINT, research methodology, robustness, international
3
Acknowledgements
I would like to express my gratitude to all those who have supported this research project. First and foremost, I would like to thank my supervisor Robbin Jan van Duijne. I came to Robbin Jan in a state of mild despair after having struggled with the approach to this topic for almost half a year. His clear vision, optimism, flexibility, and extensive guidance through meetings, skype calls and comments greatly improved my skills as a researcher and allowed me to produce this thesis. Additionally, I am grateful to everyone in the ‘OSINT community’ who showed me the ins and outs of this exciting methodology. Special thanks go out to those have contributed directly: Aliaume Leroy, Christiaan Triebert, Benjamin Strick, Milena Marin, ‘Sector35’, and Hervé. Also, I would like to thank my first supervisor Josh Maiyo for his time and efforts. Finally, many thanks to Courtney Vegelin for acting as my second reader. I hope this final product is able to ‘make a difference’.
4
Table of Contents
1
Introduction
7
2
Theoretical Framework
10
2.1
Historical perspective on Ontology and Epistemology in
10
Development studies
2.2
The position of OSINT with this methodological paradigm
13
2.3
In-depth investigation of the concept of robustness
14
2.4
Open-source investigation’s nature of scientific enquiry
18
3
Research Design
20
3.1
Research questions
20
3.2
Key characteristics of the research design
21
3.3
Approach, data analysis and operationalisation
23
3.4
Reflection of the research process and empirical chapters
26
3.5
Statement on the research location
27
3.6
Statement of the limitations of the research
28
3.7
Statement of the ethical considerations of the research
29
4
Case study: ‘Anatomy of a Killing’ (Cameroon 2018)
30
4.1
Introduction
30
4.2
Methods and Data
31
4.3
Evaluation of the Robustness
37
4.4
Findings
42
5
Case study: ‘Social media executioner’ (Libya, Benghazi 2017)
44
5.1
Introduction
44
5.2
Methods and Data
45
5.3
Evaluation of the Robustness
55
5.4
Findings
60
6
Case study: ‘Strike Tracker project’ (Syria, Raqqa 2019)
62
6.1
Introduction
63
6.2
Methods and Data
64
6.3
Evaluation of the Robustness
66
5
7.
Conclusion and Discussion
72
7.1
What is the prevailing ontology and epistemology of
72
Development studies and how does OSINT fit in this paradigm?
7.2
What are the dominant techniques of OSINT and what are their
73
limitations in terms of robustness, in researching human rights
violations in a conflict context?
7.2.1 Geolocation
73
7.2.2 Chronolocation
74
7.2.3 Photo/video analysis
74
7.2.4 Crowdsourcing
75
7.3
Conclusion and Discussion of the sub-research question three
76
7.4
Conclusion and Discussion of the main research question
77
8.
Bibliography
79
6
List of abbreviations
BBC British Broadcasting Corporation DEM Digital Elevation Model
GIS Geographic Information System ICC International Criminal Court
IS Islamic State
IDS International Development Studies NGO Non-governmental Organization
OCHA United Nations Office for the Coordination of Humanitarian Affairs OSINT Open-Source Investigation
OSM OpenStreetMap
UK United Kingdom
7
1. Introduction
On the 3rd of February 2016, the mutilated body of Giulio Regeni was found in a ditch on the outskirts of Cairo (Walsh, 2017). The PhD student from Cambridge University was conducting fieldwork in the capital to research Egypt’s Independent Trade Unions. The autopsy report showed that Regeni was tortured for up to 7 days. According to the International NGO Safety Organisation (INSO), there have been 2023 incidents involving individuals working and conducting fieldwork in a developing country since 2011; 224 of which led to injury, 89 to death and 215 were characterised as abductions (INSO, 2019).
Recent meta-analysis of research methods employed in International Development Studies (IDS) showed that 98% of research projects funded in the UK involved empirical data collection through fieldwork in developing countries (Camfield, 2019). Apart from the safety risks associated with conducting fieldwork in developing countries (Fast, 2003, 2010; Rodgers, Koonings, & Kruijt, 2019), there are growing debates within the field of IDS about the negative consequences of this fieldwork tradition. These debates include questions of researcher positionality (Dean et al., 2018; Lin, 2015; Milner, 2007), research community exploitation (Pratt et al., 2015), white saviour complex (Bandyopadhyay, 2019; Frey, 2016), research fatigue among specific populations/communities (Boesten & Henry, 2018; Brinkman et al., 2016; Clark, 2008), under-researched areas due to conflict and inaccessibility (Goodwin, 2008), and the implications of long-haul flights for fieldwork on climate
change (Desiere, 2016; Kuffer, 2016; Strengers, 2014). These debates invoke the suggestion that the
field of IDS could benefit from alternative research methods that are less intrusive, less
resource-intensive and are able to be conducted without the need for physical proximity of the researcher to the
research location. While fieldwork remains crucial to understand the context and culture of a research location, it is not unthinkable that alternative research methods might supplement, (partly) substitute or even yield more and better data in some contexts than research methods reliant on fieldwork. In this reasoning lies the academic and societal justification of this research project into an alternative/ additional research method.
In recent years, a promising research method called ‘open-source investigation’ (or OSINT) has emerged which meets the above-mentioned criteria (Hassan, 2019). So, what is open-source investigation and what does it entail? Open-source investigation is the process of obtaining intelligence from open-source data (Dupont, 2004). Open-source data is defined as “all data that is freely accessible, primarily through the internet” (Abilova & Novosseloff, 2016). When this open-source data is processed and exploited, it is referred to as open-source information, when this information is
8 subsequently analysed, triangulated and corroborated, it is referred to as open-source intelligence (Babalyan, 2001).
