Infrastructure of compassionate repression: making sense of biometrics in Kakuma refugee camp

My article focuses on the pilot of a Biometric Identity Management System (BIMS) for the distribution on in-kind aid in Kakuma refugee camp, in Kenya’s Turkana county, to examine the perception of biometric systems of verification by refugees. It explores how Somali refugees reflect on the implications of BIMS for their relations vis-à-vis humanitarian organizations, the Kenya state and other refugees, making sense of the humanitarian rationality tasked with both managing and policing populations in need. It thus argues that biopolitical technologies such as biometrics highlight, and heighten, the tension between care and surveillance as refugees challenge the official motives behind biometric infrastructures with counter-narratives situated in a specific sociopolitical milieu. Through an intense interpretative labor, which I captured in interviews and focus group discussions in Kakuma and Eastleigh, Nairobi, refugees open a crack in the apolitical veneer of humanitarianism, revealing, and challenging, the politics of biometrics.


Introduction
In June 2013, UNHCR and World Food Program (WFP) launched the pilot of a Biometric Identity Management System (BIMS) for the distribution of in-kind aid in a sector of Kakuma refugee camp, in Kenya's Turkana county. Refugee household heads in Kakuma 3 were urged to register their fingerprints to be identified at the general food distribution (GFD) points before collecting the packages with food and other basic items. The rollout of BIMS was expected to have 'significant costefficiency gains', as several UNHCR protection officers told me. 1 However, the enthusiasm of the implementers of the project was not universally shared in Kakuma. The launch of the pilot had spread anxiety among refugees in Kakuma 3 and beyond. 2 In meetings held in community centers, refugee leaders expressed their concerns to the UNHCR field officers who were explaining the logic behind this biometric transition. In October 2013, I attended two such meetings, one in Kakuma 1 and another in Kakuma 3. In both gatherings, elderly Somali refugees were particularly vocal. At the time of my fieldwork, Somalis were the second largest population of the camp after South Sudanese refugees. 3 Aid workers were taken aback when some elders suggested accompanying the introduction of the biometric registration with an increase in food rations. 'We speak different languages' 4 , a UNHCR protection officer told me with frustration after a meeting. In his view, the fact that the refugees would negotiate the acceptance of the biometric system in exchange for larger packages meant that they failed to grasp the rationale of the pilotwhich, put it as simply as the UNHCR did, was to allow only legitimate refugees to receive the food rations they were entitled to.
The Somali refugees' apprehension, nevertheless, had roots that predated the deployment of BIMS in Kakuma. Since Kenya gained independence from the British rule in 1963, the relation between the Kenyan state and its large Somali ethnic population, mostly concentrated in the North-East of the country, has been characterized by suspicion, often erupting into open violence. The outbreak of a civil war in neighboring Somalia in 1990 and the arrival of hundreds of thousands displaced people further stirred demographic anxieties in Kenya. More recently, in response to the threat of terrorism from Somali militants, the government of Kenya (GoK) has increasingly conflated refugee and security policies, reiterating its intention to have all Somali refugees in the country confined to refugee camps or back to Somalia. The contestation of BIMS thus suggested much more than a lack of understanding of the austerity logic of the project.
My article conceptualizes biometrics as an infrastructure of compassionate repression (Fassin, 2005) by asking: what are the factors that influence the way refugees understand biometric systems of verification? By exploring how Somali refugees reflect on the implications of BIMS for their relations vis-à-vis humanitarian organizations, the Kenya state and other refugees, I suggest that grappling with the anticipations and anxieties stirred by the launch of BIMS in Kakuma highlights the way refugees make sense of the 'humanitarian reason' (Fassin, 2012), tasked with both managing and policing populations in need.
My overall argument is that biopolitical technologies, such as biometrics, highlight, and heighten, the tension between care and surveillance as refugees challenge the official motives behind biometric infrastructures with situated counter-narratives. Through intense interpretative labor, which I captured in interviews and FGDs in Kakuma and Eastleigh, Nairobi, refugees open a crack in the apolitical veneer of humanitarianism, revealing, and challenging, the politics of biometric. The daily experience of BIMS and the anticipation of its long-term implications shape the Somali refugees' perception of the role of humanitarian agencies, unsettling the precarious balance between protection and control at the core of humanitarian governance. The participants to my study carried out a reflection on the implications of BIMS against a backdrop informed by the securitization of refugee policies in Kenya and the political economy of Kakuma, anticipating the so-called functioncreep of the biometric infrastructure. Furthermore, specific communities within the larger Somali refugee population of the camp emphasized that the implementation of BIMS would have exacerbated pre-existing power relations. My article focuses in particular on a group of Somali refugees in Kakuma, hailing from farming and agropastoralist communities from Southern Somalia, who have been living on the margins of the Somali society since pre-colonial times. Particularly victimized by militias of different factions after the Somali civil war broke out in 1991, and with no other option but seeking succour in UNHCR camps in Kenya, in the 1990s these refugees were lumped together by humanitarian workers under the classification of Somali Bantu, and relocated from Dadaab, the largest refugee camp in Kenya, to Kakuma. Their condition of socio-economic exclusion has continued into exile, due to a lack of access to employment and business opportunities, resulting in an overreliance on humanitarian aid.
In developing my argument, I draw on theoretical contributions spanning from critical humanitarianism studies to the literature on biometrics and humanitarian governance to infrastructure studies. I rely on critical ethnography as the methodology of choice to capture how Somali refugees in Kakuma understand the introduction of a biometric system of verification to distribute in-kind aid.
My article addresses two knowledge gaps in the relevant literature. On the one hand, despite the growing popularity of biometrics in humanitarian circles, there has been so far little discussion on the refugees' attitude towards biometric systems of identification. On the other, there is a dearth of analysis on how power differentials within specific refugee communities affect the acceptance and the interpretation of biometrics. I thus contribute to the topic of this special issue by suggesting that what makes a digital identity project political is not only the logic behind it but also the ways users make sense of it through narratives that are informed by historical trajectories and socio-political factors.
While empirically contributing to the discussion on humanitarian biometric systems of identification, at a theoretical level my article underlines the polysemic nature of biometric technologies meaning that different stakeholders may hold different, and even conflicting, views of themand highlights the needs for researchers to account for the interpretive labor that surrounds their implementation and day-to-day operation. I begin my article by reviewing and bridging the streams of literature on critical humanitarianism information systems applied to humanitarian governance and critical infrastructure studies. I then explain my methodology and describe the empirical setting of my study. After presenting the key themes emerging from my investigation, I discuss my findings and conclude with suggestions for future research.

