Who Is a Refugee? Uncertainty and Discretion in Asylum Decisions

Assessing claims for refugee status is a task often riddled with uncertainties, not least because of the challenge of establishing the credibility of the claims. The uncertainties enable divergent interpretations of both evidence and legal rules, thereby constituting a space for discretion in the refugee status determination process. This article explores how decision makers in the Norwegian asylum system reached a conviction about the outcome of asylum claims, despite numerous uncertainties. Decision makers who worked closely together developed a system of distinction that enabled them to single out applicants with protection needs from those who were not considered to be eligible. This system ordered the space for discretion, thereby reducing doubts about the outcome. It was based on the law and other formal sources, but also on recognizing patterns of difference and similarity with previous decisions. The emphasis on comparison between cases meant that the outcome of an individual application could not be understood in isolation; the distinctions between applicants who were accepted and those who were rejected depended, in part, on the case set as a whole. The findings suggest that, in a context of uncertainty, refugee status is to some extent determined by producing a local yardstick of who ‘the refugee’ is.


IN TROD UCTION
The unsettling truth is that in the overwhelming majority of cases, one will never know whether the decision was right or wrong. Refugee determination transpires under conditions of radical uncertainty. 1 The 'refugee' as a category of migrants with distinct rights developed under specific historical circumstances in post-war Europe. Although the category has been under continuous negotiation and development, the distinction between refugees and other migrants 2 has become increasingly important during the past two decades, when an increasing number of applicants for asylum have reached European borders. The distinction between refugees and other migrants is stressed by different actors in the field for different purposes: on the one hand, as a means to legitimize restrictive policies under the assumption that too many of those who claim to be refugees are 'mere' migrants who should be deterred and rejected; on the other hand, as a means to protect the institution of asylum from the very same restrictive forces, by referring to the importance of ensuring protection for bona fide refugees.
In reality, the boundaries between refugees and other migrants are often far from clear-cut. 3 Most migrants have mixed motives for leaving their home country that do not fit comfortably into a forced-voluntary dichotomy. 4 Moreover, the large discrepancies between European countries in recognition rates for certain groups of applicants from the same country illustrate that often there is not an obvious answer as to who will be recognized as a refugee. 5 One explanation for this is that even though Contracting States to the 1951 Refugee Convention base their legal frameworks on the same definition of the refugee, 6 the refugee status determination (RSD) process is organized in different ways in each national context, which may have significant implications for the way in which States assess eligibility for refugee status. 7 Such cross-national institutional differences cannot, however, explain why there is also a great deal of variation in recognition rates for the same groups of applicants within the same country. 8 In a Canadian study, Rehaag concludes that these differences could not be explained by the merits of the case or patterns in case assignments, but were largely a result of the 'luck of the draw' -the particular decision maker assigned to the case. 9 Asylum decisions appear, in other words, to involve a great deal of discretion, which means that decision makers can easily reach different conclusions about the same case. Apparently, then, it is not sufficient to study the national and institutional contexts in order to understand RSDs. It is also necessary to explore the finely tuned mechanisms that shape the decisions of those who assess claims for asylum.
The point of departure for this study was curiosity about what it is like to be on the inside of an institution that is responsible for identifying those who should be granted protection. How do decision makers exercise discretion in such a way that they reach a conviction about the outcome of asylum cases, in what Macklin has described as a context of 'radical uncertainty'? 10 To explore this question, this article draws on interviews with decision makers in the Norwegian Directorate of Immigration (UDI). As Macklin's statement at the beginning of this introduction indicates, the task is not an easy one. The assessment of refugee status is often made on the basis of uncertain information, and there is normally no way to assess whether the decision is correct. 11 Meanwhile, few bureaucratic decisions have comparable consequences: the decision may ultimately be a matter of life and death.
A substantial amount of research on RSD has been conducted from a legal perspective. 12 A social science approach contributes to another type of knowledge, described as 'law in action' . 13 Here, the law is considered to be one of several sources that shape legal decision making. Many of the mechanisms that are important for the exercise of discretion can only be examined from a bottom-up perspective, by talking to decision makers and exploring how their work is organized. In recent years, more scholars have examined the RSD process from this perspective, combining insights from migration studies, anthropology, the sociology of law, and organizational studies. 14 The current article contributes, first, to the field of refugee studies by giving an account of the RSD process from an inside perspective, and by demonstrating the importance of local practices for understanding the way in which refugee status is determined. Through their practices, decision makers who worked closely together established what can be termed a 'system of distinction' that enabled them to single out those who were deemed to be in need of protection from other applicants. Secondly, the article contributes to the sociology of law by illuminating the importance of comparison between cases in a decision-making process characterized by discretion. This means that the outcome of a case is difficult to understand in isolation, but rather depends on the case set as a whole. In a context of uncertainty, equal treatment may become the prime benchmark of what a 'right' decision looks like. While comparison produces consistency at a local level, however, the threshold for accepting cases may differ between decision makers who work with different case sets, even within the same institution. These findings may be interesting in other contexts where legal decision makers face many similar claims, the legal framework is broad, and the facts of the case are the main point of contention.