Since 2014, OSINT has gained momentum amongst NGOs and investigative journalists; most prominent among them is the ‘Bellingcat Collective’. In 2017, ‘Conflict, Security and Development’ Master’s student, and part-time Bellingcat contributor, Christiaan Triebert, became the first Dutch citizen to win the European Press Prize at the age of 26. He wrote his prize-winning article ‘The Turkish Coup through the Eyes of its Plotters’ using OSINT techniques from a dorm room in Malaysia. In 2019, he was named Rijksuniversiteit Groningen’s Alumnus of the year for his work in the field of open-source investigation (Tiekstra, 2019). Triebert showed the effectiveness, efficiency and possibilities of open-source investigation in numerous major investigations (e.g. MH17; Khashoggi; Skripal) and by explaining the methodology during regular appearances on popular Dutch talk shows. However, given the relative newness of the research method, a lot is still unclear about the research methodology. As a result, critics, online and in the media, have accused it of lacking methodological rigour and being based on cherry-picking (Bidder, 2015; Dawson, 2018; James, 2016)
Nevertheless, recent open-source investigations show the methodology to be promising in researching human rights violations in conflict-riven developing countries (Ambrose & Burns, 2019). In these investigations, most data is gathered from just three open sources: satellite imagery, social media, and photo and video material from websites such as YouTube. Photo and/or video material is in virtually all these open-source investigations the primary source of data. This is mainly due to the growing availability of smartphones in conflict areas which are used to record and share videos, photos and live streams of human right violations on social media (Tkach & Williams, 2018). This imagery, combined with satellite data and other forms of information shared through social media, results in vast amounts of open-source data on human rights violations coming out of conflict areas (Ambrose & Burns, 2019). However, in spite of the growing possibilities and attention for open-source investigation, the method gained very little traction in International Development Studies and precious little is known about its usefulness and applicability in the IDS field. Also, besides being questioned for its methodological rigour by critics online and in the media, the method has not yet been subjected to scrutiny by the scientific community. Nonetheless, the possibilities of open-source investigation for Development studies seem promising and a test of OSINT’s robustness as a scientific research methodology demands attention. Therefore, the aim of this thesis: to critically evaluate the robustness of the OSINT methodology, to investigate whether it qualifies as a viable additional research method in International Development Studies.
9 This research project critically evaluates the OSINT methodology by carrying out a robustness analysis of three case studies. These three cases are carefully selected based on a number of differences and commonalities: all three case studies make use of a distinct type of open-source investigation; all three case studies are carried out by different open source investigators; each case is located in a different geographic context. While differently focussed, the common denominator between these cases is that they all revolve around human right violations in conflict areas in developing countries. The thesis is confined to conflict areas because this is where open-source investigation is most commonly and successfully employed. Hence, this study is able to present findings on OSINT’s potential robustness, fundamental limitations and raise interesting questions about the flaws and strengths of its methodological procedures in different geographical contexts. However, as this study does not involve a random, representative sample, its findings are not generalizable to the overall employment of open-source investigation. The three case studies in this research project are: a case in the far-North of Cameroon on extrajudicial killings by the Cameroonian military; a case in Benghazi, Libya on the summary executions of hostile fighters by the Libyan National Army; and a case in Raqqa, Syria on the occurrence and denial of civilian casualties as a result of US-led Coalition bombings in 2017.
This research project is, to the best of the author’s knowledge, the first exploration of OSINT as a potential research method in Development studies. The main research question of this project is:
How robust is Open-Source Investigation (OSINT) and can it be used as a viable methodology for Development Studies in researching human rights violations in a conflict context?
After this introductory chapter, this thesis will proceed with a theoretical chapter that includes: (1) a historical perspective on the ontological conceptualisation of development and the epistemological position of research methods within Development studies; (2) an in-depth evaluation of the concept of robustness; and (3) an introduction to the Philosophy of science behind open-source investigation. Chapter three presents the research design of the thesis, lays out the operationalisation of the core concepts, describes the methodology used and comments on the limitations and ethical considerations of this project. Chapters four, five and six are the empirical chapters on the Cameroonian, Libyan and Syrian OSINT cases respectively. These chapters all centre on critically evaluating the robustness of the OSINT methods used. Chapter seven is the conclusion and discussion chapter in which the findings of this research project are summarized, put into context and in which the research questions are answered. Finally, the overall research project is discussed and reflected on. Additionally, instead of making policy recommendations, this thesis comes with an ‘Introductory guide to OSINT’. The guide is written for future IDS master students and is meant to spark their curiosity about the possibilities of using open-source investigation in their thesis projects.
10
Figure 1: The scaffold of learning. Source: Crotty (1998, p.5)
2. Theoretical Framework
Readers guide
This theoretical chapter is divided into four sections. The first section situates open-source investigation in the broader, contemporary debate on research methodology and methods in International Development Studies. In order to do so, first, a historical perspective on the prevailing ontology and epistemology within IDS is discussed. Then, an analysis of the ontological conceptualisation of ‘development’ and epistemological position on research methods within International Development Studies is made. Afterwards, a historical perspective of research methodologies in IDS is laid out and the current prevailing paradigms are discussed.
In the second section, the position of OSINT within this dominant methodological paradigm is critically weighed and laid out.
In the third section, the focus shifts to the concept of ‘robustness’. The main aim here is to untangle its various conceptualisations and present a precise and in-depth analysis of this contemporary scientific ‘buzzword’ (Soler, 2012). At the end of this section, the prerequisite conditions to robustness: reliability and validity, are introduced and laid out.
The fourth section ties the detailed conceptualisation of robustness from section three to the nature of scientific enquiry of open-source investigation. This section discusses the role of triangulation, corroboration, verification, and falsification in open-source investigation and lays out how these concepts are employed in three dominant OSINT techniques.
2.1 Historical perspective on Ontology and Epistemology in Development studies
A research method can be defined as a systematic plan to gain new knowledge or deepen understanding of reality (Bryman, 2012). As such, before ‘research methods’ can be understood properly, the nature of ‘reality’ and what constitutes ‘knowledge’ should first be understood. The branches of philosophical enquiry that deal with these questions are ontology and epistemology. Ontology informs one’s choice of epistemology which informs the choice of theory/conceptual framework, which in turn informs one’s choice of methodology and methods. This link is what Crotty (1998, p. 5) refers to as “The scaffold of learning”, and can be conceptually understood as presented in Figure 1.
