Research Guides: AI Academy Resource Guide : Select Readings

Readings of Interest

Academic Integrity

AI and Academic Integrity: Exploring Student Perceptions and Implications for Higher Education
This study examines how university students perceive academic misconduct in the context of AI tools like ChatGPT. Most view using AI to write entire papers as serious misconduct, while minor AI use is seen as less severe. Ethical education, not demographics, strongly shapes these perceptions, highlighting the need for clear AI policies.
Chatting and cheating: Ensuring academic integrity in the era of ChatGPT
This paper explores the benefits and risks of using ChatGPT in higher education, highlighting its potential to enhance engagement and accessibility while raising concerns about academic honesty. It recommends proactive strategies such as clear policies, training, and detection methods to promote ethical and responsible AI use in academic settings.
Peculiarities of use of generative artificial intelligence in codes of academic integrity in higher education institutions of Singapore
This article analyzes academic integrity policies related to AI use at all six universities in Singapore. It finds a trend toward updating policies to balance regulation with support for students and faculty. Institutions emphasize ethical, effective AI use while maintaining high academic standards, offering insights for global policy development.
Rethinking Plagiarism in the Era of Generative AI
This article examines how generative AI, like ChatGPT, is reshaping academic writing, plagiarism, and intellectual property in English composition courses. It explores educator perceptions, rising AI-generated content, and the need to rethink originality, pedagogy, and equity. The paper proposes curriculum revisions to address AI's growing role in education.
The global landscape of academic guidelines for generative AI and LLMs
This study reviews 80 academic guidelines on generative AI and large language models in education, highlighting both benefits such as improved access and collaboration, and challenges like misinformation and academic integrity risks. It recommends balanced strategies for responsible integration of these technologies in academic settings.

AI Ethics in Education

AI in education: landscape, vision and critical ethical challenges in the 21st century
This chapter reviews the rapid growth of AI in education, highlighting key applications and barriers to responsible adoption, including ethical concerns, policy gaps, and educator readiness. It critically examines AIED’s assumptions and goals, calling for coordinated efforts to ensure its effective and ethical integration into education systems.
AI and education: Guidance for policy-makers.
This publication guides policy-makers on leveraging AI in education to advance SDG 4. It outlines AI fundamentals, explores trends and risks, and emphasizes ethical, inclusive use. It offers actionable recommendations to help education systems adapt, prepare learners for an AI-driven world, and develop context-specific policies and programs.
Ethical Implications and Principles of Using Artificial Intelligence Models in the Classroom: A Systematic Literature Review
This systematic review examines the ethical implications of using AI in classrooms, identifying key concerns like bias, privacy, and transparency. It highlights six core ethical principles, with fairness and privacy most studied. The study calls for deeper research, clearer implementation strategies, and better assessment of AI's ethical impact in education.
Undergraduate students’ perspectives of generative AI ethics
This mixed-method study explores undergraduates’ views on AI ethics in education. Among five principles, students valued autonomy most and justice least. Gender and grade level influenced agreement levels, and 14 rationale themes emerged. Findings suggest instructors should address student diversity, public interest, and attitudinal change when teaching AI ethics.
Social Responsibility Attitudes Among Undergraduate Computer Science Students: An Empirical Analysis
Using a survey of 982 students (184 CS majors) graduating between 2017–2021, the study finds undergraduate computer science students report significantly lower social responsibility attitudes than peers in other STEM fields. Results highlight the effectiveness of targeted ethics or responsibility training integrated into computing curricula to boost those attitudes.
Teaching Ethics in Computing: A Systematic Literature Review of ACM Computer Science Education Publications
This review analyzes 100 papers on teaching ethics in computing education over 40 years. It finds diverse definitions, methods, and challenges, with limited consensus on best practices. To advance, the field must clarify ethical concepts, align teaching strategies accordingly, and develop better ways to assess student learning.

