Appendix H — Further Reading

This is a selective reading list for the revised book, not a systematic literature review. Research findings depend on the models, participants and tasks studied. None of these sources validates the fictional PDP-2026.1 pack or its illustrative marking threshold.

H.1 Models, context and prompting

OpenAI (2023), GPT-4 Technical Report. Describes the evaluated model and limitations; it does not disclose the parameter count. Distinguish the model from a current chat product.
Liu, N. F. and colleagues (2024), Lost in the Middle: How Language Models Use Long Contexts. Investigates sensitivity to the position of relevant information. A large context window is not a guarantee of reliable use of every source.
Xu, Xiaohan and colleagues (2024), Re-Reading Improves Reasoning in Large Language Models, EMNLP, 15549–15575. Reports a rereading method on tested benchmarks, not a universal rule that two repetitions are optimal for every task.
Vaswani, A. and colleagues (2017), Attention Is All You Need. The transformer architecture paper; architectural history is distinct from evidence of educational benefit.

H.2 Learning and judgement

Bastani, H. and colleagues (2025), Generative AI without guardrails can harm learning. The high-school mathematics study distinguishes performance while supported from later unaided performance. Its setting should not be silently generalised to all business courses.
Wineburg, S. and McGrew, S. (2019), Lateral Reading and the Nature of Expertise: Reading Less and Learning More When Evaluating Digital Information. Relevant to checking a claim beyond the page or generated answer presenting it.
Dawson, P. (2017), Assessment rubrics: towards clearer and more replicable design, research and practice, Assessment & Evaluation in Higher Education, 42(3), 347–360. Useful for specifying rubric design decisions; it is not an endorsement of this book’s sample rubric.
TEQSA (2025), Enacting assessment reform in a time of artificial intelligence. Sets out multiple approaches to assessment reform. It does not require a particular transcript metric, grade weighting or universal AI ban.

H.3 Data handling and product boundaries

Office of the Australian Information Commissioner, Guidance on privacy and the use of commercially available AI products. Start here for the Australian privacy considerations, then obtain institution-specific advice and approval.
Microsoft, Privacy and protections. Read the scope and feature-specific conditions rather than inferring permission from an enterprise label. Product terms can change.
Carlini, N. and colleagues (2021), Extracting Training Data from Large Language Models. Demonstrates training-data extraction under studied conditions.
Nasr, M. and colleagues (2023), Scalable Extraction of Training Data from (Production) Language Models. Further extraction research. Neither paper shows that an arbitrary private upload is necessarily retrievable by another user; neither supports a blanket impossibility claim.

H.4 Context-specific requirements

Fair Work Ombudsman, Bullying and harassment. Australian workplace guidance for checking the scope of a real issue. Fictional pack policies are not statements of law.
European Union, General Data Protection Regulation. Consult the applicable provisions and competent advice; consent is not the only possible legal basis.
US Department of Education, Eligible Student Guide to FERPA. Relevant rights and exceptions require context-specific interpretation.

These links were reviewed for this September 2026 revision. Reachability is separate from applicability. Before deployment, recheck current rules, product terms and the institution’s own requirements.

H.5 Companion methodology

Michael Borck, Conversation, Not Delegation, develops the wider conversation and verification approach. The present book supplies its own short recap and teaching materials; reading the companion is optional.