Drawing on data material from Norway raises questions about the relevance of the findings in this study beyond the Norwegian context. The RSD process takes place in States with different political and legal traditions. In a comparative analysis of the RSD process in Canada, the United States, and Australia, Hamlin refers to the institutions responsible for RSD, and the power dynamics between them, as different 'asylum regimes' . 15 Her study demonstrates that the autonomy of the administrative agencies responsible for the RSD process in different regimes depends, in part, on their level of insulation from the political and legal spheres. Such differences between asylum regimes will probably affect the conditions of decision making at street level as well; however, institutional autonomy and discretion among caseworkers are not equivalent. While institutional autonomy varies across migration regimes, the scope of discretion may vary within the same institution and change over time, depending on factors such as the experience and the 'institutional capital' of decision makers, 16 and the number of asylum seekers and the ensuing political attention. 17 While keeping in mind variation within and across contexts, a certain degree of discretion will inevitably be present at street level in all asylum regimes, because decision makers need flexibility in order to apply the law to individual circumstance. 18 This is particularly so in the field of asylum, where claims evolve continuously. 19 Moreover, recent studies illustrate that the dilemmas facing asylum decision makers across asylum regimes are strikingly similar: they have to make high-stake decisions on the basis of scarce and uncertain information; they need to strike a balance between the goals of protection and control; asylum law is ambiguous and provides limited guidance; and credibility is often a key point. 20 The indeterminacy that emanates from these conditions amplifies discretion and can never be fully contained by law or other policy measures that come from 'above' . Indeed, discretion is an important topic in several recent studies of the RSD process. 21 Because of these basic similarities, the findings in this article are likely to be relevant to asylum decision making in other countries.

THE R EFUGEE STAT US DETER M IN ATION PRO CE SS
The RSD process involves a number of challenges and uncertainties with which decision makers have to deal. First, the legal definition of the refugee contains ambiguous terms that have to be interpreted in order for decision makers to use the law in a concrete case. According to Norwegian law, a refugee is someone who: (a) has a well-founded fear of being persecuted for reasons of ethnicity, origin, skin colour, religion, nationality, membership of a particular social group or for reasons of political opinion, or (b) nevertheless faces a real risk of being subjected to a death penalty, torture or other inhuman or degrading treatment or punishment upon return to his or her country of origin. 22    latter is based on Norway's non-refoulement obligations under human rights law. Thus, the inclusion of the latter category in the refugee definition makes it broader in scope than the Convention definition to which most States adhere. Individuals who under Norwegian law are granted asylum under the second category will in most other States receive subsidiary protection, which usually entails more restrictions on rights such as family reunification.
If the applicant fulfils the criteria of the law, he or she shall be granted protection -at this point in the process, there is no space for discretion. 23 Because of this, decisions about asylum are in the Norwegian context considered to be 'bound by law' . There is, however, substantial discretion involved in determining whether or not the legal criteria can be considered to be fulfilled. This entails establishing the facts of the case and determining the risk of future persecution. For example, how much does it take to consider the fear of a particular applicant to be 'well founded'? Is it likely that the applicant may face treatment that can be characterized as 'persecution'? Although decision makers should heed international law and guidance from the United Nations High Commissioner for Refugees (UNHCR) pertaining to the Refugee Convention, much of the interpretative work remains in their hands. Moreover, a major challenge in asylum decisions is that the assessment of risk is oriented towards the future. This means that caseworkers have to try to predict what might happen if the applicant returns to the country of origin -an 'essay in hypothesis' . 24 This assessment rests on information about the applicant's home country which is often uncertain and ambiguous, and has to be applied to the situation of the individual applicant.
Secondly, since the great majority of asylum seekers lack documentation that corroborates their stories, the outcome often hinges on the credibility of the claim. 25 The credibility assessment constitutes perhaps the most important source of uncertainty in the process. Caseworkers use various tools in the assessment, such as whether the story appears to be coherent, plausible, consistent, and detailed, and whether the claim is in line with knowledge about the applicant's country of origin. The tools of credibility assessment do not, however, indicate 'how internally consistent, how externally consistent, or how plausible the applicant's story would need to be in order to be "credible"' . 26 Moreover, the challenge is that these indicators of credibility almost invariably can have other explanations. For example, researchers have demonstrated that difficulties with interpretation, anxiety, misunderstandings, cultural barriers, distrust, trauma, 23 This differs from, eg, the system in the United States, where such decisions are discretionary at this point as well, ie asylum may be granted if an applicant fulfils the criteria, but doing so is not required. Kate and the frailty of human memory are factors that can affect the ability of an applicant to explain him-or herself in a credible manner. 27 Applicants for asylum in Norway first lodge their applications with the police, and subsequently are summoned for an asylum interview at the UDI. The UDI is divided into different units that specialize in assessing asylum claims from specific countries. The system is highly centralized, with all asylum applications being processed at the headquarters in Oslo. The government can give the UDI political instructions about legal interpretation and the use of discretion on a general level, 28 but it cannot instruct decision makers about the outcome of individual cases. During the asylum interview, which often takes a full day, caseworkers talk to the applicant about the claim in the presence of an interpreter. Caseworkers in the UDI normally hold a master's degree but, like their counterparts in other countries, many do not have a background in law. 29 If the application is rejected, the decision can be appealed to the Immigration Appeals Board (UNE), a court-like administrative body. Immigration judges in the UNE are legal graduates, 'independent decision-makers who cannot be instructed in individual cases' . 30 A secretariat of caseworkers prepares cases for the judges. Although the lines between adversarial and inquisitorial proceedings are somewhat blurred, the appeals system in Norway can be placed in an inquisitorial tradition, promoting the ideal of an impartial decision maker who takes an active role in investigating the case. The political authorities have the power to influence the UNE through legal change and amendments, but unlike the UDI context, the government cannot instruct the UNE about legal interpretation or discretion on a general level. 31 If the case is rejected by the UNE, it can be brought to the regular courts at the applicant's own expense. This means that relatively few cases are brought at this level; moreover, Norwegian courts have generally been reluctant to overrule asylum assessments out of regard for the greater During a two-year-period after 2015, the government was able to do so. The change came together with a number of restrictive measures, but it was revoked again in 2017.
