ai-ethics
Filtering by topic ai-ethics(4)Clear all filters
- PaperComputers and Education: Artificial Intelligence14 Jul 2026
Balancing AI responsibility with privacy, safety, and utility: Unlearning in large language models for mathematics education
Chenglu Li, Gökhan Gülfidan, Yinqi Zhang-Kopf
This study applies gradient-based LLM unlearning to reduce personally identifiable information (PII) and harmful content in math tutoring models while maintaining performance on math tasks. Results show substantial decreases in PII and harm rates without sacrificing utility, demonstrating a path toward responsible AI in education.
Original abstract
Online mathematics learning platforms are increasingly adopting large language models (LLMs) to provide scalable, on-demand support, but these models may reproduce private information from training data or generate harmful language. This raises concerns about responsibility in educational settings regarding the use of pre-trained models. LLM unlearning is an emerging area for reducing a model’s ability to produce specific unwanted content and remains underexplored in educational research. This study aims to investigate how LLM unlearning reduces the model's reliance on personally identifiable information (PII) and inappropriate content in the math tutoring context, while maintaining the model's utility on both single-label and multi-label downstream math tasks. We applied a gradient-based LLM unlearning approach to three different models, which were pre-trained on approximately 3 million data points from an Algebra I online discussion forum between students and professional tutors. PII and harmful content were detected on this training data and used for unlearning in two different orders (PII and harmful content unlearning). Then, the generated outputs from these two unlearning models were compared with those of the pre-trained model in terms of PII-containing output rate and harmful rate. Moreover, unlearned models were evaluated on two different math classification tasks. The results showed that the rates of PII-containing output rate and harmfulness substantially decreased compared to the pre-trained models, and the utility of the unlearned model was still maintained. These findings demonstrate how LLM unlearning can be applied to pre-trained models to behave them more responsibly, while maintaining strong model performance on math-related tasks.
- PaperJournal of Second Language Writing4 Jul 2026
AI and scholarly writing: Ethics, authorship and ownership
Brian Paltridge
This paper examines the ethical issues surrounding AI use in scholarly writing, particularly regarding authorship and ownership.
- PaperComputers and Education: Artificial Intelligence19 Jun 2026
Enhancing AI literacy course satisfaction through empowerment in AI problem-solving and ethical awareness: Development and validation of an AI project-based learning scale
Jinyu Zhu, Siu Cheung Kong
A study developed and validated a scale measuring students' perceptions of project-based learning (PBL) in AI literacy courses, involving 446 secondary and university students in Hong Kong. Results showed that empowerment in using AI for problem-solving and AI ethical awareness mediate the relationship between PBL and course satisfaction.
Original abstract
The use of the project-based learning (PBL) approach in developing AI literacy is widely adopted. However, student satisfaction with AI literacy courses has not been thoroughly explored. This study addresses this gap by examining the associations between students’ perceived PBL, empowerment in using AI for problem-solving, AI ethical awareness, and satisfaction with AI literacy courses. A total of 1,027 secondary and university students from 102 secondary schools and a university in Hong Kong participated in an AI literacy course, among whom 446 provided complete data for the structural equation modelling (SEM) analysis. We developed a context-grounded scale to measure students’ perceptions of the PBL experiences when using AI for problem-solving (AI-PBLS) within AI literacy course. This scale was validated through exploratory and confirmatory factor analyses. A two-step SEM results confirmed that empowerment in using AI for problem-solving and AI ethical awareness mediated the relationship between PBL and their satisfaction with the AI literacy course. This study makes a valuable contribution by introducing a robust scale for researchers to assess students’ perceived PBL experiences the context of AI applications. The findings further highlight the potential of PBL to enhance AI literacy course satisfaction by fostering conditions that empower students in using AI for problem-solving and increase their awareness of AI ethics.
- PaperAssessment & Evaluation in Higher Education29 May 2026
Between delegation and responsibility: an exploratory case study of graduate educators’ conceptualizations of AI-supported assessment using the AI assessment scale
Armağan Ateşkan
Graduate educators conceptualize AI-supported assessment not as technical classification but as ethical boundary-setting about delegating evaluative responsibility. Constructive alignment was strongest when AI was embedded in design, yet designing without AI tools increased awareness of AI dependence and confidence in unassisted design, suggesting AI assessment literacy may require experiential constraint.
Original abstract
This study investigates how graduate educators conceptualize and apply the AI Assessment Scale (AIAS) within their assessment design practice. Drawing on an exploratory qualitative case study design, the study analyzed eight AI-integrated lesson plans produced by in-service teachers in a technology elective course, supplemented by semi-structured interviews with four participants. Findings suggest that AIAS level selection is experienced not as a technical classification but as an ethical boundary-setting practice, a judgment about delegating evaluative responsibility between human and AI agents. Participants demonstrated variably developed AI assessment literacy: procedural ethics (integrity, authorship) and experiential ethics (learner agency) were more fully articulated than structural ethics (algorithmic bias, data governance). Constructive alignment was strongest when AI was constitutively embedded in the design; conversely, AI integration could shift assessment criteria from content-driven toward procedurally driven evaluation, a drift originating at the design-imagining stage before any tool was deployed. Notably, designing without AI tools heightened participants’ awareness of habitual AI dependence and, in several cases, increased confidence in unassisted design, suggesting that AI assessment literacy may require experiential constraint as well as conceptual instruction. Implications are discussed for teacher education, AIAS professional development, and AI assessment literacy.