All Work

Ballsun-Stanton, B., Torrington, J., & Waked, L. (2026, September 4). AI Prompting Framework [Poster]. https://doi.org/10.5281/ZENODO.22290781

Poster

Read abstract

The AI Prompting Framework is a structured, six-stage guide (numbered 0 to 5) for working with generative AI in a way that keeps the user's knowledge, expertise, and judgment at the centre of the process. Rather than treating AI as a source of ready-made answers, it positions the tool as a collaborator whose output the user directs, scrutinises, and ultimately owns.

The framework moves from an initial decision about whether AI is appropriate for the task at hand (Step 0), through the preparation of a clear, well-contextualised first input with a built-in self-check (Step 1), to reviewing that first response (Step 2), interacting iteratively while retaining control of the exchange (Step 3), reflecting on whether the final output genuinely reflects the user's own knowledge (Step 4), and finally documenting effective prompts and being transparent about AI use (Step 5). Each stage is paired with a reflective question, such as "Have I provided enough context?" and "Am I happy to put my name to this?", that prompts users to pause and exercise judgment throughout.

Designed to be general-purpose, the framework applies across disciplines and professional contexts wherever people want to use generative AI deliberately, critically, and accountably.

Ballsun-Stanton, B., & Torrington, J. (2026, September 4). AI-Locus of Control Continuum [Graphic]. Zenodo. https://doi.org/10.5281/zenodo.17823627

Diagram

Read abstract

Our proposed AI-Locus of Control framework has emerged from our recent longitudinal study on the impact of explicit teaching about AI literacy on students' agency and control beliefs in human-AI interactions (https://osf.io/preprints/edarxiv/6mke5_v3). From our analysis, we identified four key dimensions or tensions in AI-Locus of Control:

AGENCY: Does the AI have agency, or does the human retain it? Students with external LOC felt they had to either accept AI output or reject the entire idea of using AI. Students with internal LOC realised they could steer, adjust, and control output on a sentence-by-sentence basis.

EXPERTISE: Is AI better than human, or an input to human judgement? Some students initially treated AI as an expert that couldn't be questioned. Through the unit, they learned that their own prompting and setup was crucial to producing output they felt comfortable claiming as their own.

EMBODIMENT: Is there a mind behind the mirror, or is it a tool to be used? Early in the semester, students engaged with AI as if there was an intelligence behind it. By the end, they understood they held sole responsibility for populating the context window with facts and scaffolding.

PRAXIS: Automation versus augmentation? Some students used AI to automate their assignments entirely. Others learned to dip in and out, using it for partial ideation, sense-checking, and editing - augmenting rather than automating their process.

These four tensions give us a richer understanding of what AI-LOC means in practice.

Ballsun-Stanton, B. (2026, August 4). AI Multiplies Judgement: Using AI ethically and well in research practice. https://doi.org/10.5281/ZENODO.21781864

Presentation

Read abstract

AI multiplies a researcher's judgement without caring about the sign, so the discipline built around the model matters more than the model. These are the slides from a one-hour lecture to the South Australian Research and Development Institute on using generative AI well in research practice. It opens on workslop, Niederhoffer and colleagues' term for AI output that looks polished while lacking the substance to advance the task (Harvard Business Review, 2025), and asks why capable people send it, since the reason they give is being stretched too thin. From there it sets out the prompting mechanisms that make a model interrogate your thinking rather than flatter it, and closes the first half on a live test of Macquarie's enterprise Copilot, which critiqued a bounded document competently and bluffed on an unbounded question whose sources it never looked at.

The second half works through Macquarie University's 2023 Guidance Note on generative AI in research and the 2025 disclosure and documentation Checklist, covering what each was written to protect, how disclosure scales from a footnote to an appendix, and why AI use is acknowledged rather than cited, since a model cannot be fired and so cannot take the blame. A registered report written under both supplies the worked example, with a use-of-AI declaration spanning roughly 120 sessions of drafting, simulation coding, and citation checking, and with the survey responses sealed from the models on the same terms as from the authors. The lecture ends on what makes the practice ordinary in a team, which is annotated conversations, shared prompts, and leaders going first.

Ross, S., & Ballsun-Stanton, B. (2026, July 30). Paper B: Reliability in research with large language models is a property of the human–AI system. OSF. https://doi.org/10.17605/OSF.IO/M376W

Preprint

Read abstract

Companion paper B (submitted to Royal Society Open Science, which uses open peer review): a methodological case study of large language model (LLM)-assisted research, arguing that reliability must be engineered into the human–AI system rather than expected of models alone. This component holds the preprint, electronic supplementary material, and the AB+ literature-grounding pipeline and outputs; shared data live in the Software Tools Dataset and LLM Research-Assistance Corpus components.