Uncertainty and Discretion in Asylum Decisions • Page 7 of 23 expertise of the UDI and the UNE in these matters, 32 although this has changed slightly in recent years. 33 Thus, while the courts weigh in only to a limited degree on the RSD process in Norway, the political level exerts more influence because the government can issue political instructions to the UDI.

DISCR ETION IN THE CON TE X T OF RO U TINE DECIS ION M A KIN G
The uncertainties and ambiguities throughout the RSD process enable a large degree of discretion, which leaves decision makers 'free to make a choice among possible courses of action or inaction' . 34 In other words, discretion can be connected to choice or a degree of autonomy in the decision-making process. Legal scholars have tended to focus on the discretion involved in interpreting the law and on the relationship between rules and discretion. 35 Studies in a social science tradition, on the other hand, have focused on the way various other contextual factors -organizational, psychological, political, and social -shape discretion. 36 From this perspective, discretion can be found not only at the very end of the decision-making process, when the outcome is determined, but at various points throughout. Moreover, discretion is not only related to interpretation of legal rules or to the way rules expand or confine the space for discretion; it is also connected to establishing the 'facts' of the case. 37 Thus, from a social science perspective, legal norms are considered to be one of many potential sources that enable and limit discretion. Decisions that seem highly discretionary from a legal point of view may be bound by social norms and routines in such a way that caseworkers do not feel that they have much discretion at all. As Lempert notes: 'What the law gives in discretion -that is, the authorization to reach one of a number of possible decisions and the awareness of this freedom -social forces may take away' . 38 In order to understand the boundaries of discretion, it is therefore, according to Galligan, indispensable to take the 'internal' point of view -that is, to explore the 'senses of discretion' of officials. An important social force that shapes discretion is socialization into a work community. Officials normally do not exercise discretion on their own -their discretionary powers are usually shaped by the views of colleagues and superiors. 40 The exercise of discretion is often a social undertaking that requires a shared perception of what a 'good' and 'correct' decision looks like. 41 Thus, instead of taking the individual decision makers as the point of departure, which may lead us to incorrectly focus too much on subjectivity, it is necessary to take into account the social context of decision making, and thereby inter-subjectivity.
Similarly, it is often misleading to focus on individual cases, since the decision in one case often depends heavily on what has been done in previous, similar cases. 42 As opposed to judges who make decisions on a wide range of different cases, decision makers in public bureaucracies often make decisions on many similar cases that involve the same kinds of legal and factual issues. This tends to produce shorthand ways of dealing with and categorizing cases that remain fairly stable over time. 43 How things have been done previously thus becomes a powerful influence on present decision making. 44 This kind of 'decision making by precedent' tends to take the shape of 'customs, traditions, conventions, routine practices and so on' . 45 Establishing routine ways of categorizing and assessing cases is one of the most significant sources of discretionary power among decision makers who apply the law routinely in State bureaucracies, since 'deciding in general how to decide in specific cases' 46 can potentially affect the outcome in a large number of applications. Once categories and routines have been set up, and precedents for how 'normal cases' are treated have been established, these constitute an important means of ordering and limiting the discretionary space. They can thereby be considered to function as 'enabling constraints' that help 'to simplify and facilitate choice' . 47 In other words, discretion is often exercised in ways that, in turn, limit and reduce the need for discretion. 48 Etymologically, 'discretion' stems from the Latin 'discretionem', which means the 'power to make distinctions' . 49 This meaning goes to the core of what asylum caseworkers do: they have to find ways to distinguish those who should be granted protection from other applicants who are not considered to be eligible. This article will 40 Miaz (n 14) 377. 41 Affolter, Miaz, and Poertner (n 29 refer to the interpretations, categories, and routines that limit doubts as a 'system of distinction' . As the empirical analysis will illustrate, this system consisted of interpretation of rules, regulations, and other kinds of information, but also of tacit knowledge that largely rested on experience and comparison with similar cases.

M ETHOD OLO G Y, DATA , A ND A N A LYS IS
This article is based on a doctoral thesis in sociology about the RSD process in Norway. 50 Twenty-four interviews were conducted with caseworkers in the UDI, mostly in late 2012 and early 2013. The data material also included a review of the documents in 40 asylum case files and observation of asylum cases in court. This article is based on the interviews with caseworkers. The project was approved by the Norwegian Centre for Research Data (NSD); a permit was also obtained from the Council for Confidentiality and Research.
After obtaining formal access from the UDI, the project was presented to caseworkers in some of the units and then potential participants were contacted by email. Some interviewees (six in total) were also recruited through the author's personal contacts. Most interviewees in this group no longer worked at the UDI. This mixed recruitment strategy contributed to maintaining anonymity for the caseworkers.
Most of the caseworkers were educated in the social sciences; a few had degrees in the humanities. Only one interviewee was a law graduate. Seven of the caseworkers were men and 17 were women; most were aged in their thirties. The distribution of men and women fits the overall pattern found generally in the UDI, where in recent years there have been approximately 30 per cent male and 70 per cent female employees. 51 Participants' length of employment with the UDI varied between three months and 13 years; approximately half the interviewees had worked there between two and four years. Caseworkers had experience with assessments of asylum applications from 10 different countries.