Critical humanitarianism studies
The first body of literature I draw on is critical humanitarianism studies. Scholars such as Fassin (2012), Agier (2011), Ticktin (2011) andPallister-Wilkins (2018) have focused their work on the entanglement of policing and caring, applying a Foucaldian perspective to trace the genealogy and the evolution of an apparatus entrusted with efficiently managing people affected by displacement crises while protecting the existing order. This two-prong function is vividly embodied by biopolitical devices such as migration hotspots (Pallister-Wilkins, 2018) or refugee camps (Agier, 2011), established to both assist and contain populations displaced by natural or man-made disasters in remote locations (some Greek islands in the case of migration hotspots, the Turkana region in the case of Kakuma refugee camp) to spatially reinforce the juridical separation between citizens and refugees.
This scholarship acknowledges that humanitarianism is characterized by the fraught co-habitation of logics 'developed to both effectively manage disaster and to secure (in both senses of the word) imminently mobile populations for the maintenance of liberal order alongside and through the securing of life' (Pallister-Wilkins, 2018). Scholars who focus on how rules, practices and categories shape the entanglement of control and protection (Malkki, 2002;Rabinow & Rose, 2006) mobilize the notions of biopolitics, broadly defined as the management of life, and biopower, or 'the ensemble formed by institutions, procedures, analyses and reflections, calculations, and tactics' (Foucault, 2007, p. 144) targeting populations through apparatuses of security. In this view, humanitarian governance reflects the attempts to reconcile or, at least, mitigate the detrimental effects of this tension on the populations of concern. At the same time, humanitarian organizations provide aid 'to create and maintain the impression that humanitarian assistance is distant from state interests' (Pallister-Wilkins, 2018). The attempts to depoliticize humanitarianism, though, show their limits in what Newhouse (2015) calls 'humanitarian bargains' regulating the relations between donors and host states on the one hand, and humanitarian organizations and refugees on the other. While the former interpretation of humanitarian bargain refers to the pressures placed by donors on host states to contain migration flows, the latter refers to the trade-offs for refugees, being offered minimum support in exchange for confinement in camps. In both forms, there is an asymmetry of power between the parties involved.
As I argue in this article, which is mostly concerned with the latter type of humanitarian bargain, the integration of digital technologies in the management of life in refugee camps magnifies these power asymmetries between the humanitarian apparatus and the refugees.

Biometrics and humanitarian governance
The discussion on the integration of digital technologies and humanitarianism has grown increasingly nuanced in recent years, showing the influence of Foucault's work on biopolitics (2004) and security (2009). The literature on biometrics and humanitarian governance has shown as, particularly since September 11 2001, the usage of biometricsor the assemblage of technologies to measure, analyse and process 'the digital representation of unique biological data and behavioral traits' (Ajana, 2013a)has expanded into several areas, from surveillance and security (Amoore, 2006;Amoore & de Goede, 2008) to border control (Amoore & Hall, 2009;Crampton, 2019;Duran, 2010) to administration (Martin & Donovan, 2015). The application of biometrics to immigration and asylum procedures reflects a broader interest of state security actors to harness digital technologies for what Rose (1999) calls the 'securitization of identity' (240). Although this trend has gained momentum since the 1990s, the usage of technological assemblages to shape processes of inclusion and exclusion is a legacy of colonial administrative methods aimed at rendering legible, and therefore more manageable, the subjected populations. In colonial contexts, 'the recognizing of humans on the basis of intrinsic physical or behavioral traits' (Maguire, 2009, p. 9) was used for 'identification, classification, and hierarchisation of bodies, simultaneously demarcating populations' (Nishiyama, 2015;see also Pugliese, 2010).
Recent years have seen a proliferation of digital innovations enabling new approaches to address the 'identity gap' in the Global South. (Gelb & Clark, 2013;White et al., 2019). As a result of humanitarian budget constraints 5 , donors' calls for greater accountability, and falling technology costs, humanitarian agencies have embraced biometric systems of identification as strategic tools for the management of populations of concern, addressing the need to make the distribution of limited resource to refugees more efficient (Jacobsen, 2017). This popularity is due, according to an Oxfam report, to three types of benefits: identifiability and traceability of the people targeted for assistance; accuracy and integrity to reduce fraud and duplication; simplicity and efficacy to simplify registration and identification (The Engine Room and Oxfam, 2018; see also Pilegaard, 2011).
Biometric projects appear thus shaped by rationalitiesaccountability, audit, capitalism, solutionism and securitizationthat 'represent the parallel and often conflicting agendas of different stakeholders within the aid field' (Madianou, 2019: 586). Of particular interest for my article are the logics of solutionism and securitization and the way they influence the deployment of biometric initiatives in humanitarian spaces. Refugee camps are turned into living labs for experimentation without properly accounting for the risks associated with each technological component of the biometric assemblage (a concept with which Madianou refers to the integrated ensemble of biometrics, AI, and blockchain). Scott-Smith (2016) has coined the expression 'humanitarian neophilia' to define a techno-utopian ideology that 'combines an optimistic faith in the possibilities of technology with a commitment to the expansion of markets' (2230). Sustained by the growing importance of the private sector in humanitarianism, this trend has influenced discourses, strategies and priorities, as underlined by the attention that corporate-philanthropic actors are paying to the need to 'financially include' refugees (Tazzioli, 2019). Humanitarian organizations, international donors and NGOs are therefore spearheading a view of biometrics resting on the idea that technology is merely a means to achieve greater efficiency in aid distribution and refugee registration, with beneficial implications for humanitarian protection (Jacobsen & Fast, 2019; see also Read et al., 2016).
However, the diffusion of biometric technologies in humanitarianism has recently come under mounting criticism. These critiques mostly fall into four groups: reliability, referring to the possibility of false matches; reusability, or the repurposing of the biometrics data by other actors; security, related to the loss or theft of data; and social impact, referring to the unintended consequences of the datafication of identity, such as a greater risk of social exclusion (The Engine Room and Oxfam, 2018). Jacobsen and Fast (2019) draw attention on the 'agentic capacity' and the constitutive effects of technology to argue that digital technologies and data are transforming humanitarian governance, dismantling 'the boundaries between care and control, emancipation and domination' (158) and raising questions on the governance of both technological experimentation and their outcomes in humanitarian settings (ibid.). The major risk of looking at the mere material dimension of biometric infrastructures through which data are collected, stored and processed is the depoliticization of displacement and the exacerbation of 'power inequalities between refugees and humanitarian agencies' (Madianou, 2019: 583). The work of Btihaj Ajana (2013a, 2013b) bridges the discussions on critical humanitarianism and biometrics in humanitarian governance by conceptualizing biometric identity systems as socio-technical assemblages that reconfigure forms of identity, citizenship and belonging according to criteria of security and risk management (1). She also stresses the need to 'challenge the label of newness that is often stapled on [biometrics] and to draw attention to the fact that the body has for so long been the subject of control, measurement, classification and surveillance' (45). In fact, she argues that we need to situate biometrics within a history of practices and technologies of management of bodies viewed with suspicion by the political power, concluding that 'the digitalization aspect of biometrics has certainly intensified such processes and opened up the body to further dynamics of power and control.' (ibid.)