11 Within the social sciences, Ontology is concerned with the nature of social entities: “The central point of orientation here is the question of whether social entities can and should be considered objective entities that have a reality external to social actors, or whether they can and should be considered social constructions built up from the perceptions and actions of social actors” (Bryman, 2012, p. 32). The two central positions within research ontology are objectivism and constructivism (see Crotty, 1998). In brief, objectivism states that social phenomena are external facts: they are beyond the reach and influence of humans (Crotty, 1998, p. 33). Within the ontology of objectivism, social phenomena exist independent of actors (Bryman, 2012, p. 575). Bryman (2012) uses the example of an organisational hierarchy: employees filling in the hierarchy come and go, but the organisational framework remains intact. The organisational hierarchy exerts pressure and exists independently from the people that make up the hierarchy. Constructivism poses the exact opposite thesis. Constructivism asserts that social phenomena and their meanings are continually being manifested into the world by their actors (Scott & Marshall, 2009, p. 484). Social phenomena are, therefore, not only the product of social interaction, they are continually being revised by social interactions. In this view, a researcher is only able to ‘construct’ his version of social reality, rather than one that can be regarded as categorical. Epistemology is the branch of philosophy concerned with what is often referred to as the ‘theory of knowledge’: what is meant by ‘knowledge’, and how acceptable knowledge is arrived at (Crotty, 1998, p. 8). The central issue of epistemology in the social sciences can be summed up in the question “Can and should the social world be studied according to the same principles and procedures as the natural sciences?” (Bryman, 2012, p. 27). People who affirm this central issue are called ‘positivist’, and often take an objectivist ontological position in their research approach. Positivism is centred on the notion that there is an external reality that is perceivable through the senses (empiricism). Positivists collect data in order to arrive at theories and laws (inductivism) and then collect more data to test these theories and laws (deductivism). On the other hand, people who oppose the central issue assign to ‘interpretivism’. Interpretivism is centred on the notion that the subject of the social sciences – people and organisations – is fundamentally different from that of the natural sciences. Therefore, the study of the social world requires a different logic of research procedure; it requires social scientists to grasp the subjective meaning of social action. Interpretivism is based on an ontology of constructivism. Since the beginning of the ‘Development debate’ in the 1950s, there have been various ontological conceptualisations of ‘development’. Its first ontological conceptualization in academic literature is as a ‘process of structural societal change’ (Gore, 2000, p. 794) or ‘a process of historical change’ (Thomas, 2004, p. 1). While it relates development to various socio-economic changes, it does not relate to any specific objectives. This first conceptualization of development is not necessarily prescriptive. It does not expect any two countries to develop in the same manner. All countries change over time, the
12 economic growth (even negative growth) and societal change these countries experience can be accepted as ‘development’ (Sumner & Tribe, 2008, p. 12). In the 1980s, the second view of development materialised; referred to as ‘post-development’ (Escobar, 1995; Sachs, 1992; Shiva, 1988). Post-development is grounded in structuralism and the post-modern critique (Hickey & Mohan, 2005, p. 38). Post-development states that concepts such as ‘poverty’ and ‘development’ are social constructs that merely exist in and through discourse (a body of ideas, concepts and theories) (Shiva, 1988, p. 10). Post-modernists argue that we can only know reality through discourse, and that discourse is employed to shape and define reality (Foucault, 1973). The ability “to define reality is a crucial aspect of power and one of the major means by which certain groups … are silenced and suppressed.” (Booth et al., 1999, p. 12). In a way, post-development is not so much an ontological conceptualisation of ‘development’ but rather; a frontal assault on the practices of the ‘development industry’ (Sumner & Tribe, 2008, p. 14). Around the change of the Millennium, a third ontological conceptualisation of development came up. This conceptualisation is often referred to as big-D Development and is described as a ‘project’ of intervention in the ‘Third World’ (Hart, 2001, p. 650). It is characterised by ‘a vision or measure of progressive change’ (Thomas, 2004) witnessable in the Millennium Development Goals and their successor, the Sustainable Development Goals.
The interdisciplinary nature of Development studies makes it hard to point towards a ‘signature’ epistemology or research methodology (Camfield, 2019, p. 243). Different development challenges require different epistemologies and research methods. Nonetheless, there is some literature that sheds light on dominant research paradigms. In a recent meta-analysis of the most used research methods in Development studies, Camfield (2019, p. 251) finds over 60% of Development studies research projects submitted in the UK in the past ten years to be employing a mixed methods design. However, within these mixed methods research designs, surveys (93%, 51% of which secondary), semi-structured interviews (81%) and focus groups (33%) predominate. The remaining 40% is equally divided between quantitative and qualitative, both 20%. Where the quantitative studies predominately consist of a survey research method, the qualitative studies employ more diverse research methods (Camfield, in personal communication). Camfield (2019) suggests that this tendency to incorporate a quantitative component in the research design reflects a fear among qualitative researchers that their research will otherwise be flagged as “insufficiently rigorous, lacking in reliability and representativity, and not satisfy donor body requirements”. Camfield (2019, p. 253) found that the majority of researchers are not mixing methods in an innovative way as the designs generally lack integration of qualitative and quantitative components, stating ‘a survey is just a more structured interview’. Camfield (2019, p. 254) also notes these findings might be prescribed to the seniority of lead authors and principal investigators, stating that this cohort of scientist learned to do research in
13 a time when there was a different focus in research and different tools available. Additionally, more than 98% of the UK Development studies research projects involved primary empirical data generation through fieldwork (of which 99% in the Global South), which was assigned to the expectations of funding bodies (Camfield, 2019, p. 256).
From these statistics, it can be argued that the dominant epistemological position with Development studies is (still) interpretivism. Despite the fact that various incentives appear to have driven scholars to adopt more quantitative (positivist) methods into their research, the core of Development studies research appears to rely heavily on traditional interpretivist research methods such as (semi-structured) interviews, focus groups, and ethnography. This reminisces of what Max Planck said, later echoed by Thomas Kuhn, about scientific revolutions and paradigm shifts: “A new scientific truth does
not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it.” (Kuhn, 1962; Planck,
1949).
2.2 The position of OSINT within this methodological paradigm
So, what implications does this have for the positioning of OSINT in Development Studies? Naturally, first, an exploration of the ontological and epistemological position of OSINT needs to be made. According to Sumner and Tribe (2008, p. 59), OSINT’s ‘position’ can be determined by asking three general questions: “What is the nature of reality?”; “What is the aim of knowledge enquiry?”; and “What is the relationship between the researcher and the researched?”. OSINT assumes that there is one objective reality that is observable through the internet. OSINT aims to approach this ‘truth’ by gathering objective facts through techniques such as verification, falsification, triangulation and corroboration. In OSINT, the researcher is presumed to be objective and independent from the research subject. OSINT refrains from reasoning from authority and gives the highest priority to full disclosure of information, validity, reliability, transparency, and replicability of the research. The credo of OSINT could be loosely summarized in the sentence: “Don’t believe me, but look what I found and where I found it, and piece the puzzle together for yourself” (Babalyan, 2001). Judging from these answers, OSINT is clearly rooted in an ontological and epistemological position that appears to be diametrically opposed to the position of the traditional research methods in Development studies; OSINT’s position is rooted in objectivism and positivism.