Prior to the interviews, the author developed an interview guide that was reworked during the course of the data collection. The goal was to learn from the interviewees how they conducted their work and to access participants' 'work knowledge' 52 -that is, to obtain detailed information about both formal and informal aspects of everyday work processes. This involved asking concrete questions about an individual's work, and probing to ensure that the actual meaning was understood -practically speakingof the terms and categories in use.
Most interviews lasted between one and two hours. The interviews were recorded, transcribed, and coded in Nvivo, an electronic software for analysing qualitative data. The coding was largely done thematically, in order to systematize the topics that emerged during the interviews. The analysis itself was the result of going back and forth between writing about the empirical material in a detailed and descriptive manner, reading theory and previous research, and integrating these insights with the empirical 50 Liodden (n 17). material. All quotations from interviewees in the article have been translated from Norwegian by the author.
Confidentiality was a major concern throughout the writing process. It was addressed at two levels: vis-à-vis an internal and an external audience. 53 External confidentiality was ensured by removing names and personal information about interviewees from the data. Safeguarding internal confidentiality was more challenging. It entailed taking measures to avoid identification of interviewees by colleagues and others who knew the actors in the field from the 'inside' . This was crucial for the interviewees. In response to this concern, the age of the caseworkers and the country with which each person worked are not mentioned. Moreover, the passage of time since the interviews has also aided anonymization. The UDI has a high staff turnover and undergoes frequent reorganizations. It is likely that some caseworkers who participated in the study no longer work at the UDI or may now work in different units.

'PR ACTICE': A S YSTE M OF DISTIN CTION
The most difficult thing is being as certain as you can be that you have made the right decision. That's the hardest thing, the uncertainty. That you never get a clear answer about whether what you have been doing during the past five years is right or wrong.
In this quotation, a decision maker expressed the uncertainty she faced every day in her work. Although some caseworkers echoed her sentiments, uncertainty and doubt were less important themes in the interviews than might have been expected. Overall, most caseworkers seemed relatively confident about the outcome of the majority of cases. When asked how they made decisions, they often answered, simply, that they 'followed practice' . Drawing on the framework outlined above, practice can be considered to function as 'a system of distinction' that enabled decision makers to order and manage the discretionary space.
Formally speaking, the 'practice' of public institutions is a legal source of lesser status than the law and its preparatory materials that can guide decision makers' interpretation of general rules and legal provisions. 54 Basically, it is made up of knowledge about previous decisions in similar cases, which serve as precedents for future decisions. Since 2008, UDI units have regularly published updated 'Practice Notes' for the most important applicant countries of origin on the UDI website. The notes are a compilation of information about a given country, the legal framework, and a description of the most common claims and how they are usually assessed. The notes have a similar function to 'secondary implementation rules' in the Swiss asylum determination procedure and to 'internal guidelines' in the German system. Changes in practice that could have substantial consequences for the number of acceptances had to be approved by the Ministry of Justice (now the Ministry of Justice and Public Security), but several interviewees emphasized that caseworkers as a collective had a great deal of influence over the way practice developed because of their country expertise. This influence seemed, however, to be somewhat contingent on which country they worked with. If there were a large number of arrivals -and thereby increased political attention -some caseworkers described practice as largely beyond their influence, even as opaque. Conversely, if they worked with 'small' countries, from which there were few applicants, there seemed to be more scope for influence over practice at the caseworker level. Perceived influence also depended on the status of the caseworker. Experienced caseworkers who had worked in the units for a longer time -and thereby had more 'institutional capital' 56 -considered themselves to hold more authority and influence over practice than inexperienced colleagues. Thus, the ability to determine 'in general how to decide in specific cases' 57 varied, depending on the specific conditions in the unit and the experience of caseworkers.
Once practice had been settled, it meant that country information, the legal framework, and political instructions had been interpreted in such a way that they could be applied directly to the practical realities that caseworkers dealt with. In decision letters, practice was further crystallized in the form of 'autotexts' -standardized paragraphs containing arguments about country information or the law, which could be pasted directly into the document. In this way, practice relieved caseworkers from having to make sense of background information individually, which saved time and resources, and reduced the risk of different people reaching different conclusions about similar cases.
It was not, however, possible to understand practice only by reading the Practice Notes, country reports, and legal texts. In addition to the formal sources of practice outlined above, practice seemed to involve a certain degree of tacit knowledge 58 that was difficult to articulate. As one caseworker put it, 'there's no one who can put that into words, what kind of knowledge we have' . In her 2011 study on asylum decision making in the United Kingdom and Spain, Jubany similarly found that caseworkers relied heavily on 'professional instinct': 'The fact that an officer "just knows" is in most cases as far as the rationalization will go' . 59 Her interviewees explained that these skills had to be acquired through hands-on work with cases. 60 Although this tacit knowledge may be related to several issues in the decision-making process, it appeared to be particularly tied to the assessment of credibility. This resonates with the view of caseworkers in Sweden who considered the assessment of credibility to be an 'almost mystical' skill 56 Miaz (n 14) 386. 57 Hawkins (n 46) 86. that could be acquired over time. 61 Similarly, Affolter found that while caseworkers in Switzerland often could not articulate why they found a claim credible, they had a clear conviction that was underpinned by a 'feeling' . 62

Comparison and pattern recognition
Considering the findings above, one could easily draw the conclusion that these are simply different ways of saying that caseworkers often rely on 'gut feeling' in their work and, thereby, that the decisions are highly arbitrary. Although subjective perceptions probably can play a role, the 'feeling' caseworkers referred to seemed to be shared by colleagues who worked closely together. In their study of the Swiss asylum system, Affolter, Miaz, and Poertner describe how caseworkers in an office belong to different 'communities of interpretation' . Members of these communities hold common norms about how they should conduct their work and what a just decision looks like. 63 New decision makers are socialized into an 'institutional habitus' that they share with their co-workers. 64 In Norway, the existence of such communities of interpretation was particularly visible with the arrival of new caseworkers to a unit, or when caseworkers transferred from one unit to another. Learning how to assess cases according to the norms in the unit was very much a social enterprise. As Miaz outlines, during 'the process of institutional socialization, caseworkers internalize institutional categories of thinking and ways of working, which guide their perceptions of the cases' . 65 When new caseworkers learned practice, they acquired understandings that to a large extent had been settled by more experienced members in their community of interpretation.