Infrastructure studies
In response to the apolitical reading of donors and humanitarian organizations hailing the efficiency and traceability afforded by digital infrastructures, the infrastructure studies literature is relevant to bring politics back in the picture. This debate acknowledges the dual nature of infrastructures as both 'things and relations between things' (Larkin, 2013: 329) and the way their materiality is entwined to biopolitics (Collier, 2011). Infrastructures spatialize and reproduce political rationalities, ideologies and ethics by 'constantly ranking, connecting, and segmenting spaces and people' (Larkin, 2004: 292; see also Collier, 2011;Donovan, 2015). Beyond the nuts and bolts of the infrastructures lie their relational nature, articulating the relation between power and subjects and among the users, and informing the strategies and aspirations of the users. The design, the governance and the daily operations of the infrastructures produce specific subjects who are disciplined, or compelled, to adopt behaviors and perform practices that comply with the classificatory logic of the actors behind the infrastructure. For instance, Von Schnitzler's (2008) study on the introduction of meters to regulate water provision in the Soweto township, South Africa, argues that this technological innovation has the twofold aim of reducing water by producing a 'calculating subject' (906). Infrastructures thus influence the way power is deployed, sociality is experienced, and agency is performed. These three aspects are entangled in Star's (1999) relational and ecological approach to infrastructuresmeaning that they are defined by the interactions they mediate and are integrated into a pre-existing environment. In the case of a specific type of infrastructure such as digital systems of identification for refugees, as I will show in this article, biometric technologies are designed to reformat a more rule-abiding and traceable refugee population.
The dynamics of power and control encapsulated in biometric infrastructures have different implications for different groups. Nevertheless, the unintended exclusionary aspects of biometrics systems, related to gender, cultural and power inequalities, are often neglected. The hype surrounding biometrics has hitherto eclipsed 'the lack of evidence demonstrating whether these technologies actually solve humanitarian problems' (The Engine Room and Oxfam, 2018). Latonero et al. (2019), for instance, argues that the technologies and processes involved in digital identity will not provide easy solutions in the migration and refugee context. Taylor and Broeders (2015) suggest that the lack of context that often accompanies the digitization of knowledge production risks increasing pre-existing vulnerabilities. Technologies that rely on identity data introduce a new sociotechnical layer that may exacerbate existing biases, discrimination, or power imbalances.
As I shall show in the remainder of the article, this aspect emerged against a backdrop characterized by the securitization of refugee policies, steeped in the county's postcolonial history. This volatile situation had implications also for my methodological choices and for the challenges I met in the field.