The OSINT methodology can be considered both quantitative and qualitative. This is because OSINT might involve content analysis of social media data such as tweets (qualitative) but also ‘following the money’ of abusive firms and governments operating in conflict areas through tax declarations and other open documents (quantitative) as well as geolocating damning video and photo material (verification/falsification). OSINT is pragmatic in its application of research methodologies; it uses
14 ‘what works’. Since ‘what works’ is often unclear at the beginning of an investigation, OSINT should depart from a mixed methods approach and clarify the specific methods used afterwards. As such, OSINT can be considered ‘exploratory research’, in a non-traditional sense.
There is also the question of whether OSINT works with primary or secondary data. While an initial thought into this might lead one to conclude that it solely uses secondary data, the ‘OSINT community’ would disagree. While it is true that the data has been previously generated by someone else, it often contains ‘first-hand accounts’. Data from open sources is generally spontaneously generated through intrinsic motivation, not prompted and steered by interview or survey questions. The relative anonymity of the internet also allows for more honest sharing of opinion. Hence, social media is increasingly viewed as a primary source (Coleman, 2013).
In conclusion, OSINT draws from ontological and epistemological positions opposite to those currently dominant in Development studies. This contradiction might explain why the method has not found its way into this discipline. Also, the fact that the method is inherently technological might decrease its adoption rate in a field where the leading scholars are of senior age and are described as ‘sticking to their methods’ and ‘not innovative in mixing methods’ (Camfield, 2019).
2.3 In-depth investigation of the concept of Robustness
As this thesis is now situated in the discussion of research methodologies in Development studies, this theory chapter will proceed with a discussion of the main concept of this study: robustness.
The concept of robustness has a long tradition in science – dating back to the work of Aristotle – but has seldom been the primary focus of research (Stegenga, 2009; Wimsatt, 2012). “In recent years the term robustness has become a common platitude in the social sciences” (Stegenga, 2009, p. 650). As with many ‘buzzwords’, the popularity of the term tends to water-down its original meaning. Therefore, a rigorous substantive review of the concept and clarification of the terminology is needed. At first sight, robustness appears to be a concept that is intuitively clear, yet, it is far from easy to give a precise account of just what is implied by this notion (Soler, 2012). As pointed out by Nickles (2012): “In recent years the term robustness, its cognates and neighbours (solidity, persistence, hardiness, resilience, flexibility, healthiness, etc.) have been applied to just about everything.” In fact, ‘robust’ has become such a buzzword in popular culture that it can be applied to anything that “exhibits strength of some sort” (Nickles, 2012, p. 330). In academic literature, the concept of robustness first shows up in the works of Richard Levins (1966, 1968) and Donald Campbell (1958, 1966, 1969, 1974), although the latter mainly refers to it as ‘multiple means of determination’ (Wimsatt, 1981). Currently, the main point of reference for the concept of ‘robustness’ in the Philosophy of science canon is the seminal
15 1981 paper by William Wimsatt (Soler, 2012). In this paper, Wimsatt defines ‘robustness’ as “the use of multiple means of determination to triangulate on the existence and character of a common phenomenon, object, or result”(Wimsatt, 1981, p. 5).
Wimsatt starts his general analysis of the concept of robustness by quoting Feynman’s (1967) popular lectures on the character of physical law. In these lectures, Feynman distinguishes two approaches to the structure of physical theory: the Greek and the Babylonian approach. In the Greek approach, the fundamental principles of science are taken as axioms: statements that are taken to be true, to serve as a premise for further reasoning. ‘Greek’ theorists achieve simplicity and elegance by taking only a small number of axioms, from which all theorems are derived (Wimsatt, 2012, p. 64). Most illustrative of these ‘Greek’ theorists is Rene Descartes, known for his fundamental axiom ‘Cogito, ergo sum’ (I think; therefore, I am). The principal rationale for this axiomatic view of science is to attempt to make the structure of scientific theories as reliable as possible by starting with, as axioms, the minimal number of assumptions which are ‘as certain as possible' and operating on them with rules which are as precise as possible (deductive rules which are truth-preserving) (Wimsatt, 2012, p. 65). In attempting this highest possible level of reliability, the focus is entirely on eliminating errors from occurring, not on recognizing errors that do occur and controlling their effect. In this structure: if no errors are introduced – through faulty axioms – and no errors are made in applying these axioms and the rules of logic, no errors will occur (Feynman, 1967). However, this structure is not the best in dealing with errors – by minimizing their effects or making them easy to identify – if they do occur. The ‘Greek’ structure can be illustrated as a chain of reasoning. Popular mantra has it that “a chain is as strong as its weakest link”, according to Wimsatt (2012, p. 66), a chain of reasoning is always weaker than its weakest link. Chains will always fail at their weakest link, chains of reasoning not necessarily. If a link in a chain breaks, the release in tension protects the other links from breaking: failures in a chain are not independent (2012, p. 66). In contrast, theorems in a chain of reasoning are more likely to fail if one of the assumptions they are based on fails. Therefore; “fallible thinkers should avoid long serial chains of reasoning” (2012, p. 66). Luckily, an antidote to the inevitability of failure in long serial chains of reasoning can be found in robustness. This is represented in the ‘Babylonian’ structure of theory; it focusses less on the exclusion of errors and more on the triangulation of theories and concepts from ‘multiple means of determination’. The Babylonian structure is not built up from central axioms, reasoning out towards theorems at the periphery, but by certain man-made conventions from where everything can be constructed. Feynman, in his 1965 lecture, describes it as:
16 “What I have called the Babylonian idea is to say: I happen to know this, and I happen to know that, and maybe I know that, and I work everything out from there. Tomorrow I may forget that this is true, but remember that something else is true, so I can reconstruct it all again. I am never quite sure of where I am supposed to begin or where I am supposed to end. I just remember enough all the time so that as the memory fades and some of the pieces fall out, I can put the thing back together again every day” (Feynman, 1967, p. 46)
The ‘Babylonian’ structure is the reasoning structure behind mathematics. Mathematicians can ‘start’ in different places; all the theorems are interconnected; there is no way of identifying any fundamental axioms (Feynman, 1967, p. 46). This interconnection triangulates the theorems with one another; “It is like a bridge with lots of members, and it is overconnected; if pieces have dropped out you can reconnect it another way” (1967, p. 46). While the ‘Greek’ structure was exemplified by a chain, the ‘Babylonian’ is best described as a cable; “a cable whose fibres may be so slender, provided they are sufficiently numerous and intimately connected” (Peirce, 1868, p. 141). This is the big advantage of ‘multiple means of determination’, a theorem in a sequence of reasoning is not reliant on the weakest link, but rather, supported by its cognates and neighbour theorems. An example illustrating the advantages of the Babylonian structure is the case of the Indian mathematician Srinivasa Ramanujan.