In Norway, there are normally two caseworkers involved in each decision. 66 The person who drafts the decision is described as the 'first hand' and the colleague who checks the decision is the 'second hand' . New caseworkers take the position of the first hand when they begin. Several interviewees explained that they often experienced considerable doubts about the outcome of cases when they were new in the job. One In some countries, such as Germany, caseworkers work individually. As Schittenhelm and Schneider (n 14) point out, however, individual caseworkers also appear to rely on different forms of tacit knowledge in their evaluations of asylum applications. In other countries, such as France (Dequen (n 14) 467) and Switzerland (Miaz (n 14) 386), superiors are involved in the decision-making process. The number of people involved, whether they are at the same level in the organization or hierarchically positioned, and whether or not it is known beforehand who will check or approve decisions, are factors that may influence the exercise of discretion.
Uncertainty and Discretion in Asylum Decisions • Page 13 of 23 caseworker described the trepidation she felt when she started in the position, because she often had no idea if the outcome she had reached was 'correct': 'I could sometimes get the feeling that I never knew what kind of feedback I would get from the second hand. It was like: I simply did not get it!' When the interviewees spoke about their trajectory from inexperienced to experienced, some described it as a process of learning how to become less gullible. At the beginning, they were inclined to accept claims that their colleagues would reject. 67 As one caseworker noted: '[Y]ou start from scratch, for all I knew everything could be plausible, you have nothing to compare it with, so people could say one thing or another thing, I was not able to say how credible or likely it was' . She stressed the importance of comparison -as long as she did not have experience with other cases and knowledge about their outcomes, the credibility of the claim was difficult to establish. Another caseworker explained how she gradually learned to 'read' the outcome of cases by making decisions that, in turn, were corrected by her supervisor: You had to try a bit and see. Sometimes I felt like it [the outcome] really could go both ways, but I simply had to settle on one side. Rely on a gut feeling and then see what the second hand said, and through that I got some guidance about what was an acceptance and what was not, and [the second hand] explained why and things like that. So little by little I managed to straighten things [out] in my own mind. But you depend a lot on receiving good training from those who are experienced. I think you cannot learn it on your own. Over time, she learned where to draw the line -how to make distinctions according to norms which she learned from her experienced colleagues. This resembles the findings of Affolter, Miaz, and Poertner in the Swiss asylum office, where a 'correct decision is often understood by decision makers as one that one manages to get past the superior' . 68 When new caseworkers over time made decisions that were in line with the settled ways in the unit and with the views of superiors, they increasingly felt more competent, and much of the uncertainty related to assessing cases decreased. 69 So, too, did the sense of individual discretion, which was tempered by a 'feeling that is quite similar among caseworkers' and the principle of equal treatment: You work up a kind of feeling that is quite similar among caseworkers in a unit, about what's okay and what's not. You discuss a sufficient number of cases in order to obtain that kind of feeling. And that feeling probably overshadows the sense that you're sitting down and thinking 'now I'll use discretion' . 67 This implies that caseworkers appeared to become stricter in their judgments with experience, as they became socialized into their role. Relatedly, Affolter, Miaz, and Poertner, (n 29) 281, found that in the Swiss system being lenient or naïve was considered unprofessional. In his study of the RSD process in Switzerland, Miaz also suggests that positive decisions were scrutinized more by superiors than negative ones, (n 14) 390. Becoming professional in the role as a caseworker appears -at least in some contexts -to be related to an emphasis on strictness in the assessments. 68 Affolter, Miaz, and Poertner (n 29) 268. 69 See also Schittenhelm and Schneider (n 14) 1703-04.
