examining the potential challenges | 10.1163/22131035-14020001 15 actual records existed. In contrast, searches associated with white persons did not prompt such advertisements.74 In State of Wisconsin v Loomis, the Wisconsin Supreme Court assessed whether the use of an algorithmic risk assessment tool to determine if the defendant could be supervised within the community rather than detained violated the defendant’s right to due process. As the court noted, risk scores are intended to predict the general likelihood that those with a similar history of offending are either less likely or more likely to commit another crime following release from custody. However, the court found that the risk assessment used in this case did not predict the specific likelihood that an individual offender will reoffend.75 Instead, it provided a prediction based on a comparison of information about the individual to a similar data group (African-America) as it was more likely to falsely predict that they would reoffend compared to white defendants.76 The United States Citizenship and Immigration Services (uscis) uses Asylum Text Analytics, an ai tool to identify fraud in asylum applications. This tool employs machine learning to identify plagiarism-based fraud in applications for asylum. The technology scans the narrative text of applications and looks for duplicate language repeated across applications. It flags when applicants’ stories do not align. uscis states that the machine reviews an applicant’s narrative not only individually but also compared with other applicants’ narratives. The tool can be susceptible to discrimination against those applicants who do not speak English well.77 Regarding bias leading to discrimination, Chen asserts that the idea that ai processes are inherently ‘objective’ and ‘neutral’ is a misconception.78 Despite ai’s sophistication, the systems can still be classified as ‘garbage-in-garbageout’ since they are unable to distinguish between a biased and unbiased dataset.79 Consequently, the system might automate systemic and implicit bias that might pass for neutrality while posing as algorithm objectivity.80 Therefore, algorithms risk aggravating human biases in decision-making. 74 75 76 77 78 79 80 Zhisheng Chen, ‘Ethics and Discrimination in Artificial Intelligence-enabled Recruitment Practices’ (2023) 10 Humanities and Social Sciences Communications 6. State v Loomis, 881 n.w.2d 749 (wis. 2016) (2017)130(2) Havard Law Review <https://harvard lawreview.org/print/vol-130/state-v-loomis/> accessed 30 September 2025. Ibid. Julie Mao and others, ‘Automating Deportation: The Artificial Intelligence Behind the Department of Homeland Security’s Immigration Intelligence Behind Immigration Enforcement Regime’ June 2024 <https://mijente.net/wp-content/uploads/2024/06 /Automating-Deportation.pdf> accessed 30 September 2025, 23. Chen, (n 74) at 2. Awotula, (n 2) at 17. Ibid. International Human Rights Law Review (2025) 1–31