Learning, Skill Development, & Challenges

AI Literacy in Teacher Education: Empowering Educators Through Critical Co-Discovery
This study explores a critical co-discovery approach to enhance AI literacy in teacher education. Through collaborative activities, educators developed understanding of AI concepts, ethics, and applications. The findings emphasize the importance of integrating sustained, reflective AI literacy into teacher training to empower educators in navigating AI's role in education.
ChatGPT-generated help produces learning gains equivalent to human tutor-authored help on mathematics skills
This study compares ChatGPT- and human tutor-generated help in math learning. ChatGPT led to significant learning gains over no help, with results comparable to human tutors. While 32% of ChatGPT responses initially failed quality checks, applying a hallucination-reduction technique greatly improved accuracy, especially in algebra topics.
Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
This review of 69 experimental studies finds that ChatGPT improves academic performance, motivation, and higher-order thinking while reducing mental effort, though it does not affect self-efficacy. Most interventions occurred in university settings. The authors call for stronger methods, including complex assessments, long-term studies, and objective measures to validate findings.
Is Chatgpt a menace for creative writing ability? An experiment.
This experimental study of 600 university students found that using ChatGPT negatively impacts creative writing skills, particularly originality, content presentation, accuracy, and elaboration. The findings provide quantitative evidence supporting earlier qualitative concerns. The study urges cautious AI integration in writing education and suggests policy and pedagogical changes to protect student creativity.
An empirical investigation of the impact of ChatGPT on creativity
This study finds that ChatGPT (GPT-3.5) enhances creativity in problem-solving across five experiments, outperforming both no tools and Google searches. It was especially effective for generating incrementally novel ideas by combining remote concepts coherently, regardless of problem constraints or empathy requirements.

Course Development, Assessments, and Grading

Collaborative Intelligence: Towards Practical, Critical and Cooperative Teaching & Learning with AI
The framework offers strategies for teaching amid AI disruption, emphasizing algorithmic literacy, course outcome review, and collaborative learning. It focuses on adaptable practices developed with students rather than specific technologies.
Unveiling Scoring Processes: Dissecting the Differences Between LLMs and Human Graders in Automatic Scoring
This study reveals that large language models often diverge from human grading rubrics when scoring student responses, relying on shortcuts instead of deep reasoning. By prompting LLMs with high-quality, human-aligned rubrics, scoring accuracy improves. The findings stress the need for careful alignment to ensure reliable AI assessment in education.
The Textbook of Tomorrow: Rethinking Course Material Interfacing in the Era of GPT
Despite decades of LMS use, digital course readings remain static replicas of print texts. This paper proposes integrating smart textbooks, AI-enhanced tools that let students interact with readings through summaries, explanations, and quiz generation. The goal is to reduce learning inequities and evolve toward more personalized, context-aware digital learning experiences.

Other

ChatGPT and Artificial Intelligence in Higher Education.
This Quick Start Guide introduces ChatGPT and its use in higher education, outlining its functionality, benefits, challenges, and ethical considerations. Based on GPT-3.5, it offers practical steps for institutions while emphasizing the need to stay updated on rapid technological and ethical developments in AI.
Out of the laboratory and into the classroom: The future of artificial intelligence in education.
This paper assesses the current state of AI in education, focusing on intelligent tutoring systems and AI teaching agents. It contrasts potential benefits and risks, explores future scenarios, and recommends peer review and stakeholder engagement to guide socially responsible AIEd development and policy.
Why AI will not Democratize Education: a Critical Pragmatist Perspective: Why AI will not Democratize Education: a Critical Pragmatist Perspective
This paper, drawing on Dewey’s philosophy, argues that current commercial AI tools hinder democratic education by promoting individualization, automation, and limited student agency. It critiques narrow views of AI’s role in education and offers alternatives, like simulations, diverse learning networks, and teacher-augmenting tools, to better support democratic values.
The Dawn of the AI Era: Teens, Parents, and the Adoption of Generative AI at Home and School
This report highlights teens' rapid adoption of generative AI, mainly for schoolwork, while most parents remain unaware. It urges schools to consider these usage patterns in policy development. Open communication between students, parents, and educators is key to navigating AI’s educational benefits and risks effectively.