Wang, H.-C., Lai, J., Ballsun-Stanton, B., Van Bergen, P., Waked, L., Green, G., Colenbrander, D., Jones, T., & Robinson-Jones, C. (2026, July 17). AI-enabled Academic Vocabulary Learning for Multilingual Children. EdArXiv. https://doi.org/10.35542/osf.io/p8mab_v2

Preprint

Read abstract

Vocabulary knowledge is fundamental to academic achievement, yet multilingual learners with English as an Additional Language or Dialect (EAL/D) consistently demonstrate vocabulary gaps compared to monolingual peers. Despite evidence supporting explicit vocabulary instruction, scalable and personalised interventions remain limited. This study evaluated an AI-enabled vocabulary learning web application designed to pre-teach academic vocabulary to EAL/D students and support curriculum content learning. The application utilises multimodal large language models (LLMs) for conversational instruction, automatic speech recognition, response evaluation, and text-to-speech production. Thirty-one multilingual learners in Grades 2–4 were recruited in Sydney, Australia. Using a counterbalanced within-subjects design, 24 target science vocabulary words were assessed at pre-test. Participants then received AI teaching for 12 of these words over six at-home sessions across 2 weeks (the remaining 12 served as untrained control), followed by a matched post-test. Over the following 2 weeks, all participants received content lessons on four science topics, two containing pre-trained vocabulary and two that did not. Comprehension was assessed via topic-specific questions. Linear mixed-effects models showed significantly AI-training effect (d = 1.45), compared to the untrained words, and significantly higher comprehension accuracy for lessons paired with trained vocabulary (d = 0.21). Students and parents viewed the application positively, particularly its personalised, multilingual feedback, while highlighting technical reliability as an area for improvement. These findings provide initial evidence that AI-enabled, curriculum-aligned vocabulary pre-teaching can support EAL/D students’ academic vocabulary development and content learning, warranting evaluation at scale in school settings.

Ballsun-Stanton, B., Torrington, J., & Meissner, E. (2026, July 13). Registered Report: A Longitudinal Study of AI-Locus of Control in Higher Education. OSF. https://doi.org/10.17605/OSF.IO/CS8BH

Registered report

Read abstract

The study measures whether students' beliefs and orientation about AI shifts across a semester of teaching. This orientation, a student's AI-Locus of Control, is whether they direct the AI tool or defer to it. The design is longitudinal, mixed-methods, and Bayesian. It runs across three cohorts at one university. Both surveys are collected and sealed, and no response has been examined. This deposit lets a reviewer inspect the parts of the analysis plan fixed before the data are seen: the survey instrument, the qualitative codebook, the design simulations that set the evidence thresholds, and the record of how AI assisted the work. The survey responses are not here. They are sealed now and embargoed to Stage 2.

Ballsun-Stanton, B., Johnston, V. H., Jensen, H. S., Kjeldsen, C. K., & Thøgersen, J. (2026, April 15). LLM Document Discovery (Version Container) [Computer software]. Zenodo. https://doi.org/10.5281/ZENODO.19591245

Software

Read abstract

LLM Document Discovery v0.2.0 — Apptainer Container Pipeline

Reproducible pipeline for classifying historical web documents (1996–2005) using large language models. Extracts linguistic and structural features from children's web

content archived by the Internet Archive, producing a structured SQLite database of classifications with supporting blockquote evidence.

This release adds a containerised execution pipeline using Apptainer/Singularity, enabling reproducible deployment on both local GPUs and HPC clusters (NCI Gadi).

What's new in v0.2.0:

Apptainer container: wrapping vLLM v0.19.0 + llm-discovery for fully offline, reproducible execution on HPC compute nodes

Single-command deployment: deploy assembles the data directory, syncs to HPC, and submits the PBS job

Multi-model support: tested with google/gemma-4-E4B-it (local RTX 4090), google/gemma-4-31B-it (Gadi V100), and openai/gpt-oss-120b (Gadi H200)

Prompt engineering: rationale-first instruction format achieving 98% structured output compliance (up from <1%)

Crash-safe resumability: killed containers resume from the last completed document–category pair

CLI commands: build, init, download-model, deploy, status –watch, retrieve, for end-to-end HPC lifecycle management

Offline tokeniser support: tiktoken vocab files baked into the container for OpenAI gpt-oss model family

Associated paper:

Johnston, V.H., Ballsun-Stanton, B., Jensen, H.S., Kjelsen, C.K., & Thøgersen, J. (2026). Exploring the Archived Web through AI-Assisted Document Discovery. https://github.com/WEB-CHILD/exploring-the-archived-web-through-ai-assisted-document-discovery

Johnston, V. H., Ballsun-Stanton, B., Jensen, H. S., Kjeldsen, C. K., & Thøgersen, J. (2026, April 13). Exploring the Archived Web through AI-Assisted Document Discovery. Journal of Digital History. https://journalofdigitalhistory.org/en/notebook-viewer/JTJGcHJveHktZ2l0aHVidXNlcmNvbnRlbnQlMkZXRUItQ0hJTEQlMkZleHBsb3JpbmctdGhlLWFyY2hpdmVkLXdlYi10aHJvdWdoLWFpLWFzc2lzdGVkLWRvY3VtZW50LWRpc2NvdmVyeSUyRm1haW4lMkZhcnRpY2xlLmlweW5i/?v=3

Paper

Bower, M., Torrington, J., & Lai, J. W. M. (2026, February 2). 2026 Typology of Generative AI Tools for Education. EdArXiv. https://doi.org/10.35542/osf.io/4eqrk_v1

Preprint

Read abstract

The Typology of Generative AI Tools for Education provides educators with a list of generative AI tools arranged into nine categories that are currently being used in educational contexts. This typology follows the previously published Typology of Free Web-based Learning Technologies (Bower &amp; Torrington, 2020) and the Typology of Web 2.0 Learning Technologies (Bower, 2015), representing the evolution of educational technology into the generative AI era. To create this typology, 211 educators from nine countries spanning early childhood through higher education completed a survey in late 2025 about their generative AI tool usage. Tools that were reported by two or more educators were included in the typology. A total of 50 unique AI Educational tools were included in the Typology, and have been arranged into nine overarching categories: General-Purpose Large Language Models, Image Creation Tools, Audio and Music Generation Tools, Video Generation Tools, Presentation Generation Tools, Research and Study Tools, AI Tutoring and Chatbots, Custom Education Tools, and Other Miscellaneous Tools. Brief descriptions and links are provided for each tool to support educators in making informed decisions about which tools might suit their teaching and learning contexts. The generative AI landscape is rapidly evolving, and it is noted that some tools offer functionality across multiple categories. This typology represents a snapshot of educator-reported usage patterns in 2025-2026, and offers educators a touchstone for the appropriate selection of Generative AI technologies in their teaching.

Webb, M., Bower, M., Carvalho, A. A., Røkenes, F. M., Torrington, J., Cohen, J. D., Chtouki, Y., Maccallum, K., Linden, T., Butler, D., Raffaghelli, J. E., Vartiainen, H., Ronci, M., Tiernan, P., Smith, D. M., Shelton, C., Malyn-smith, J., & Gorissen, P. (2026, January 13). Thematic Working Group 5 – Artificial Intelligence (AI) literacy for teaching and learning: design and implementation. arXiv. https://doi.org/10.48550/ARXIV.2601.08380

Preprint

Read abstract

TWG 5 focused on developing and implementing effective strategies for enhancing AI literacy and agency of teachers, equipping them with the knowledge and skills necessary to integrate AI into their teaching practices. Explorations covered curriculum design, professional development programs, practical classroom applications, and policy guidelines aiming to empower educators to confidently utilize AI tools and foster a deeper understanding of AI concepts among students.

Smith, S. M., Tate, M., Freeman, K., Walsh, A., Ballsun-Stanton, B., & Lane, M. (2026, January 2). A university framework for the responsible use of generative AI in research. Journal of Higher Education Policy and Management, 48(1), 17–36. https://doi.org/10.1080/1360080X.2025.2509187

Paper

Ballsun-Stanton, B., & Torrington, J. (2025, October 17). Effective AI Prompting Strategies from the Humanities [Poster]. https://doi.org/10.5281/ZENODO.17373463

Poster

Read abstract

Poster to support successful prompting by teachers and researchers.

Good AI prompts are essential to predictably and consistently getting a desired output.There are no single phrases that work across all situations - AI will simply amplify the user’s capabilities, creativity, judgement and discernment.

These effective prompting strategies will help develop your own skills and strategies with AI prompting.

Text:

Start with a goal: what is the single thing you are trying to achieve?

Add context and details: requirements, instructions and parameters. AI needs clear scaffolding so it’s set up for success. Don’t ask it to search for facts - it’s not a search engine.

State your role, so it knows what register to use. The use of words influences its sample and response. Tell it what you want and who the output will be for.