Methodology
I negotiated my access to the Somali refugee population in Kakuma, and in particular to the Somali Bantu community in Kakuma 3, through relationships I had established in the mostly Somali-inhabited estate of Eastleigh, Nairobi, before traveling to the camp. Indeed, this article is part of a broader research project on the reproduction of relations of power and solidarity within the Somali diaspora. My fieldwork in Kakuma did not begin with the intent of exploring Somali refugee attitude to biometrics. It started as an investigation of the mechanisms behind the reproduction of dominance and marginalization among Somalis in a humanitarian space. In particular, I was interested in understanding how power relations tracing back to pre-war Somalia, and further exacerbated by the eruption of the conflict, were re-spatialised in exile (Iazzolino, 2020). As I was doing fieldwork in Eastleigh, I was informed of the biometric pilot that UNHCR and WFP were rolling out in Kakuma during the same period by some of my informants. Hailing from the Somali Bantu community, they had moved to Nairobi from Kakuma in the first place to leave their food cards with members of their households still in the refugee camp. Now, they were dismayed by the fact that the adoption of biometric identification prevented their relatives in Kakuma from accessing muchneeded cash through the sale of extra food rations. I thus set off to Kakuma, where I relied on the introduction of an INGO and on the contacts with refugees I was given in Eastleigh.
I endeavored to grasp Somali refugees' nuanced feelings and multi-layered anxieties towards biometrics by relying on an interpretative approach, producing 'an understanding of the context of the information system, and the process whereby the information system influences and is influenced by its context' (Walsham, 1993, pp. 4-5) and acknowledging the existence of 'subjective meanings already there in the social world […] to reconstruct them, to understand them, to avoid distorting them, to use them as building-blocks in theorizing' (Goldkuhl, 2012, p. 138). I also drew inspiration from the critical ethnography literature (Denzin, 2001;Madison, 2012), effectively used in other studies to understand how marginalized people make sense and interact with ICTs in the light of their socio-economic and political conditions (Nemer, 2018;Rangaswamy & Nair, 2012). As Nemer (2018) suggests, this perspective places emphasis on the users of digital infrastructures, 'positing that people themselves define the value of these services in their lives.' (462) It is worth pointing out that my fieldwork in Kakuma started in the wake of a terror attack staged by the Somali Islamist militant organization al-Shabaab on Nairobi's Westgate mall on 21 September 2013 in which 71 people lost their lives. To quell the anxieties of the Kenyan citizens, the government of Kenya (GoK) threatened to repatriate Somali refugees to Somalia (BBC, 2013) and further securitized its refugee policy, based on confinement in camps. Before travelzing to Kakuma, I had received the approval of the GoK Department of Refugee Affairs and UNHCR officials on site. However, the exceptional circumstances following the Westgate terror attack added a further layer of scrutiny to my movements and meetings across the camp. 6 This article draws on interviews with 40 Somali refugees in Kakuma and 12 in Nairobi who regularly traveled to Kakuma to visit their families. Furthermore, I organized two mixed male/female focus group discussions in Kakuma. The face-to-face interviews and the FGDs were primarily aimed at identifying patterns of exploitation and dependence within the Somali community.
The fieldwork in Kakuma was conducted between October and November 2013, but the data I collected in the camp were matched against the knowledge I had gleaned during 11 months in Nairobi's Eastleigh area, also known as Little Mogadishu because of the large presence of Somali refugees and Kenyan Somalis, between April 2013 and March 2014, with subsequent visits in 2016, 2017 and 2018. I also draw on interviews with 14 Ethiopians, Congolese and South Sudanese refugees, Somali traders (10 in Kakuma and 23 in Eastleigh), and 12 UN and NGOs workers. As I interviewed humanitarian workers to understand their perspective on the biometric system of identification, I tried to capture the way the biometric assemblage did not reflect a monolithic logic of aid distribution but was rather the outcome of a negotiation between the UNHCR and the WFP.
While, in Eastleigh, I had the opportunity to regularly meet with my informants and collect data both in interviews and during everyday interactions, the data collection in Kakuma was exclusively based on semi-structured interviews. In both contexts, the interviews were conducted with the help of Somali translators, who also acted as researcher assistants. Given the sensitivity of the issues at the center of my study, I had extensively reflected on the kind of approach to be used in the field and how to maintain cultural and individual sensitivity before starting my fieldwork, in discussions with Somali leaders in Eastleigh, UNHCR protection officers, and agents of Refugees United, a Danish NGO operating in Kakuma 3 which provided me with contacts in the field. Being introduced by well-known and trusted members of the Somali community was crucial to ensure the informed consent of the participants and reassure them that I would have complied with the requirement of anonymity. In both interviews and FGDs, no participant was pressed to answer to my questions or join the discussion. No minor was interviewed or included in the FGDs. All the contributions were fully anonymised. The data collected during the interviews and FGDs were grouped into core themes and analysed using Nvivo qualitative software.
There were some clear limits to my data collection. In the FGDs, female participants were averagely less active than male in taking the floor, although some women stood up and lead the discussion. Also, some pieces of information were for me impossible to verify. For instance, many respondents argued that Kakuma hosted an undetermined number of Somali Kenyans who had been pushed by adverse circumstances (such as having lost their cattle because of drought) to seek humanitarian aid by claiming to have been displaced from Somalia. Although none of the refugees I met in both FGD and one-to-one interviews self-identified as a member of this category, several refugees and UNHCR protection officers referred to this group of crypto-Kenyan citizens when reflecting on the implications of the biometric program. Nevertheless, in my analysis, I stress that no data is available to corroborate this information.