AI and the Environment

AI Energy Usage Climate Footprint
Big Tech’s AI boom is driving a sharp rise in energy, water consumption, and carbon emissions. Data centers now rival aviation in global emissions, power grids strain, and freshwater supplies are stressed. While efforts toward efficiency, nuclear, and renewables are underway, transparency and regulatory action remain urgent.
The Climate and Sustainability Implications of Generative AI
This MIT report warns that generative AI’s rapid growth, driven by efficiency gains and demand, is causing significant environmental harm through expanded data-center energy, water use, and resource strain. It urges adoption of benefit–cost frameworks, stakeholder coordination, and sustainability-centered policymaking to guide responsible Gen-AI development.
Estimating the Carbon Footprint of BLOOM, a 176B Parameter Language Model
In the present article, we aim to quantify the carbon footprint of BLOOM, a 176-billion parameter language model, across its life cycle. We estimate that BLOOM's final training emitted approximately 24.7 tonnes of CO 2 eq if we consider only the dynamic power consumption, and 50.5 tonnes if we account for all processes ranging from equipment manufacturing to energy-based operational consumption.
The Environmental Impacts of AI -- Primer
Hugging Face’s primer outlines how AI's entire lifecycle, from hardware manufacturing and energy-hungry training to cooling, water use, and mineral mining, drains resources and emits greenhouse gases. It highlights gaps in transparency and regulation, and calls for efficiency techniques, sustainable practices, and policy oversight to mitigate AI’s environmental toll
Environmental footprint of GenAI – Changing technological future or planet climate?
A 2024 Czech survey found that frequent GenAI users view climate change as less urgent and show less willingness to adopt eco-friendly tech. The study highlights a need to integrate environmental awareness into AI literacy to promote sustainable GenAI use amid rising energy demands and shifting digital behaviors.

Global and Regional AI Governance Approaches

Artificial Intelligence and the ‘Good Society’: the US, EU, and UK approach
This article compares 2016 AI policy reports from the U.S., UK, and EU, highlighting shared ethical and social concerns but noting a lack of unified long-term vision. It proposes a two-pronged strategy to guide the development of a "good AI society" through clearer political direction and coordinated efforts.
Transnational Digital Governance and Its Impact on Artificial Intelligence
Rapid technological change, especially AI, challenges existing governance. This chapter examines current AI regulation, focusing on the EU’s approach as a leading model. It highlights how EU rules shape global standards through the “Brussels effect” and emphasizes the need for innovative, value-driven, and globally influential regulatory frameworks.
Introducing the AI Governance and Regulatory Archive (AGORA): An Analytic Infrastructure for Navigating the Emerging AI Governance Landscape
AGORA is a public, curated dataset of 330+ AI laws and policies from diverse regions, organized by a unique taxonomy of risks, strategies, and domains. It addresses gaps in AI governance monitoring by enabling reliable analysis through structured, validated data and a publicly accessible platform.
Non-western AI ethics guidelines: Implications for intercultural ethics of technology.
This study analyzes non-Western AI ethics guidelines to uncover cultural elements beneath seemingly universal language. While resembling Western phrasing, these guidelines reflect distinct cultural values. The authors argue that ethical “universals” arise from a shared desire for coexistence, not from a singular, culture-independent moral standard.

Ethics Principles and High-Level Frameworks

Principled Artificial Intelligence: Mapping Consensus in Ethical and Rights-Based Approaches to Principles for AI
This white paper analyzes 36 major AI principles documents, revealing consensus around eight core themes like fairness, transparency, and accountability. Despite variations, shared values are emerging. The study aims to support policymakers and stakeholders in advancing a more unified global approach to ethical AI development and governance.
Principles alone cannot guarantee ethical AI.
AI ethics often mirrors medical ethics, embracing high-level principles like fairness and accountability. However, key differences, such as lacking shared goals, professional norms, and accountability, limit their effectiveness in AI. Mittelstadt warns that apparent consensus may obscure deeper political and practical challenges in applying principles to AI governance.
Democratising AI: Multiple Meanings, Goals, and Methods
AI democratisation is a complex, sometimes conflicting concept encompassing use, development, profits, and governance. This paper argues that meaningful progress requires recognizing governance as central to managing trade-offs, rather than equating democratisation solely with broader AI accessibility.
University Governance for Responsible AI
This study analyzes AI-related policy documents from U.S. universities to examine emerging ethics concerns and institutional responses in higher education. It identifies key issues around academic integrity, appropriate AI use, and policy development, offering recommendations based on observed strategies and evolving norms in educational policymaking.