Work as thinking partners: Tell the AI to “Ask me one critical question at a time until we achieve our goal”

Check for clarity: Include a final sentence for the AI to scaffold the task to make sure what you said is what you meant. “To begin, please functionally decompose the task.”

Use examples. AI is effective at imitation, and if you give it a few examples of what you want, it can copy their form.

Longer prompts, rather than shorter: Make sure that you provide all the details it needs. Edit your prompts if errors occur, or start over.

Clear expectations: At the end of each prompt, end with a single, specific task.

Have a real conversation: Go back and forth a few times before you reach your answer. Asking it to ask you for clarification highlights blank spots in your thinking.(Also, you can prompt it to help you design prompts…)

Avoid “AI-slop” - Would you be comfortable putting your name to this? Have you interacted using some effort? Make sure you know enough to be able to validate the output.

Ballsun-Stanton, B. (2025, September 25). How Teaching Must Become Process-Oriented instead of Knowledge-Oriented. https://doi.org/10.5281/ZENODO.17171573

Presentation

Read abstract

Keynote for "eTeach-Jahrestagung” | 2025 Sep 25

Slides in German and English, plus  prompt logs of Claude use for development

Resources used:

https://osf.io/preprints/edarxiv/6mke5_v3

https://doi.org/10.1080/1360080X.2025.2509187

https://doi.org/10.1007/s44204-025-00247-1

https://zenodo.org/communities/iacap-aisb-25-teachingai/

https://osf.io/rd24y/

https://denubis.github.io/KI-Summercamp-2025/basic/

https://dataverse.ada.edu.au/dataset.xhtml?persistentId=doi:10.26193/CDCZA7

Torrington, J., Ballsun-Stanton, B., & Lai, J. W. M. (2025, July 28). Teaching Students How to Effectively Interact with LLMs at University: Insights on the Longitudinal Development of AI-Locus of Control. EdArXiv. https://doi.org/10.35542/osf.io/6mke5_v3

Preprint

Read abstract

Background: As AI disrupts education systems worldwide, there is an urgent need for students to develop AI literacy while maintaining human agency. While extensive research has examined self-efficacy in technology contexts, there is a notable gap in understanding how locus of control (LOC) specifically applies to human-AI interactions in educational settings. LOC is defined as the belief that outcomes are contingent on one's own behaviour versus external forces. Objectives: This study examined how explicit instruction and guided practice with multiple frontier Large Language Models (LLMs) impact students' agency and control beliefs in human-AI interactions. Methods: A longitudinal qualitative case study followed 12 undergraduate students through a semester-long experimental AI unit. Data collected through surveys, focus group discussions, and weekly observations were analysed thematically, using LOC theory as an interpretive framework. Results and Conclusions: Students demonstrated a clear progression toward internal AI-LOC, developing critical awareness and agency in their AI interactions. The findings revealed that sustained, scaffolded AI instruction enabled students to view themselves, rather than the technology, as the primary driver of outcomes. This study contributes to AI literacy theory by demonstrating that locus of control, rather than self-efficacy, may be the critical construct for understanding human agency in AI interactions, introducing the concept of AI-LOC as a new theoretical framework for education research.

Ballsun-Stanton, B., Torrington, J., Laurence, R., Khalid, M., & Atkin, A. (2025, June 30). Exploring Student and Faculty use of Generative AI and Large Language Models in Classrooms and Research, ARTS3500 Classroom Transcripts and Assessments Semester 2 2024 (Version 1.0) [Dataset]. ADA Dataverse. https://doi.org/10.26193/CDCZA7

Dataset

Read abstract

This dataset contains educational materials from an AI-focused course, including: - Student assessments from AI History, AI Philosophy, and AI Politics streams - Class transcripts from various sessions - Course materials, schedules, and rubrics

Ballsun-Stanton, B., & Khalid, M. (2025, June 16). The Emperor’s New Clothes: A Manifesto for Universities in an AI-Haunted World. https://doi.org/10.5281/ZENODO.15671962

Paper

Read abstract

Universities face an existential crisis revealed, but not caused, by artificial intelligence. Using Lakatos' Philosophy of Science, we argue higher education has become a degenerating research programme, maintaining formal structures while abandoning educational substance. Drawing on evidence from an experimental AI-integrated unit at Macquarie University, we pose three interlocking questions: What constitutes content-increasing education when AI handles information transfer? How can universities reconnect education to meaningful world-making? What academic practices enable transformative rather than transactional learning? Student transformations documented in our unit suggest design principles for post-transactional universities that embrace productive failure, reward process mastery, and develop new capabilities beyond credentialism.