Experience with other, similar cases -and their known outcomes -constituted the horizon against which new cases were assessed. The contours of a decision became visible against this backdrop. The description of these more subtle aspects of practice bears similarities to what March and Simon refer to as the 'logic of appropriateness' that is intimately related to recognizing patterns in a situation and applying rules that have been used in similar situations previously. 70 Simon has suggested that intuition, and particularly the kind of intuition connected to expertise, can be conceptualized as pattern recognition. 71 Caseworkers often assessed many cases with similar claims. Over time, they seemed to acquire a 'sense of the outcome' that -at least in part -was based on recognizing patterns in the case that were similar to, or departed from, previous ones. 72 The conviction about the outcome that comparison produced could not, however, be used to justify a decision. 73 The reasons need to be 'public' -the kinds of arguments that are legally legitimate and understandable to an external audience. As Molander, Grimen, and Eriksen note, however, discretionary judgments are often considered to 'express tacit knowledge, intuition, holistic thinking, bodily sensibilities, receptiveness and many other things. Such factors, at best, generate non-public reasons, that is reasons that cannot be assessed by others' . 74 When caseworkers talked about justifying their decision, they often described what seemed to be a tension between such public and non-public reasons. They had a sense of the outcome that rested largely on experience and comparison between cases, but it could be difficult to find sufficient public reasons to justify the decision. Determining the outcome and finding justifications for that outcome sometimes appeared to belong to two different processes. A caseworker described this disjuncture: Interviewer: Do you often feel uncertain or are you usually quite certain now that you've been in this job for quite a long time, do you see quite quickly which way the case is going?
Caseworker: Most of the time I do, and then it happens from time to time that I am uncertain and then I have to run it through with the second hand. Then I may be uncertain about the outcome and perhaps most often I am completely blank on how I will be able to justify a rejection. It happens, not often, but it happens that: 'This is a rejection, but I just don't know how I will justify it' .
Caseworker: It's kind of the totality of it all. And usually, once you begin to discuss it with someone, you've got this thing and that thing and this does not make sense and that does not make sense. But I guess I get this feeling because . 73 See also Affolter's discussion about the need to produce 'on file' facts to justify the 'feeling' caseworkers had about the outcome. Affolter (n 14) 71-78. I've worked here for a while. Kind of what I mentioned earlier that when you've got people with the same kind of story who are simply so much more credible and who sort of substantiated it in a much better way.
In this context, the 'totality of it' seems, at least in part, to refer to tacit knowledge that is connected to comparison with other claims. Several caseworkers said that, although they might have a strong 'hunch' about a rejection, they had to accept the applicant if they could not mobilize sufficient arguments to support a rejection. In Norway, evidentiary assessment is, however, guided by the principle of 'free assessment' , which means that the weight of a piece of evidence is not assigned a priori, but depends on the other available evidence. 75 In the literature on credibility assessments in asylum cases, this is referred to as decision making based 'on the totality of the evidence' or 'in the round' . 76 In this context, it is likely that impressions or reasons that are 'non-public' influence the weight given to different pieces of evidence and thereby the view of the totality. Thus, an element that seems relatively insignificant may appear to have been given quite a lot of weight, but the force of the argument does not necessarily depend on that particular element, but rather rests on a conviction about the outcome that cannot be articulated directly.

Case-set effects
The previous section illustrated, in line with Hawkins, 77 that discretionary decisions are often not handled individually, but collegially; moreover, it showed the centrality of comparison with similar cases in the past in the decision-making process, turning equal treatment into a shorthand for the 'right' decision. The question is: what consequences does comparison between cases have for the assessment of credibility and, ultimately, for the outcome of the case? As Emerson proposes, in a fascinating article, such comparison may result in what he labels 'case-set effects' . 78 In line with the findings above, Emerson suggests that decision makers rarely consider each case as a discrete unit that is treated independently of other cases. He describes how the overall stream of cases provides a backdrop against which the assessment of any individual case will be made. The distinctions that decision makers create in a given case set will depend on how many other similar cases there are and how they are perceived.
As an example, Emerson refers to a study described by Freidson about the judgments of doctors about children's need for tonsillectomy operations. 79 From a total of 389 children, the operation was judged necessary in 174 cases, with no need to operate 75 Schjatvet (n 61). on the remaining 215 children. When another group of doctors was given the task of evaluating these 215 children about their need for a tonsillectomy, almost half were judged to be in need of an operation. Freidson concludes that the doctors employed a 'sliding scale of severity' in their assessments, which resulted in different distinctions being made, depending upon the case set as a whole.
In the context of asylum decision making, this means that the case set as a whole may contribute to a slight shift in the thresholds of what is perceived as sufficiently credible or not, or what is considered to be a sufficient risk of persecution. One caseworker reflected on some of the effects this comparative approach might have -returning again to the principle of equal treatment as the bottom line: If each case were very different -and they are not, they are very similar -but if each case were very different …, then you might assess them in a very individual manner every time, and you would not relate that much to a kind of threshold. If you assess many cases that are similar at the same time, you might assess them more similarly, and that might mean that it gets stricter. It shouldn't be the case, but it might be. At least it's equal treatment. And that's good.
One explanation for the increased strictness may be that credibility rests in large part on the perception of authenticity -that the applicant has a 'story of his or her own' . A stream of similar claims can be considered to be a symptom of the fact that regular patterns of persecution produce similar claims. It can also be considered a sign that asylum seekers have picked up a 'successful story' and try to reproduce it in order to be accepted. This means that a credible applicant may look quite different, depending on the context in which the claim is read. One caseworker noted the difficulty of distinguishing real cases from fabricated ones: 'We know we are being tricked when we receive one, two, three, four, five, six, seven, eight, nine, ten applicants who say [the same thing]. But in which case is it true?' Many similar stories will -at least after a while -come across as fabricated if they do not appear in a shape that underscores a personal dimension that has not been encountered previously. In other words, stories down the line are probably evaluated according to somewhat different criteria than those at the beginning, because of the case-set effect. Another caseworker explained how experience with many similar claims affected her assessment: Interviewer: How does that affect you? If you've got many similar claims?