Setting the context
A mix of control, mistrust and repression has characterized the historic Somali condition in Kenya visà-vis the Kenyan state since Independence. In 1963, when an insurgency in the Northern Frontier District, a semi-arid area that was once part of Somalia, before this latter was ceded to the British by the Italians, prompted Kenya's President Jomo Kenyatta to declare a state of emergency in the entire district. The ensuing three-year war, remembered as the shifta war after the local word for bandit that Kenyatta used to label the insurgents, was bound to leave an open sore between Kenya and its citizens of Somali origin (Weitzberg, 2017).
Since the onset of the Somali displacement to Kenya triggered by the collapse of the Somali state in 1991, advocacy organizations have criticized the Kenyan authorities for the treatment of Somali refugees, contributing to creating an enabling environment for the targeting of vulnerable refugees by predatory security officers (HRW, 2010;Otunnu, 1992;Teff & Yarnell, 2012). The rise of al-Shabaab in Somalia in 2006 further worsened the popular perception of Somalis in Kenya. Increasing insecurity in the borderlands and humanitarian spaces foreshadowed the risk of a spillover of the Somali conflict into Kenya. Security concerns seeped into refugee policies, to the point where public discourses in Kenya conflated the two issues and Somali refugees were increasingly portrayed by Kenyan politicians and media as a potential fifth column of al-Shabaab (Okungu, 2011). In the aftermath of the Westgate siege, the government of Uhuru Kenyatta agreed with its Somali counterpart and the UNHCR to the repatriation of more than 500,000 Somali refugees currently living in the country (UNHCR, 2015). 7 The announcement of repatriation sowed apprehension among refugees, both in urban areas and in the two Kenyan refugee camps, Dadaab and Kakuma.
Kakuma refugee camp is located in the outskirts of Kakuma town, the largest urban center in northern Turkana, along the road linking Kitale to the South Sudanese border, and is jointly managed by UNHCR and the Camp Manager's Office of the GoK Ministry of Refugee Affairs. At the time of my fieldwork, in 2013, the complex hosted approximately 180,000 refugees from 20 different countries in four camps: Kakuma 1, the first camp to be established, Kakuma 2 and Kakuma 3 (UNHCR, 2015). Kakuma 4 was established in December 2013, after my fieldwork, to accommodate South Sudanese displaced, who have since outnumbered Somali refugees. In November 2013, the UNHCR in Kakuma had registered 53,890 Somali refugees (26,609 female and 27,281 male). However, several humanitarian workers warned that this number, and any refugee population numbers, was not reliable as it was well known that an undefined number of refugees, despite being registered in the camp, lived elsewhere. Kakuma 1 was the business hub of the camp, also known as Mogadishu because of the commercial dominance of businesspeople from dominant Somali clans. 8 The area featured a variety of shopsmostly owned by Somalis and, in smaller part, Ethiopiansselling clothes, mobile phones and electronics, TV booths and drinking dens. Most of the goods traded in the camp were shipped from Nairobi, while local Turkana pastoralists provided the meat for sale in the area. The source of the cash circulating in the camp was mainly remittances or, as suggested by the presence of bags of rice and flour labeled WFP in shops in Nairobi, the sale of foodstuff.
It was also to curb the sale of food rations that, in June 2013, UNHCR and WFP launched BIMS in Kakuma. 9 The rationale of the project was elucidated in a report that a UNHCR protection officer shared with me. According to the document, The lack of a rigorous identification method leaves the process open to abuse. Ration cards may be passed on and used by non-refugees, or by refugees already holding a card. Some refugees unofficially return to their home countries and most deaths are unreported until verification exercises are held; in the meantime, the ration cards continue to be used to access services in the camps, including food. Ration cards also risk being stolen or taken from their legitimate owners and used to receive food rations until reporting and de-registering are completed. These abuses result in the sale of a portion of relief food on local markets. 10 The implementation of the BIMS pilot in Kakuma was meant to strengthen the identity verification system by introducing, alongside the scan of the barcode of the food card, also a fingerprint check. Figure 1.
The distribution procedure for food consisted of 5 steps: Subsequent reports have taken stock of the limits and shortcomings of BIMS and modified the model accordingly. According to a survey conducted by WFP in November 2014, 77 per cent of refugees polled were satisfied with how the new controls had been implemented, and 60 per cent responded that the new procedures made food distribution faster and more orderly. 11 However, as I verified in a follow up in 2016 among Somali refugees registered in Kakuma but living in Eastleigh, most of the issues I highlight in this paper have, to the date, not been fully addressed. In the next section, where I present the main findings from the fieldwork, I will discuss the perception of BIMS' implications for refugee livelihood.

Repatriation anxieties and limitation of movement
Somali refugees in Kakuma interviewed for this study expressed the fear that BIMS could be used to tighten police control over their movement. The launch of BIMS coincided with fresh announcements of the Kenyan government to repatriate Somali refugees. These announcements further reinforced the perception of BIMS as an instrument to prevent those who would be transferred to Somalia from re-entering Kenya and seeking assistance again as refugees. When, in the meeting in Kakuma 1, an authoritative Somali elder asked: 'What are you going to do with our fingerprints? What if we need to travel elsewhere and then we want to come back? 12 ', murmurs of approval filled the space of the gathering. His comments were often echoed during the interviews.
'What if we are forced to leave Somalia again? People there continue dying every day. We may go back and things could be fine for a while, but you never know what could happen in the future. Things are very uncertain' (Interview #14 13 ).
This concern were shared particularly by those refugees originating from areas in Somalia still ravaged by the Al Shabaab insurgency. During the FGDs in Kakuma 3, Somali Bantu refugees from the area of Southern Somalia comprised between the rivers Shabelle and Juba, at the time firmly under the control of the Islamist movement, were particularly anxious about the possibility of being denied assistance if they were forced to flee Somalia again. Among these was A., a 50-yearold from Buale, in Middle Juba who was held in great esteem by other Bantu refugees.
'I left my farm on the bank of the river when my mango trees were destroyed, my father was strangled and one of my three wives was raped. I have spent many years in Dadaab before being transferred here. I'm still waiting for resettlement. I speak sometimes with my brother, who is still in Buale, and he tells me that the situation is bad one day and worse the next. If I go back to Somalia I cannot return to my land. I will end up in a camp outside Mogadishu, which is worse than here. If I want to come back here I will have to burn my fingertips, otherwise I will be sent back again ' [FGD#2] A.'s reference to erasing his fingerprints by applying chemicals to or sanding off the fingertips to avoid the risk of being re-identified by the Kenyan security is common also to refugees registered in the camp but regularly traveling to urban centers. In an FGD in Kakuma, several participants, including local businessmen, stressed the importance of being able to travel to Nairobi for medical check-ups, attending school, or looking for business opportunities. During an FGD in Kakuma 1, several participants shared the anxieties of O., a Somali Darood refugee from Galkayo, in central Somalia.
'What happens if the police stop me in Eastleigh and check my fingerprints and realise that I should be in Kakuma? Will they cut my food rations? If they see that I don't live in Kakuma anymore, I may lose the right to resettlement. ' (FGD#1) There was a widespread concern was that, once their fingerprints were recorded in the camp, the refugees stopped and identified by the Kenyan police in Nairobi would be sanctioned, lose the right to reside in Kakuma and, with it, many believed, a fast-track to resettlement (although this belief, as confirmed by UNHCR officers, was unfounded). Also, some refugees pointed out that being registered in the camp would prevent them from seeking business opportunities elsewhere. One of these refugees was F., a young woman I first met in Kampala, Uganda, and then in Nairobi. Although still registered as a resident of Kakuma, F. regularly traveled between Kenya and Uganda, having built a reputation as a respected seller of fabrics in Eastleigh, where she had moved three years earlier, and later in Kisenyi, an area of the Ugandan capital (Iazzolino & Hersi, 2019).
Will they check my fingerprints at Busia [a crossing point between Kenya and Uganda] if they see that I'm a resident of Kakuma? Will they let me cross into Uganda or they will send me back to Kakuma? (Interview #3 14 ) Another young refugee interviewed in Kakuma 3 raised the concern that the UNHCR would share its database with the Kenyan government and the police would stop him on the way to Nairobi, where he regularly traveled to look for work, and send him back to Kakuma.