Ethics Implementation and Practice

Companies committed to responsible AI: From principles towards implementation and regulation?
This study examines whether 24 AI companies backing responsible AI principles have acted on them. While some have made meaningful progress, others risk “ethics washing.” Many support government regulation, especially for high-risk AI, as a balanced alternative to either self-regulation or outright bans.
Challenges of responsible AI in practice: scoping review and recommended actions
Responsible AI guidelines promote democratic values but often fail to influence real-world practice. This study identifies five key barriers and explores interdisciplinary, socio-technical solutions, grounded in concrete examples, to help translate ethical principles into actionable outcomes within AI development processes.
Responsible AI Pattern Catalogue: A Collection of Best Practices for AI Governance and Engineering
Responsible AI is essential for adoption but often lacks practical guidance beyond high-level principles. This article introduces an RAI Pattern Catalogue, offering actionable system-level practices across governance, process, and design to help stakeholders implement responsible AI throughout the entire development lifecycle.
The emergence of artificial intelligence ethics auditing
AI ethics auditing is expanding alongside new regulations, yet remains underdeveloped. Interviews with 34 auditors reveal challenges like limited stakeholder input, unclear standards, and resource constraints. Despite these, auditors are shaping the field by developing frameworks, interpreting regulations, and promoting responsible AI practices.

Sector-Specific Governance and Ethics

AI and Procurement.
This study examines how AI use in procurement affects supplier pricing. Automation alone leads to higher price quotes for chatbot buyers, but combining automation with smart AI reduces discrimination and adds value. Effective procurement requires integrating both automation and intelligence rather than using automation in isolation.
AI Ethics and Governance in the Job Market: Trends, Skills, and Sectoral Demand
AI ethics and governance skills are increasingly in demand, yet most AI literacy efforts remain technical. Analyzing over 4 million job postings (2018–2023), this study finds growing demand for professionals with combined technical, managerial, and regulatory skills, especially in finance and tech, highlighting a need to expand ethical AI education.
AI Ethics in the Public, Private, and NGO Sectors: A Review of a Global Document Collection
This study analyzed 112 global AI ethics documents from 2016–2019, revealing key differences across public, private, and NGO sectors. Public and NGO documents covered more ethical topics, emphasized law and participation, and reflected greater focus on social responsibility than those from private entities, which prioritized economic interests.

Cognitive and Behavioral Studies

AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking
This study finds that frequent AI tool use is linked to lower critical thinking skills, largely due to increased cognitive offloading. Younger users are more reliant on AI and score lower in critical thinking, while higher education correlates with stronger skills. The study urges strategies to promote critical engagement with AI.
The Memory Paradox: Why Our Brains Need Knowledge in an Age of AI
This chapter links the decline in IQ scores (reversal of the Flynn Effect) to increased cognitive offloading via AI and digital tools. It argues that underusing memory systems weakens reasoning and learning. The authors advocate for educational and policy strategies that use AI to support, not replace, human cognitive development.
Use of large language models might affect our cognitive skills.
Large language models can produce advanced text or code with minimal input, potentially weakening human writing and thinking skills. It is essential to examine their cognitive impact and consider whether such reliance aligns with our educational and developmental goals.

A Brief History of AI: How to Prevent Another Winter (A Critical Review)
Artificial intelligence has witnessed exponential growth in the past decade. Advances in computing power and the design of sophisticated artificial intelligence algorithms have enabled computers to outperform humans in a variety of tasks. Yet, artificial intelligence’s path has never been smooth, having essentially fallen apart twice in its lifetime after periods of popular success. We provide a brief rundown of artificial intelligence's evolution, highlighting its crucial moments and major turning points from inception to the present. In doing so, we attempt to learn, anticipate the future, and discuss what steps may be taken to prevent another winter
Teach Better or Show Smarter? On Instructions and Exemplars in Automatic Prompt Optimization
This study finds that exemplar optimization (EO) methods, even simple ones, can outperform instruction optimization (IO) in prompt tuning for large language models. Combining EO and IO yields the best results, highlighting the need for more research into EO, which has been underexplored compared to IO.