Ramanujan was a poor, uneducated man born in India in 1887. Ramanujan was able to prove and (re)discover many mathematical theorems in solitude of the rest of the mathematical world from just a few books he had access to in his childhood. No consensus on the fundamental axioms was needed for Ramanujan to accomplish his extraordinary work. Ramanujan merely needed knowledge of man-made conventions such as the numeral system from his few books, from there, he was able to, in Feynman’s words, ‘reconstruct it all again’. (see Kolata, 1987)
However, the Babylonian structure also has its disadvantages. As Feynman (1967, p. 46) said: “so physicists do Babylonian mathematics and pay but little attention to the precise reasoning from the fixed axiom”. One of these days, the field of physics might experience the repercussion of this, demonstrated in the following text on a likely upcoming paradigm shift in the field of Physics:
Within the field of Physics, ‘quantum physics’ has very successfully described the behaviour of tiny particles such as atoms, while ‘relativity’ has accurately described forces at a cosmic scale. However, in some cases, mainly gravity, the two fields yield incompatible results. Currently, scientific experiments are carried out which are thought capable of disproving the ‘equivalence principle’; one of the main assumptions of
17 Einstein’s General Theory of Relativity. This would mean that a significant error pervaded in theory, which might have severe implications for other theorems in the field of Physics. To remain with the analogy in the previous text; the equivalence principle can be viewed as one of the thicker strings in the cable, without which the cable loses considerable internal consistency. (see Dooley, 2018)
Illustrated in the paragraphs above are the fundamentals of robustness and its advantages and disadvantages. A clear advantage of robustness is that findings are fortified: they are not solely dependent on one ‘means of determination’ or research methodology. A disadvantage is that findings of a method should not be taken as true merely on the basis that they corroborate the findings on other research methods. Therefore, it is important that reliability and validity remain preconditions to critically appraising the merit of research methods before they are used to bolster the robustness of a finding.
Reliability of a method is fundamentally concerned with whether the measures that are devised for concepts in a study are consistent. A research method is reliable when it responds to the same phenomena in the same way regardless of the circumstances of its implementation (Krippendorff, 2005a). Reliability is not concerned with the world beyond the research process itself: it cannot tell the researcher what the phenomena he/she is researching ‘really’ are, it can only assure him/her that their procedures can be trusted to have responded to real phenomena (Krippendorff, 2005a, p. 212). A lack of reliability limits the chance of validity, however, reliability does not guarantee validity (Krippendorff, 2005a, p. 213). Hence, its relationship is correlative, not causal.
Validity is concerned with the integrity of the conclusions that are generated from a piece of research (Krippendorff, 2005b). Facets of validity relevant to this research project are internal, external, and face validity. Internal validity is concerned with the question of whether a conclusion that incorporates a relationship between two or more variables holds water (Krippendorff, 2005b, p. 315). External validity is concerned with the question of whether the results of an observation can be extrapolated outside the specific observation (Krippendorff, 2005b, p. 315). Face validity is the logical judgement of whether research findings ‘make sense’, whether they are “plausible and believable on their face”(Krippendorff, 2005b, p. 313).
18
2.4 Open-source investigation’s nature of scientific enquiry
The previous two sections situated open-source investigation in the prevailing ontology and epistemology in Development studies and laid out the concept of robustness as the theoretical framework. In this section, a brief overview of open-source investigation’s nature of scientific enquiry is presented in order to set the stage for the operationalisation of this research project and to expand the reader’s knowledge of the internal dynamics of the methodology before commencing with the case study chapters. Also, this section ties together the need for the rigorous substantive review of robustness with the practice of doing OSINT research. As mentioned before, because of the methodological nature of this study, there is some overlap with the next chapter, chapter three on research design.
Below, first, the nature of scientific enquiry in open-source investigation is introduced. Afterwards, three OSINT methods crucial to this project are laid out.
Video and/or photo material is almost without exception the primary source of data in an open-source investigation into human rights violations in a conflict area. Open-source investigation starts by extracting clues from the photo/video material to answer the archetypal questions: what, where, when, who and why; and to support or challenge certain claims. These clues are triangulated and corroborated with other pieces of open-source intelligence, collected mainly from satellite imagery and social media. Effectively, this is a verification/falsification process based on visual clues. The concept of falsification comes from the Philosophy of science concept of deduction, or deductive
reasoning (Popper, 2013). Deduction is the process of reasoning from one or more premises to a logical
conclusion which is certainly correct (Evans, Newstead, & Byrne, 1993). Deduction is what was described by Feynman (1967) in the last section as the ‘Greek approach’. A statement, hypothesis or theory is ‘falsifiable’ if a contradicting statement can disprove it. According to Popper, science should aim for falsification, as this provides a higher degree of certainty than verification. This is exemplified in Popper’s famous quote:
No number of sightings of white swans can prove the theory that all swans are white. The sighting of just one black swan may disprove it.
The concept of verification comes from the Philosophy of science concept of induction, or inductive
reasoning (Popper, 2013). Induction is defined as a method of reasoning in which you use individual
ideas or facts to come to a general rule or conclusion (Popper, 2013). The problem of induction is that it presupposes that in the future, events will occur as they have in the past; this premise is described by David Hume as the ‘uniformity of nature’ (Hume, 2016). According to Hume, there are no grounds to accept this premise. Therefore, findings obtained through verification have a lower degree of
19 certainty. Unfortunately, the majority of OSINT methods rely on a process of verification rather than a process of falsification. However, as rigorously explained in the previous section, findings obtained through inductive reasoning can be bolstered through the concept of robustness. As OSINT uses multiple means of determinations to triangulate and corroborate its findings, it is still able to arrive at robust, valid and reliable results.
In this thesis, the verification process is mainly employed in three distinct OSINT ‘techniques’. These techniques are the dominant mechanisms through which source data is processed into open-source intelligence, upon which findings are based. The techniques are geolocation, chronolocation, and crowdsourcing.