Identification of Kenyan Somali citizens
During the FGDs, several participants remarked that the UNHCR biometric system would allow singling out Kenyan Somali citizens. 15 It is widely known among both the humanitarian personnel and refugees in the camp that a large number of Kenyan Somali citizens claim to be refugees to benefit from the assistance of humanitarian organizations, although there are no data on the size of this population. In the course of my fieldwork, I have not had the opportunity to interview any resident of the camp self-identifying as a Kenyan citizen. However, UNHCR workers were aware that drought and other natural disasters regularly pushed ethnic Somali pastoralists from the North-Eastern Province of the country mostly to Dadaab and, to a limited extent, to Kakuma in search of assistance. Some of these humanitarian workers evoked fairness as the leading principle of the program, thus couching the rationale of the biometric shift in moral terms. As a UNHCR protection officer said, 'It is a matter of fairness, because not everybody is entitled to aid. Those who are claiming aid without being refugees, or using the card of someone else are robbing other refugees' (Interview #13 16 ) The representative of another international organization linked BIMS to the need to circumscribe the population of concern of the humanitarian system by identifying Kenyan Somalis trying to pass for people displaced from Somalia 'There are many Kenyans to claim to be refugees. I understand that they too need help, but it should be the state to support them, not us. We cannot do the job of the state. (Interview #15 17 ) The participants of an FGD in Kakuma 1 also suggested that Kenyan Somalis trying to pass for refugees would try to sand off their fingerprints to avoid being identified as Kenyan citizens.

Loss of income
Somali refugees from minority groups proved particularly concerned with the implications of the biometric verification for the possibility to access much-needed cash. The cash was used for diet diversification (like purchasing goat meat from Turkana pastoralists, firewood or charcoal) or invested in small entrepreneurship initiatives, such as setting up a shop. It was also necessary for purchasing phone credit, to pay for transport to the UNHCR compound and the market in Kakuma 1 and to travel to cities to undergo medical visits.
Although money transfer operators and mobile money kiosks were present in the camp, mostly clustered in Kakuma 1, not all refugees had equal access to remittances. The trade of food rations was a critical source of cash for those refugees, such as Somali Bantu refugees, with no or very limited access to remittances from abroad. Most Somali Bantu households practices a split household strategy, in which a household member, typically a young male, would move to the city looking for work. If he was also the registered head of the household, his card could be used by another member of the family and his food ration could be resold to the traders in Kakuma 1 or used to pay local Turkana pastoralists in exchange for goat meat. In some cases, it was an entire household to relocate elsewhere, leaving their food cards with relatives who could then collect and sell their food package. It was widely feared that the introduction of BIMS risked disrupting this widespread strategy.
A., a Somali Bantu refugee in his twenties, had recently moved with his spouse and young son to Eastleigh, where I met him. In the past, since he was registered as the head of the household, he had simply to leave his card with his mother, who still lived in Kakuma 3 with her other daughters.
We need that money because we don't have anybody to rely on. I can work here [in Eastleigh] as an English teacher but I don't earn enough to support my family and my mother. With my wife, we decided to move here together so that my mother could use our card and make a little money and open a little shop in the camp. Now she has to go through the procedure to become the head [of the household], but she has barely the money to take a piki-piki [mototaxi] to travel to the UNHCR office. How will she buy meat and pasta? She cannot even use our card. (Interview #5 18 ) In an FGD 19 held in Kakuma 3, Somali Bantu women confirmed that 1 Kg of rice or wheat flour provided by the UNHCR could be resold to traders in Kakuma 1 for 30 KSh. While it was possible to register the change of the household composition at the UNHCR compound, most refugees considered the procedure cumbersome and time-consuming. Moreover, as in the case of A.'s mother, traveling from Kakuma 3 to the UNHCR offices entailed paying for the transport.