Caseworker: It depends on how special they are. Sometimes a case comes along and -wow, this is a bit new! It can count in favour of the applicant: this is new! And that can make it more credible in a way, unless four claims that are similar come along during the next few days, then it [credibility] decreases again and then we pay extra close attention.
Stories that are similar and do not offer anything new compared to previous claims will easily be deemed less credible. This is not to suggest that the caseworker above incorrectly labelled these stories as fabrications -her hypothesis may well be true. It can, however, be challenging to differentiate fabrications from more general case-set effects, where the view of any one case is linked to the larger stream of cases.
The case-set effect can also be related to the relative distribution of acceptances and rejections for claims from a specific country. One caseworker said that decision makers in the UNE were stricter than caseworkers in the UDI because they only saw rejections: Let's imagine that knowledge is based on experience. They [UNE] work with rejections, cases that we have rejected already. That's the empirical data that fills them every day. They don't see the convincing cases that we see. That probably affects their hermeneutics. They establish a kind of tacit knowledge that heightens their threshold, which means that if they had in turn assessed cases in the first round, they probably would have rejected more of them.
She suggested, in other words, that decision makers in the UNE might have a different threshold for accepting cases because they have a different case set, one largely comprising rejections. In the same way, she considered caseworkers who worked with case sets where the majority of cases were rejected to have slightly different work modes from their colleagues who worked with case sets that mostly involved acceptances: People are creatures of habit and if you are used to rejecting, it can be more difficult to accept -the threshold might get higher. If you constantly question credibility and sense that things are not right, it can be difficult to see the things that are right. It's the same thing for units that largely work with acceptances.
If, over time, caseworkers assess cases from the same country, they may approach new cases with an attitude that has been shaped under circumstances that no longer apply. Different approaches shaped by different case sets became particularly visible if someone changed units or if they received help from colleagues with whom they did not talk regularly. When this happened, caseworkers came face to face with different ways of assessing cases that were at odds with the way they normally worked. One interviewee said that she was 'startled' to see how colleagues who worked with cases from a different country drafted rejections, and described them as 'very, very, very strict' . Another caseworker described a similar situation: I have supervised a new caseworker now who came from a unit with a lot of rejections. We have a lot of acceptances. And that caseworker wanted to reject everything. It was really interesting, because, what on earth: 'No, no, no! You can't say that this did not happen! You can't say that these things do not take place!' But she thought that the case was not substantiated well enough.
The new caseworker appeared to consider certain issues 'sufficient' to reject a case, but this did not suffice in the view of the interviewee in the comment above. Thus, experience with a particular case set may create different case-set effects, where the view of credibility, and perhaps even of risk, may look different in different units. Such perceived differences are in line with one of the main conclusions in a 2014 study of the UDI, which suggested that caseworkers' approaches to the assessment of credibility varied across the institution. 80 The researchers found that there were 'large differences in terms of what caseworkers perceive as contradictions and what they consider to be serious contradictions', and further that there were 'no established rules as to how consistent a story needs to be, how much external inconsistency caseworkers should accept, and how plausible a story needs to be before an applicant fulfils the criteria' . 81 Case-set effects may be one explanation for these variations.

CON CLUS ION
The complexity of refugee status determination calls for 'the confidence to make decisions, which, if wrong, may result in a person being forced to return to their country of origin where they may face further traumatization and even death' . 82 One interviewee put it this way: 'I mostly manage not to take it in too much, even if I am very aware of the seriousness of what I do. I simply think it would be unbearable if I did' . This article has aimed to explore how decision makers in Norway reached a sense of conviction about the outcome in a context of uncertainty.
As the analysis illustrates, they did so, in part, by establishing a system of distinction that allowed them to draw a line between refugees and applicants who were not considered eligible for protection. The system included interpretations of legal sources, country information, and political instructions, but it also seemed to be connected to experience-based knowledge about the outcomes of similar cases. Once the system had been set up, the exercise of discretion was not so much about individual choice as about making decisions in line with the established norms and patterns. As a result, decision makers did not often feel there was much choice or discretion, and doubts about the outcome were largely reduced.
Because asylum assessments involve high stakes and regular encounters with stories of suffering and violence, they entail substantial emotional labour and the risk of vicarious traumatization among decision makers and others involved in the RSD process. 83 Research has shown that professionals working in the RSD process may respond by detachment, disbelief, and denial of responsibility as a means of shielding themselves emotionally. 84 The mechanisms described above, whereby decision makers over time gain a sense of certainty about the correctness of their decisions, may similarly serve as a means to mitigate the emotional labour of asylum decision making and thereby the burdens of discretion. Even though asylum decisions can be considered to be 'inexorably subjective' , 85 decision makers may not necessarily perceive them as such. They may instead consider their decisions to be based on relatively specific norms and interpretations that they share with others in the work environment. As Hamlin points out, each country tends to organize its asylum procedure according to national conceptions of administrative justice. 86 This article illustrates that such conceptions of justice are produced -and reproduced -at the caseworker level as well, through an inter-subjective understanding of what a 'just' and 'correct' decision looks like.
Moreover, this article shows that caseworkers' conviction about the outcome seemed to be based not only on external sources, such as country information or the legal framework, or even on the features of the individual case in isolation. Their conviction stemmed -perhaps equally -from seeing the case in relation to other, similar cases. The importance of comparison with previous cases illustrates that the decision in an individual case cannot be understood in isolation. The assessment of many similar cases produces case-set effects, where distinctions between cases are drawn differently, depending on the case set as a whole. 87 The composition of cases and the sequence in which cases appear may affect individual outcomes.