Geolocation is defined as “The process or technique of identifying the geographical location of a person, device or structure by means of digital information processed via the Internet.” (Oxford Dictionary, 2019). In geolocation, clues from photo/video material are compared to satellite data. The goal is to find out where the imagery is recorded. In researching human rights violations in conflict contexts, geolocation is mostly done through inductive reasoning. The verification process is reliant on clues such as vegetation and small uniform buildings. As such, if more clues can be verified, the robustness of the correct geolocation increases. Hence, the geolocation process usually continues until all clues are exhausted.
Chronolocation follows the same reasoning, but this time the goal of the technique is to find out when the imagery is recorded. Chronolocation is done in two ways. First, it compares the presence and/or absence of structures (e.g. buildings) in the video to historical imagery in GIS software. This gives an initial indication when the imagery was recorded. Secondly, it utilizes the shadows in the video to calculate the Solar position and subsequently determine the time and date when it was filmed. Crowdsourcing is used in two different manners. The first form is the use of expert knowledge throughout the research process from social media ‘knowledge workers’. Knowledge workers are actors that have expert knowledge of a particular topic and freely share this knowledge through social media, blogs, forums and comments (Majchrzak, Faraj, Kane, & Azad, 2013). Knowledge workers are consulted for the same reasons ‘normal’ experts are consulted: to gain answers and speed up the investigative process. The second form of crowdsourcing is the systematic acquirement of crowd analysis. As open-source data can be overwhelmingly rich and vast, this type of crowdsourcing is mainly used to increase the efficiency and robustness of the verification process (Howe, 2006).
As the position of OSINT in the Development studies paradigm is clear, and the main concept of ‘robustness’ has been tied to the nature of scientific enquiry of open-source investigation. This thesis now continues with the research design of this project.
20
3. Research Design
Readers guide
In this chapter, the research design of the project will be laid out, the concept of ‘robustness’ will be operationalised, the research location will be discussed, a reflection of the research process and empirical chapters is presented and the limitations and ethical considerations are elaborated on. First, the main research question is re-iterated, and divided into three manageable sub-research questions.
3.1 Research questions
The main research question of this research project, as presented in the introduction, is:
How robust is Open Source Investigation (OSINT) and can it be used as a viable methodology for Development Studies in researching human rights violations in a conflict context?
The first sub-question aims to situate the OSINT methodology within a framework of the prevailing ontological and epistemological positions in Development studies. This research question is posed prior to the commencement of the empirical chapters. It reads:
(1) Sub-RQ1: What is the prevailing ontology and epistemology of Development studies and how does OSINT fit in this paradigm?
The second sub-question further guides the empirical chapters. It is focussed on the OSINT methodology itself, and the applicability of the methodology in researching human rights violations in conflict settings. The second sub-RQ reads:
(2) Sub-RQ2: What are the dominant techniques of OSINT and what are their limitations in terms of robustness, in researching human rights violations in a conflict context?
The third sub-question looks at the different geographical contexts in which OSINT is employed in this research project and aims to draw lessons from this. The goal is to increase understanding about what makes a suitable context for the application of open-source investigation.
(3) Sub-RQ3: What lessons can be drawn from open-source investigations that were conducted in various geographical contexts?
21
3.2 Key characteristics of the research design
As the nature of this research project is methodological, its research design is distinctly different from those normally employed in Development studies. In the following paragraphs, I will detail the key characteristics of this innovative design, and elaborate on why this design is most appropriate for critically evaluating OSINT’s robustness in development studies.
This research project is the first of its kind in Development studies. The first key characteristic of the design is, therefore, that this project concerns exploratory research, predominantly aimed at raising important methodological questions that could benefit the discipline, and not to write hard/definitive conclusions about OSINT.
The second key characteristic is that this project employs three distinct case studies, and as such, it qualifies as a multiple case study design (Yin, 2009). In order to assess the robustness of the open-source investigation method, three successful open-open-source investigations, that are considered ‘good practice’ within the OSINT community, are chosen as case studies. ‘Successful’ investigations are chosen because this project is interested in exploring the potential of the OSINT research method rather than assessing the robustness of a representative sample of OSINT investigations. In order for open-source investigations to be eligible as a case study in this research project, the subject of the investigation had to be ‘human rights violations a conflict context’. The investigated human rights violations had to have taken place in a developing country. In order to control for bias, the investigations needed to be carried out across various subjects, across various geographical contexts, and across various political actors. Eventually, the three investigations chosen as case studies are; The BBC Africa Eye investigation into extrajudicial killings in Cameroon; the Bellingcat investigation into ‘the social media executioner’ in Benghazi, Libya; and the Amnesty International investigation into civilian casualties in Raqqa, Syria.
The third characteristic that requires some elaboration is that the project does not have a singular unit
of analysis. Again, because of the methodological nature of the research, in each case study the unit
of analysis changes. This will become clearer in the empirical chapters, but as an example, in the Cameroonian case the unit of analysis is men in army fatigue executing 2 women and children, while in the Syrian case study (called “Strike Tracker”) the unit of analysis is geographical in nature, as the focus of this case is on individual buildings hit by Coalition airstrikes.
The fourth key characteristic of the design relates to OSINT’s methodological approach. As discussed in the previous chapter, OSINT employs a mixing of methods in a non-traditional way. The methodology is pragmatic in its approach, it uses “what works”. This means OSINT relies on both
22 quantitative and qualitative data that can be of both primary and secondary nature. As such, open-source investigation is considered mixed methods research (Creswell & Clark, 2017).
Overall, the research design of open-source investigation, and therefore also of this research project, is an advanced mixed methods research design called ‘concurrent triangulation design’, well suited for critically evaluating OSINT’s robustness (Creswell, Plano Clark, Gutmann, & Hanson, 2003). In a concurrent triangulation design, researchers use both quantitative and qualitative methods to cross-validate and corroborate findings within a single study. ‘Concurrent’ signifies that the data from the various techniques in each case study is analysed separately and then compared and combined (Creswell et al., 2003, p. 183). Ideally, in this research design, the priority would be equal between these two methodologies, but in practice, priority is skewed in either the direction of the quantitative or the qualitative methodology. In OSINT, priority is given through pragmatic decision-making. As this research replicates the open-source investigations, it is also prioritized through pragmatism. In Figure 2, a technical representation of the research design is presented.
23
3.3 Approach, Data Analysis and Operationalisation
The concept of robustness was unpacked and discussed in great detail in the previous chapter, in this section, I will operationalise the concept and discuss my approach to the data analysis.