Worsening conditions of indebtedness
The shift to the biometric system of verification thus deprived refugees left to their own devices of a critical source of income. As a result, several interviewees anticipated a worsening of relations of dependence and indebtedness that already tied many Somali Bantu refugees (mostly women) to Somali households from dominant clans in Kakuma 1 and 2. The chores that most participants to the FGD were required to do included cleaning latrines, plastering house walls and collecting firewood. One of the participants said that 'depending on the task, they would pay me up to 150 KSh (1.5 USD) per day, but often I have to work to repay debts for money I have already borrowed. (FGD #2 20 ) There was a widespread agreement that linking the distribution of food aid to the biometric verification was already forcing many women, who were no longer able to resell the extra food rations of absentee household members, to borrow money from better-off Somalis. Another participant to the FGD reported having worked for three months in the past to repay a 1000 KSh (10 USD) debt. 21 Since she could no longer use the food card of his brother and needed cash to purchase canned food and meat, she had recently sought help from the same family she had previously worked for. Despite having suffered harassment from this family in the past, she had refrained from reporting the abuses to the UNHCR because of the lack of alternative sources of income.

Discussion
The reflections around BIMS that refugees shared in the meetings and in the interviews upon which this article is built revealed the intense interpretative work on infrastructures, meaning 'different things for different groups' (Star, 1999).
Two main themes emerged from my ethnography: 1) the shift to a biometric system of identification was largely perceived as a constraint to refugees' agency and instrumental to tighten control on refugee mobility; 2) the implications of the efficiency logic underlying the deployment of the biometric system of identification were viewed as particularly alarming by historically marginalized groups within the Somali refugee community, who feared that their condition of subordination could be exacerbated.

Mobility constraints
Somali refugees framed BIMS as an extension of the Kenyan security apparatus. The refugees' perspectives went beyond the solutionist veneer to achieve greater efficiency and accountability to shed light on, and call into question, the underlying political rationality of the project (Larkin, 2013). The biometric verification system was largely perceived as an instrument of mobility control. Refugees anticipated what Ajana (2013a) calls biometrics' function creep, 'whereby techniquest hat were initially adopted for specific uses and purposes have beengradually spreading into much wider spheres and practices of governance' (576), placing the issue 'within a wider political perspective and in light of biometrics' spillover from exceptional domains and uses into the entire migrant/asylum body'(591). Furthermore, biometric technologies interact with what Newhouse (2015), in her study of South Sudanese refugees in Kakuma, calls 'the violence of verification', in which BIMS reinforce the impact of headcounts in 'fixing refugees in place and containing their movement' (2296). The perception of function creep of biometric verification infrastructures is not uncommon in humanitarian crisis, as recent cases in camps for displaced Rohingyas in Bangladesh (Thomas, 2018) and Yemeni in Yemen's Huthi-controlled areas (El Yaakoubi & Barrington, 2019) clearly show. The concern that sensitive information might be shared with actors that, in case of forced repatriation, would use it to prevent those fleeing Somalia from accessing humanitarian aid again shaped the position of the refugees vis-à-vis the biometric registration process.
The refugees' anticipation of function creep was not misplaced. Although none of the interviewees working for international organizations linked BIMS to security, the camp manager, a Kenyan Somali, was quite adamant that the logic behind the biometric system of identification was aligned with the GoK policy of encampment and with the renewed efforts, in the aftermath of the Westgate terror attack, to police refugees' movements.
'The policy is clear: refugees must reside in the camp. They are not supposed to travel outside the camp. There has been tolerance for a long time, but now the situation has changed. With what happened in Westgate we need to keep our eyes open.' (Interview #8 22 ) Furthermore, the camp manager anticipated that the GoK would have in the future a greater involvement in the management of BIMS and gain full access to the biometric data of the refugees. The GoK had been pressing international organizations to share refugee data for a long time because of both the national security agenda and donors conditionalities (Jacobsen, 2016(Jacobsen, , 2017. At the geopolitical level, the securitization of the Somali population in Kenya should be seen in the context of the broader global agenda that, in the post-September 11 scenarios, strengthened the security-development nexus (Faist, 2005). For a country such as Kenya, located on the front line of the so-called War on Terror because of its proximity to Somalia, a 'failed state' in most geopolical narratives and a potential breeding ground for terrorists in policy discourse, donors' conditionalities were attached to the commitment to side with the US in the fight against terrorism (Bachman & Hönke, 2010). For instance, a 2009 US cable released by Wikileaks revealed that the US government pressured the GoK to improve the identification of 'terrorists posing as refugees by biometrically register all refugees near the Somali border and to cross-check this data with the US' Terrorist Interdiction Program' (Jacobsen, 2017). After Kenya launched a military operation in southern Somalia against Al Shabab, thus triggering an escalating sequence of terror attacks and counter-terrorism measures targeting the Somali refugee population, the government pressure increased until, in 2017, Kenya's Refugee Affairs Secretariat (RAS), supported by UNHCR, took over the reception, registration, documentation, refugee status determination (RSD) and refugee management (UNHCR, 2020).