# Further Reading {#sec-further-reading .appendix} This is a selective reading list for the revised book, not a systematic literature review. Research findings depend on the models, participants and tasks studied. None of these sources validates the fictional PDP-2026.1 pack or its illustrative marking threshold. ## Models, context and prompting - OpenAI (2023), [GPT-4 Technical Report](https://arxiv.org/abs/2303.08774). Describes the evaluated model and limitations; it does not disclose the parameter count. Distinguish the model from a current chat product. - Liu, N. F. and colleagues (2024), [Lost in the Middle: How Language Models Use Long Contexts](https://arxiv.org/abs/2307.03172). Investigates sensitivity to the position of relevant information. A large context window is not a guarantee of reliable use of every source. - Xu, Xiaohan and colleagues (2024), [Re-Reading Improves Reasoning in Large Language Models](https://aclanthology.org/2024.emnlp-main.871/), *EMNLP*, 15549–15575. Reports a rereading method on tested benchmarks, not a universal rule that two repetitions are optimal for every task. - Vaswani, A. and colleagues (2017), [Attention Is All You Need](https://arxiv.org/abs/1706.03762). The transformer architecture paper; architectural history is distinct from evidence of educational benefit. ## Learning and judgement - Bastani, H. and colleagues (2025), [Generative AI without guardrails can harm learning](https://hamsabastani.github.io/education_llm.pdf). The high-school mathematics study distinguishes performance while supported from later unaided performance. Its setting should not be silently generalised to all business courses. - Wineburg, S. and McGrew, S. (2019), [Lateral Reading and the Nature of Expertise: Reading Less and Learning More When Evaluating Digital Information](https://journals.sagepub.com/doi/10.1177/016146811912101102). Relevant to checking a claim beyond the page or generated answer presenting it. - Dawson, P. (2017), [Assessment rubrics: towards clearer and more replicable design, research and practice](https://doi.org/10.1080/02602938.2015.1111294), *Assessment & Evaluation in Higher Education*, 42(3), 347–360. Useful for specifying rubric design decisions; it is not an endorsement of this book's sample rubric. - TEQSA (2025), [Enacting assessment reform in a time of artificial intelligence](https://www.teqsa.gov.au/sites/default/files/2025-09/enacting-assessment-reform-in-a-time-of-artificial-intelligence.pdf). Sets out multiple approaches to assessment reform. It does not require a particular transcript metric, grade weighting or universal AI ban. ## Data handling and product boundaries - Office of the Australian Information Commissioner, [Guidance on privacy and the use of commercially available AI products](https://www.oaic.gov.au/privacy/privacy-guidance-for-organisations-and-government-agencies/guidance-on-privacy-and-the-use-of-commercially-available-ai-products). Start here for the Australian privacy considerations, then obtain institution-specific advice and approval. - Microsoft, [Privacy and protections](https://learn.microsoft.com/en-us/copilot/privacy-and-protections). Read the scope and feature-specific conditions rather than inferring permission from an enterprise label. Product terms can change. - Carlini, N. and colleagues (2021), [Extracting Training Data from Large Language Models](https://arxiv.org/abs/2012.07805). Demonstrates training-data extraction under studied conditions. - Nasr, M. and colleagues (2023), [Scalable Extraction of Training Data from (Production) Language Models](https://arxiv.org/abs/2311.17035). Further extraction research. Neither paper shows that an arbitrary private upload is necessarily retrievable by another user; neither supports a blanket impossibility claim. ## Context-specific requirements - Fair Work Ombudsman, [Bullying and harassment](https://www.fairwork.gov.au/employment-conditions/bullying-and-harassment). Australian workplace guidance for checking the scope of a real issue. Fictional pack policies are not statements of law. - European Union, [General Data Protection Regulation](https://eur-lex.europa.eu/eli/reg/2016/679/oj/eng). Consult the applicable provisions and competent advice; consent is not the only possible legal basis. - US Department of Education, [Eligible Student Guide to FERPA](https://studentprivacy.ed.gov/resources/eligible-student-guide-family-educational-rights-and-privacy-act-ferpa). Relevant rights and exceptions require context-specific interpretation. These links were reviewed for this September 2026 revision. Reachability is separate from applicability. Before deployment, recheck current rules, product terms and the institution's own requirements. ## Companion methodology Michael Borck, [Conversation, Not Delegation](https://michael-borck.github.io/conversation-not-delegation), develops the wider conversation and verification approach. The present book supplies its own short recap and teaching materials; reading the companion is optional.