algorithmic-bias
Filtering by topic algorithmic-bias(3)Clear all filters
- NewsThe PIE News24 Jul 2026
AI in Admissions: please don’t do it just because you can
The article warns against using AI to fully automate admissions decisions, noting that systems trained on past data risk reproducing historical biases and overlooking nontraditional applicants. It advocates for using AI as a support tool rather than a replacement for human judgment, highlighting dangers such as institutional repetition, bias, and over-reliance on automated scores.
Original abstract
New tools promise faster processing, better student-course matching and improved risk assessment. The Uni Guide, for example, uses AI to help students identify universities and courses that fit their aspirations, while UniReady Global has developed AI capabilities to help providers assess whether applicants are likely to be genuine students. Others are going further, seeking to automate significant parts, or even all, of the admissions process. The attraction is obvious. AI processes information at a scale no human team can match. It can triage applications, identify missing documents, flag inconsistencies, detect potential fraud and highlight where an applicant’s background, ambitions and chosen programme may not align. For institutions under growing pressure to recruit effectively while managing compliance and costs, AI offers genuine advantages. But there is a crucial distinction between using AI to support admissions decisions and allowing Agentic AI to make them. The danger begins when a risk flag automatically becomes a rejection, a course-fit score becomes a gatekeeper, or an automated recommendation is treated as more objective than the human judgement it was designed to inform. Three risks deserve particular attention. The first is institutional repetition. Professor Marnie Hughes-Warrington has observed that the large language models underpinning AI are inherently historical. They learn from what has happened before. That is precisely what makes them effective; but it also means they risk reproducing yesterday’s patterns rather than identifying tomorrow’s potential. Universities rightly talk about preparing students for a changing world. Yet admissions systems trained on previous cohorts may repeatedly favour applicants who resemble those already admitted. Efficiency could come at the expense of opportunity, overlooking talented students whose potential sits outside historical norms. The second challenge is bias. Admissions systems trained on previous coho
- PaperJournal of Learning Analytics8 Jul 2026
Toward Reliable Estimation of Algorithmic Bias for Minority Groups
Jaeyoon Choi, Shamya Karumbaiah, Jeffrey Matayoshi
Predictive models in learning analytics often show performance disparities across demographic groups. This study identifies that small group sizes, common among marginalized students, cause unreliable bias estimates due to sampling error. The authors recommend using confidence intervals, multiple metrics, and larger samples to improve bias assessment.
Original abstract
While predictive models are widely used in learning analytics, several studies have shown that the performance of these models can vary significantly across different demographic groups of students. The first step to audit for and mitigate these group biases is to accurately estimate them. However, the current practices for identifying and measuring group bias often suffer from reliability issues. In this paper, we use simulations and real-world data analysis to explore statistical factors that impact the reliable estimate of group bias and suggest approaches to improve their statistical robustness. Our analysis revealed that small group sizes lead to high variability in group bias estimation due to sampling error -- an issue that is more likely to impact students from historically marginalized communities. We then suggest statistical approaches, such as bootstrapping, to construct confidence intervals for a more reliable estimation of group bias. Based on our findings, we encourage future learning analytics researchers to ensure sufficiently large group sizes, construct confidence intervals rather than relying on p-values, use at least two metrics, and move beyond the dichotomy of the presence or absence of bias for a more comprehensive evaluation of group bias.
- PaperJournal of Second Language Writing29 Jun 2026
“When the Editor Detected AI — But the ‘AI’ Was Me”: Algorithmic misjudgement and the crisis of authorship in TESOL publishing
Yusop Boonsuk
A study reveals how AI-detection algorithms falsely flagged a human author as AI-generated, highlighting the crisis of authorship in TESOL publishing due to algorithmic misjudgment.