Shifting the Burden: towards more equitable tests for Language Analysis in the Asylum Procedure

庇护程序中的语言分析(LAAP)被一些政府使用，特别是在欧洲。其目的是对不能或不愿通过文件手段证明身份的庇护申请者的原籍申请进行语言评估。LAAP在学术界和媒体中都极具争议性，但到目前为止，验证其实践的直接相关实验工作却非常少。LAAP取决于申请者的语言产出(即他们的言语)；事实上，LAAP的问题是，申请人的言语在多大程度上可以与他们声称的发源地相一致？目前，申请者必须在一次面试中证明其讲话的真实性，对所得数据的评估由为LAAP机构工作的母语非语言学家(NSNL)的分析员进行。然而，由于过去的流动性、社会人口因素和当天的表现，言语模式可能会有所不同，这使得LAAP评估受到批评。在这种情况下，我的研究核心本质上是经验性的。它的核心不是演讲，而是申请者的语言感知--特别是他们辨别说自己方言的人的能力。感知是申请者语言能力的一个重要方面，但到目前为止，LAAP忽略了这一点。我工作的最终目标是开发新的、稳健的、基于证据的知觉测试，这些测试可以证实--或者对--基于言语的初级LAAP评估产生怀疑。这些认知测试将通过引发申请者本身而不是NSNL分析师的判断来补充目前的LAAP实践，以确定他们自己(声称)语言变体的真实样本。PDF将允许我向学术受众传播这些新颖测试的实验结果，并将它们用于LAAP机构。我的实验主要涉及叙利亚阿拉伯语。在我的博士学位上的定量数据分析表明，在识别叙利亚阿拉伯语的话语时，叙利亚语的准确率明显高于非叙利亚的NSNL。这一发现与我最初的假设一致，让人对涉及相同任务的新感知测试产生了乐观情绪，因为在我的博士实验中，没有理由相信真正的叙利亚NSNL庇护申请者的表现会比叙利亚NSNL的听众差得多。在我的论文和在PDF期间进行的进一步实验工作中，这项知觉测试将建立在目标测试组--(叙利亚语)讲阿拉伯语的NSNL与(非叙利亚语)说阿拉伯语的NSNL--的经验性验证的基础上。PDF还将允许我在博士学位期间对定性数据进行分析，以开发第二种类型的知觉测试。我将研究听众在识别他们自己或其他语言变体时引用的提示，并创建一系列音频文本，其中的特征被声音调整或放置在没有出现在叙利亚阿拉伯语中的上下文中。例如，/e/可以通过使用计算机软件被人为地拉长，而/q/可以放在不属于它的单词中。在第二个测试中，庇护申请者将听取这些文本，并被要求识别异常情况。我的假设是，真正的叙利亚申请者会认为超长的/e/和/q/不是自然出现在叙利亚同胞的语言中，而假装是叙利亚人的申请者将缺乏识别异常现象所需的语言感知。机构接受这些测试将为现有的LAAP做法增加几层验证，从而增强其公平性、严格性和透明度。我将在PDF中投入时间与LAAP机构和媒体接触，以提高人们对LAAP公平问题的认识。我还将探索合作，通过这些合作，这些测试可以适用于说其他方言和语言的人。

Language Analysis in the Asylum Procedure (LAAP) is used by a number of governments, particularly in Europe. Its purpose is to produce a linguistic assessment of claims to origin by asylum applicants who cannot or will not prove their identity by documentary means. LAAP is highly controversial in both academia and the media, yet remarkably little directly relevant experimental work (in)validating its practice has so far been conducted. LAAP depends on applicants' language production (i.e. their speech); in fact, the question in LAAP is, 'To what extent is the applicant's speech identifiable with the area from which they claim to have originated?'. At present, applicants must demonstrate the authenticity of their speech in a one-shot interview, and the assessment of the derived data is carried out by analysts who are native speaker non-linguists (NSNLs) working for LAAP agencies. Yet speech patterns may vary due to past mobility, socio-demographic factors and performance on the day, leaving LAAP assessments open to criticism.This being the case, the core of my research is empirical in nature. It centres not on the speech but on the language perception of applicants-in particular, their ability to identify fellow speakers of their own dialect. Perception is an important aspect of applicants' linguistic competence that has hitherto been overlooked in LAAP. The ultimate objective of my work is the development of novel, robust and evidence-based perception tests that can confirm-or cast doubt on-the primary, speech-based LAAP assessment. These perception tests would complement current LAAP practice by eliciting judgements from applicants themselves, not an NSNL analyst, in identifying authentic samples of their own (claimed) language variety. The PDF will allow me to disseminate to academic audiences the results of experiments piloting these novel tests, and to put the case for their use to LAAP agencies.My experiments mainly concern Syrian Arabic. Quantitative data analysis in my PhD demonstrates significantly superior accuracy for Syrian over non-Syrian NSNLs when identifying utterances in Syrian Arabic. This finding, which is consistent with my initial hypothesis, engenders optimism about a novel perception test involving the same task, since there is no reason to believe that genuine Syrian NSNL asylum applicants would perform significantly worse than the Syrian NSNL listeners in my PhD experiments. This perception test would be founded on the empirically validated performance of the target test groups-(Syrian) Arabic-speaking NSNLs vs. (non-Syrian) Arabic-speaking NSNLs-in my thesis and in further experimental work, conducted during the PDF. The PDF will also allow me to build on analysis of qualitative data in my PhD to develop a second type of perception test. I will examine the cues that listeners cited in identifying their own or other language varieties, and create a series of audio texts in which features are acoustically adjusted or placed in contexts where they do not occur in Syrian Arabic. For example, /e/ could be artificially 'elongated' by the use of computer software, and /q/ placed in words where it does not belong. In the second test, asylum applicants will listen to these texts and be asked to identify the anomalies. My hypothesis is that an authentic Syrian applicant will perceive the extra-long /e/ and the /q/ as not naturally occurring in the speech of a fellow Syrian, while an applicant pretending to be Syrian will lack the linguistic perceptions necessary to recognise the anomalies. Uptake by agencies of these tests would add several layers of validation to extant LAAP practice, thereby enhancing its fairness, rigour and transparency. I will dedicate time within the PDF to engage with LAAP agencies and the media to raise awareness about issues of fairness in LAAP. I will also explore collaborations through which the tests could be adapted for speakers of other dialects and languages.