The research method employed in this research project is called robustness analysis. In his seminal paper on robustness, Wimsatt (1981) describes a robustness analysis to include the following procedures:
1. To analyse all the various means of determination used to arrive at the conclusions and results for their validity and reliability.
2. To analyse whether the various means of determination are unchanging and identical in their conclusions and results.
3. To determine the limitations and boundary conditions of these conclusions and results. 4. To analyse and explain any relevant failures in the robustness of findings
Key to this robustness analysis is thus to check whether the various means of determination used are reliable and valid and whether the findings they yield can be triangulated and corroborated to arrive at robust conclusions and results. In order to do this, a robustness analysis entails the systematic, step-by-step replication of all the various means of determination used in the open-source investigation. The various ‘means of determination’ used in these case studies are detailed in the operationalization Table 1 under ‘OSINT techniques’. As OSINT values its transparency and replicability highly, open-source investigation reports describe the application of these various OSINT techniques in detail. Hence, I found the replication of these investigations to be straightforward. Any instance in the research process when the steps taken by investigators were unclear, personal communication was established successfully to clarify the procedures. I found investigators generally willing to corporate and easy to reach through social media.
As can be seen in the conceptual model in Figure 3, the first step in critically evaluating the robustness is evaluating the validity and reliability of the various OSINT techniques. In the operationalization Table, the indicators for the reliability and validity are presented per OSINT technique. The evaluation of the validity and reliability of the techniques is the main area where limitations, and thus potential improvements, come up. Hence, this first part of the conceptual model is enclosed by a dotted line which includes the keywords: limitations and improvements.
In the second step of the robustness analysis, corroboration and triangulation of the OSINT techniques and findings are performed. Corroboration is done by analysing whether the findings of each technique confirm a uniform conclusion. Triangulation is done by evaluating whether the findings of each
24 technique support the overall finding of the investigation. In the conceptual model, this requirement for the various means of determination to yield identical result is illustrated by the dotted enclosure with the words, borrowed from Wimsatt’s original text, ‘condition of invariant results’.
The more ‘means of determinations’ triangulating and corroborating a finding, the more robust a finding becomes. This causal relationship is illustrated by the ‘+’ sign in the conceptual model. A similar ‘+’ sign is present between reliability and validity. This is because the more reliable a means of determination is, the more likely it is to be valid. However, as this relationship is correlative rather than causal, it presented by a smaller ‘+’ sign.
25
Table 1: Operationalisation of the concept of Robustness across the dominant OSINT techniques in this research project.
OSINT Technique Concept Dimension Indicator
Photo and video analysis
R
O
B
U
S
T
N
E
S
S
Multiple means of determination Corroboration and Triangulation of different clues Reliability Quality of the imagery, indication its was doctored Validity Logical and effective in terms of its stated aim
Geolocation Multiple means of determination Corroboration and Triangulation of video footage with satellite imagery Reliability Quality and availability of imagery, the stability of observations
Validity Logical and effective in terms of its stated aim
Chronolocation Multiple means of determination Corroboration and Triangulation of video footage with satellite imagery Reliability Quality of imagery, procedural rigour, the stability of observations Validity Logical and effective in terms of its stated aim
Social Media Content Analysis Multiple means of determination Corroboration and Triangulation with video footage and satellite imagery
Reliability Indications of deceit
Validity Logical and effective in terms of its stated aim
Crowdsourcing Multiple means of determination Corroboration and Triangulation by multiple actors, some of which experts Reliability Stability of observations, intercoder reliability
26
3.4 Reflection on the research process and empirical chapters.
There are three important statements to be made on the empirical chapters before the reader commences reading them. First, it should be noted that the author was submerged in the online ‘open source community’ for over eight months. Although many of these activities did not lead to a direct collection of data for these case studies, it did lead to an enhanced understanding of the research method. These activities included, but were not limited to: partaking in investigation related discussions; organizing a ‘Hackathon’ at the UvA with over a dozen renowned open-source investigators from France, Germany, England, the USA and Scandinavia; meetings and dinners with investigators which effectively functioned as focus groups and informal interviews; digital ethnography (Murthy, 2008) was done through being present in investigators’ secure chat environments; partaking in weekly OSINT quizzes; familiarizing with techniques used in open-source investigation; etc.
Second, the critical evaluation of some techniques in the empirical case studies is sometimes really thorough. The author can imagine the reader experiencing a ‘So what?’ feeling in the initial read. However, in the findings section of each empirical chapter, these in-depth critical evaluations are brought to a higher level of abstraction; therein highlighting the importance of the detailed enquiry. Third, and perhaps most importantly; the empirical chapters include explicit pictures of people seconds before they were executed. This statement is meant as a warning, before each chapter, an additional warning is given. In one case, imagery is particularly explicit. These images are printed on a separate page without any text, a premature warning is presented so skipping this page is at the discretion of the reader. This chapter now continues with three additional statements on the research location, the limitations and the ethical considerations of this research.
27
3.5 Statement on the Research location
The main goal of this research project is to examine a potential research method that does not rely on ‘fieldwork’. Therefore, the location where this full research is carried out is distinctly different from nearly all other research projects in Development studies: the UvA library. In open-source investigation, the medium enabling access is ‘the Internet’. Hence, this research method can be carried out anywhere where a researcher is able to connect to the internet. Like my own research, the actual open-source investigations used as case studies in this research project were also carried out remotely. Nevertheless, the human rights violations investigated in the case studies were carried out in the conflict zone of Benghazi, Libya; Northern Cameroon and Raqqa, Syria. In order to have a close as possible feel for these conflict zones, the researcher longitudinally consulted numerous open sources throughout the research. This included reading NGO reports on the conditions in the countries, watching dozens of documentaries and tv reports, reading news articles, familiarizing with the environment in GIS software such as Google Earth and the daily tracking of social media (Instagram, Twitter, Facebook) data coming out of the cities and/or regions. This tracking was done by identifying and following open source experts on the three conflict areas, following dominant hashtags, and distilling data coming out of the area through using ‘search operators’. Below, an overview map of the locations in which the open-source investigations take place is presented.