Reproduction of power relations
The dependency of most Somali Bantu on the sale of food rations of absentee household members made them particularly vulnerable to the unanticipated consequences of policies to curtail the diversion of humanitarian aid. The implementation of BIMS was thus regarded by most Somali Bantu participants as likely to exacerbate their condition of subordination to Somali traders from dominant clans. The importance of selling food rations helps explain why, as I mentioned in the introduction, some elders suggested accompanying the introduction of the biometric registration with an increase in food rations.
The rollout of the biometric food distribution program integrated the double aim of policing and protecting with the purpose to increase efficiency and effectivity by regulating who has access to assistance (Fassin, 2012). Humanitarian practices produce categories to render refugees more legible (Scott, 1998) and sift 'vulnerable victims' from 'cunning crooks' (Horst, 2006). This distinction also highlights the gulf between the normative assumptions of the humanitarian apparatus and the practices enacted by recipients of aid to increase their agency. Biometric assemblages are shaped by, and operate according to, procedures that incapsulate rigid moral categories, premised on what is licit and illicit, right and wrong, according to which humanitarian actors classify specific practices and behaviors. Yet, the moral economy on the ground was more nuanced. Playing the system may be seen as particularly necessary by members of some groups, such as Somali Bantu, who had very limited support from outside the camp in the form of remittances from the diaspora or business partnerships.
Forms of 'refugee cheating' have been an object of study in several humanitarian settings (Kibreab, 1991;Oka, 2014;Pérouse de Montclos & Mwangi Kagwanja, 2000). Kumsa (2006) interprets refugee cheating in terms of entrepreneurship, while Harrell-Bond (2004) suggests that 'perhaps the ubiquitous welfare cheating of the aid regime that occurs in refugee camps could be interpreted as evidence of the re-development of 'solidarity' among these populations ' (27). In Kakuma, Jansen (2008) observes, refugees were used to UNHCR and NGO employees and police officers accepting or asking for bribes -'TKK' was a well-known phrase: toa kitu kidogo in Kiswahili, meaning 'bring something small': it was part of the system as they knew it. Perhaps this explains why some people thought that the only way of organising resettlement was by cheating (575) As Newhouse (2015) argues, in refugee camps, 'illicit economic activities (such as brewing alcohol and selling rations) are at once material strategies to improve living standards, and practical political critiques that contest the sufficiency of humanitarian assistance '(2293). The anxiety of Somali refugees stemmed from the awareness that, by enforcing encampment as the mandatory condition to receive aid, BIMS limited their movements outside of the camp and, consequently, their ability to pursue livelihood opportunities and life aspirations. For communities with a long history of marginalization, such as the Somali Bantu, the crackdown on the food sale collided with the rhetoric of self-reliance muchtouted by aid agencies (Easton-Calabria & Omata, 2018), further worsening the dependency on humanitarian support and the vulnerability to exploitative working conditions at the service of better-off refugees.
The humanitarian workers interviewed for this study emphasized that BIMS was likely to increase efficiency minimizing the risk of aid diversion and better sort the legitimate recipients from the illegitimate ones. 23 Before being an object of interpretation by refugees, BIMS meant different things for the different stakeholders involved in its design and implementation. Moreover, the reflection on the pilot brought to the fore conflicting views on humanitarian governmentality among the stakeholders. Commenting on the aim of curbing the trade of foodstuff, a UNHCR officer said: WFP is very strict when it comes to money. They say no. For them, it's only foodstuff and other types of in-kind. We think differently because we know that refugees need money sometimes and not all of them receive help from abroad. But we need to find a compromise with the other agencies, with the government. The way we operate is given by a mediation among different agendas. We need to find common ground with the government to continue our job of assisting refugees in Kenya. (Interview #10 24 ) Moreover, reflecting on whether biometrics reduces fraud, a report by The Engine Room and Oxfam (2018) concludes that 'using biometrics on beneficiaries only allows for accountability checks to be done at the 'downstream' part of the process' (8), but does not address issues of aid diversion along the supply chain. Thus, biometrics risks putting 'the burden of accountability checks on the beneficiaries, when the real problems with fraud are elsewhere in the ecosystem' (ibid.).

Conclusions
In this article, I have examined how Somali refugees make sense of biometric infrastructures of verification. I have suggested that humanitarian actors deploy biometrics as a disciplining technology that reflects ideal categories of licit and illicit to determine who is entitled to aid and who is not. In so doing, they gloss over the way structural inequalities force members of specific communities to blur the boundaries of these categories. I have also sought to link the issue of biopolitical technologies to the broader discussion on the digitization and datafication of identity finalized to the allocation of services and resources. The unintended consequences of biometric-based initiatives to digitize identity are amplified by the ambiguous role of the refugee camp as a device shaped by multiple, and often diverging, rationalities and sources of authority, and characterized by a tension between protecting and controlling the population of concern. I have thus conceptualized biometrics as an infrastructure of compassionate repression to go beyond the materiality of infrastructures and to understand the significance of the socio-political context in shaping the way they are perceived by the stakeholders. In the state of exception that characterizes the condition of ethnic Somalis in Kenya, the perception of function creep further erodes the trust relationship between the humanitarian sector and the recipients of aid.
Despite the declared aim of harnessing the potential of biometrics to extend social protection and render 'legible' to the state segments of the population previously invisible, in some case this goal has proved to be controversial. There is, therefore, the need to approach biometric assemblages not simply as a technological solution to improve efficiency and accuracy in the management practices of information systems, but also as devices to reformat an otherwise complex reality to make it accountable to the donorsoften at the expense of the recipients of aid. Ignoring contextual dynamics when rolling out biometric systems for aid distribution in volatile settings rises critical issues as refugee policies are securitized by host states. Furthermore, the disciplining function of biometric systems of identification risks weighing particularly on communities for which playing the system is the only way to overcome livelihood constraints. This is why we need to broaden the gaze from the individual recipients to the social networks in which they are embedded, and upon which their livelihoods depend.
By suggesting that biometric systems of identification unevenly affect different socio-economic groups, I have argued that the case here examined throws in sharp relief the risk that the fixation on manageability would lead to the oversimplification of contextual, historical information, a critical issue that should be accounted for in debates on digital identity. This is particularly the case for those groups whose identity and livelihoods are entangled. By affecting the livelihoods and the socio-economic practices of specific groupslivelihoods and practices that are stigmatized by humanitarian actorsthe design, set up and management of infrastructures to digitize and verify identities risk contributing to the protracted marginalization of these communities. Neglecting the risk of function creep and the significance of entrenched relations of dominance and subordination in the design and implementation of biometric systems of identification can undermine the moral authority of humanitarianism.

Notes
South in reshaping processes of inclusion and exclusion and instigating new forms of organization and practices of contestation.