The emphasis on comparison moves equal treatment to the centre of the decision-making process. The distinction between comparative and non-comparative justice 88 may be useful to understand its function. Comparative justice means that the yardstick of justice is produced by an internal comparison of cases; equal treatment is here the main yardstick of justice. Non-comparative justice means that each individual is assessed on his or her merits, independently of others, in accordance with a yardstick that is external. In this context, equal treatment plays no role in determining the outcome, but is merely a by-product of assessing similar cases with the same, external yardstick. The relative emphasis on a comparative or non-comparative approach probably depends on the degree of discretion involved and whether there are many similar decisions in the past that can provide guidance.
If cases are determined more on the basis of comparison than on external criteria, however, there is a risk that the yardstick of making the right decision will become largely internal to the case set. In this situation, equal treatment will contribute to perpetuating a 'practice' that has taken on a life of its own, since the 'forces that establish a precedent are not necessarily those that keep it in motion' . 89 Moreover, while equal treatment contributes to consistency and predictability -at least at a local level -it is worth keeping in mind that it is a normatively empty concept. 90 It is entirely possible that decisions are 100 per cent consistent, but substantially wrong. 91 Although a focus on consistency is important in order to avoid the outcome of a case depending mainly upon the decision maker, it is equally important to ensure that equal treatment does not contribute to perpetuating patterns of practice that are no longer valid. Indeed, the primary role of equal treatment may, in some cases, be to provide a sense of security for decision makers, as a proxy for making the right decision in a context of uncertainty.
There may be a particular inclination towards a comparative approach in the assessment of credibility, since it is open to a large degree of discretion. One caseworker said, for example, that she sometimes had difficulty articulating why a case should be rejected, but she suggested that this was based on experience with other cases where applicants were much 'more credible' . Compared to the articulate person who explains his or her case in detail, the next applicant may appear to be less credible -but does it make the claim less likely to be true? A comparative approach to credibility is likely to primarily produce a measure of applicants' ability to present their claims. The ability of an applicant to present a claim can, however, be affected by a large number of factors, such as trauma and socio-economic background. 92 In line with the standard of proof, the question should be whether the claim is 'reasonably likely', rather than whether it is more or less convincing than previous claims. Similarly, with regard to the assessment of risk, Bailliet suggests that there is a need for a 'non-comparative approach -cases should be evaluated with respect to whether the applicant's risk is of sufficiently serious harm and is linked to a Convention Ground, not whether it is greater than the risk of others' . 93 Heavy reliance on experience with other cases may produce a strong conviction of the correctness of the decision, but this conviction appeared to emerge prior to conscious reasoning, as a kind of intuition that at times was difficult to articulate or corroborate with 'public' reasons. Similar findings have been made in the field of legal decision making, where scholars have noted that the formal arguments are in part constructed after the fact, and do not necessarily reflect all the substantial reasons for the outcome. 94 Even though the reasoning in asylum determinations is open to analysis and appeal, the challenge is that the determinations are 'to an unknown and variable extent, constructed post facto in order to justify conclusions reached on more intuitive, less formally logical, grounds' . 95 This has implications for an applicant's ability to respond adequately during a complaint or appeal. In response to complaints, the UDI often wrote that it had taken into account the applicant's arguments or clarifications, but with a view to 'an assessment of the whole' it did not change the outcome. If that 'whole' is largely inaccessible to the applicant, it is difficult to understand and appeal a decision.
In conclusion, let us return to the question contained in the title of this article: 'Who is a refugee?' This article adds to the body of research that demonstrates that there are likely to be large variations in the answer to this question; however, the contingent nature of refugee status rarely becomes visible in public debates about refugee protection. In these debates, the representation of 'refugeehood' as an objective identity given by law appears to be tenacious. The clear-cut distinction between 'refugees' and 'migrants' is stressed by different actors in the field for different purposes. 96 UNHCR has been particularly vocal in pointing to the uniqueness of the refugee category, suggesting that when 'the line between "migrant" and "refugee" blurs, so does the distinction between migration control and refugee protection' . 97 It is, however, with reference to the distinction between refugees and migrants that restrictive migration policies can be implemented as well. 98 In the same way that the control of borders is justified in part with the need to save lives, 99 a restrictive approach to asylum can be justified by the need to uphold the value of protection. 100 Describing the uncertainties in the system may provoke concerns that not all refugees who are accepted are 'genuine', which could be used to call the legitimacy of the system into question. It may equally create worries that applicants who are rejected suffer subsequent persecution. It could create more understanding for the very real difficulties involved in making decisions in asylum cases, and it could be seen as a call for greater humility in the face of potential mistakes. It is perhaps inevitable that the current refugee system will only provide contingent, local forms of justice, and simultaneously produce other kinds of injustices. At a deeper level, the dilemmas in the system are a symptom of global inequalities, where security, freedom, wealth, and life chances are unequally distributed between countries. It may be unreasonable to expect justice from a system that is itself the product of fundamental inequality. The provision of protection to refugees bears similarities to emergency aid: it does not change the root causes of the problem. It is unevenly distributed, but it makes a difference to those who receive it -even a vital difference. The idea of the refugee as a non-negotiable identity across time and space may largely be fictional, but as Whyte points out, it is a 'crucial fiction' that has very real consequences for those who are granted -or denied -refugee status. 101 The act of determining someone's refugee status embodies, in condensed form, some of the most difficult and contentious political and ethical questions of our time.