28
3.6 Statement on the Limitations of this Research
The research criteria by which research is assessed are internal validity, external validity, reliability, and replicability (Bryman, Becker, & Sempik, 2008). As limitations are mainly derived from shortcomings in these areas, they will now be discussed in subsequent order. It is important to stress that these research criteria are discussed here in regard to this research project, not in regard to the OSINT methodology. First, internal validity is concerned with the question of whether a conclusion that incorporates a relationship between variables holds water (Bryman, 2012, p. 47). In the context of this project, internal validity is concerned with the question of whether the robustness of a research methodology can be determined by looking at its reliability, validity and various means of determination. Since the robustness analysis method is derived from widely accepted theory which is laid out in detail in this thesis, the internal validity of this project is considered high. Second, external validity is concerned with the question of whether the findings of a research project can be extrapolated beyond the scope of the project; this is often also referred to as ‘generalizability’. The main limitation for the generalizability of this research is the non-representative sample of case studies chosen in this project. However, as previously stated; it is not the goal of this research project to make decisive conclusions about the overall robustness of open-source investigation. Instead, this project aims to raise interesting questions about the limitations, robustness and applicability of OSINT in development studies in researching conflict contexts. By looking at not one, but three distinct case studies, across multiple dimensions, this project increases the generalizability of the questions and findings it is able to raise. Third, reliability is concerned with the question of whether the results of this research project are repeatable (Bryman, 2012, p. 46). In assuring its reliability, OSINT does not reason from an authoritative stance, but simply lays out the methods and findings and allows each observer to make up its own mind about the reliability of the conclusions. However, as such, OSINT is inherently subjected to an observer’s positionality and intrinsic biases. Hence, those observers that do not prescribe to the ontology and epistemology behind OSINT, might question its reliability a lot more than those who do prescribe to objectivism and positivism. Lastly, replicability is concerned with how well other scholars are capable of replicating this research project. Replicability comes in three stages: Is the data used in this research project open to other scholars? Are the procedures used in this research project clear to other scholars? Do other scholars arrive at the same findings when replicating this research project? (Bryman, 2012, p. 47) The first question is decisively answered yes: the fundamental idea of OSINT is that the investigation, as well as all the sources used, are accessible to everyone with a connection to the internet. The second question is again in the hands of peer reviewers; this whole chapter is meant to clarify the research set-up in detail. The third question is at the discretion of individual observers.
29
3.7 Statement on the Ethical Considerations in this Research
As a general rule, this research project acted from the ‘no-harm principle’. Meaning, the ethical considerations in this research project are focussed on not endangering or interfering with the lives of people who have and have not agreed to contribute to this research project.
Before discussing measures taken in this thesis, a short discussion of using open source data is needed, specifically social media data. In the article “Introducing Social Media Intelligence (SOCMINT)”, the question is asked: ‘what is private or public, and […] what counts as intrusive or not?’ (Omand, Bartlett, & Miller, 2012). One train of thought in this debate argues for a very judicial view of this dilemma: if people use social media services, they automatically agree to the terms of these service providers. In their ‘terms of agreement’, Twitter (2019), for example, declares that information shared through their platform is “‘information you are asking us to make public”. The other dominant train of thought takes a milder approach, arguing that it is unreasonable for users to be aware of all the possible ways their shared (open) data can be used. This train of thought argues for better protection of user data. While this thesis supports the second train of thought, this does bring forth additional difficulties. For example, imagine a video of an extrajudicial killing, in which multiple people are recognizable, is shared through social media. It would be unethical for this severe human rights violation to be ignored by the global community. However, at the same time, the investigation process might endanger the lives of innocent bystanders in the video. In this sense, OSINT has obligations to the victims as well as to the
potential perpetrators.
In this thesis, the only real ethical dilemma was about using the real names of the men that allegedly carried out the atrocities recorded in the videos. In consultation with my supervisor, I have decided to partly anonymise their names.
Also, in addition to the measures and considerations described above, I will be asking the Department of International Development Studies not to make my thesis public. Being submerged in the world of OSINT for the past eight months has taught me how such open-source data can be exploited and possibly abused. Therefore, I don’t feel comfortable with having my name directly linked with researching these sensitive case studies. Data that I am willing to share openly will be found in the ‘Introductory Guide to OSINT’ that comes with this thesis. As this is specifically designed for people interested in further exploring the OSINT methodology, this is of most use to other students anyway. Also, upon personal request, I am willing to consider sharing my full thesis.
30
4. Case study: BBC Africa Eye’s ‘Anatomy of a Killing’ (Cameroon, 2018)
Readers guide
This first part of this empirical chapter is descriptive: the case is introduced and the methods and data used are outlined in detail. The second part is a critical reflection of this case’s methods and data with a specific focus of the robustness of the findings. In the end, the findings of this case study are presented and placed in the broader context of research into human right violations in Development studies. Readers are warned: this case contains imagery captured moments leading up to the killing of two women and children.
4.1 Introduction:
In July 2018, a horrifying video in which two women and children were executed for allegedly belonging to Boko Haram began circulating on social media. Rumours accompanying the video on social media claimed the killing ground was either in Cameroon or Mali. On the 11th of July, the government of
Cameroon denied any links and branded it ‘Fake News’. The Cameroonian Minister of Communications told journalists “The video that is currently going round is nothing but an unfortunate attempt to distort actual facts and intoxicate the public. Its sincerity can easily be questioned …” (Kouagheu, 2018). BBC Africa Eye, the OSINT team of BBC Africa, analysed the footage to reveal the facts behind these murders. Using OSINT tactics and two clandestine local sources, the investigation was able to provide a very compelling narrative on where, when, why, and by whom these killings took place. Eventually, the Cameroonian Minister of Communication, Issa Tchiroma Bakary, issued a press release with the names of 7 soldiers that were arrested and under investigation for ‘extrajudicial killings’ (Bakary, 2018). The investigation was reported over the whole world. The ‘Twitter Thread’ laying out the methodology and the findings received over 15 million views and helped frame an international response to the atrocities happening in Cameroon (BBC Africa Eye, 2019). In February 2019, The United States military withdrew over 17 million dollars in security aid to Cameroon because of concerns about increasing violence and human rights violations across the country (Browne & Hansler, 2019). In April 2019, the European Parliament passed a strong resolution condemning the “torture, forced disappearances, extrajudicial killings perpetrated by governmental forces.” (European Partliament, 2019). One of the initiating members of parliament, Bodil Valero, explicitly mentioned the ‘Anatomy of a Killing’ investigation as a reason to address the issue (BBC Africa Eye, 2019). The investigation won both the prestigious Peabody award for Best Storytelling and the Royal Television award for News Technology (Peabody, 2019; Royal Television Society, 2019).