Burns, E., Torrington, J., & Bower, M. (2025, June 11). Complementary or contradictory? The double-edged sword of AI’s impact on effort regulation in higher education. EdArXiv. https://doi.org/10.35542/osf.io/dh6rj_v1

Preprint

Read abstract

Effort regulation, or the ability to manage and control one’s effort during learning, is a consistent predictor of academic achievement in higher education. The rapid rise of artificial intelligence and its impact on higher education learning environments raises questions about how students are adjusting how they apply and manage effort in their learning. Examining the dynamic interplay between use of AI and effort regulation is necessary to develop effective higher education practices and policies that support improved learning processes for students in the context of AI. This study employed a mixed methods design to explore 1,316 undergraduate pre-service teachers’ perceptions of the relationships between effort regulation and AI use. Quantitative findings showed that students with stronger effort regulation were significantly less likely to use AI. Similarly, qualitative findings showed over half of the sample (52.59%) perceived that using AI could lead to overall disengagement from university work, learning loss, and lack of skill development. When triangulated, findings suggested that students who perceived effort regulation as critical to their learning and skill development were less inclined to use AI because they perceived AI use as undermining their effort regulation and meaningful engagement. These findings have implications for higher education practice: these suggest a need to explicitly communicate to students the importance of effort regulation in learning and the appropriate ways in which AI can be used to enhance effort regulation.

Ballsun-Stanton, B., & Torrington, J. (2025, June 3). Teaching the Unknown: A Pedagogical Framework for Teaching With and About AI. Zenodo. https://doi.org/10.5281/ZENODO.15583013

Preprint

Read abstract

Generative artificial intelligence (AI) has disrupted education systems worldwide. This disruption necessitates pedagogical approaches that embrace uncertainty while developing student agency. We examined how decoupling task success from assessment outcomes created environments where students developed critical AI literacy through structured risk-taking. Drawing on Transformative Learning Theory and Rumsfeld's epistemological matrix as interpretive frameworks, we analysed an experimental undergraduate AI unit across three disciplinary streams: Ancient History, Philosophy, and Politics and International Relations (N=23N=23). Data included student reflections, classroom observations, and AI interaction logs collected over a 13-week semester. Our pedagogical framework operationalised four interconnected pillars: risk-embracing assessment structures, intentional classroom culture development, systematic navigation of technological uncertainty, and facilitation of transformative learning experiences. This paper presents the implications of these pillars for 1) educational theory, where productive failure serves as an effective pedagogical strategy; 2) educator praxis, viewing AI as a textual technology that extends the capabilities of the humanities; and 3) implications for the university teaching context, where AI-enabled teaching should focus on reflection and process rather than demonstratable competencies.

Ballsun-Stanton, B., & Smith, S. (2025, April 30). Generative AI in Research: Disclosure & Documentation Guidance / Checklist. Macquarie University. https://doi.org/10.5281/ZENODO.15307492

Guidance

Read abstract

The use of generative AI in research needs to align with the principles and responsibilities outlined in the Australian Code for the Responsible Conduct of Research 2018. Guidelines about the appropriate disclosure and documentation to accompany the use of Generative AI in research also need to recognise the practical considerations of current or innovative research practices. An approach combining these elements is detailed in this document for Macquarie University researchers.

Ballsun-Stanton, B. (2025, February 18). Research Use of LLMs: Beyond the chatbot. https://doi.org/10.5281/ZENODO.14885316

Presentation

Read abstract

Slides for the Research Data Alliance event:

Title: AI in Action: How Researchers Leverage AI (Asia/Oceania Friendly Time)

https://www.rd-alliance.org/event/ai-in-action-how-researchers-leverage-ai-europe-americas-friendly-time-2/

Kudina, O., Ballsun-Stanton, B., & Alfano, M. (2025, February 1). The use of large language models as scaffolds for proleptic reasoning. Asian Journal of Philosophy, 4(1), 24. https://doi.org/10.1007/s44204-025-00247-1

Paper

Read abstract

Abstract This paper examines the potential educational uses of chat-based large language models (LLMs), moving past initial hype and skepticism. Although LLM outputs often evoke fascination and resemble human writing, they are unpredictable and must be used with discernment. Several metaphors—like calculators, cars, and drunk tutors—highlight distinct models for student interactions with LLMs, which we explore in the paper. We suggest that LLMs hold a potential in students’ learning by fostering proleptic reasoning through scaffolding, i.e., presenting a technological accompaniment in anticipating and responding to potential objections to arguments. Here, the technical limitations of LLMs can be reframed as beneficial when fostering anticipatory reasoning. Whether their outputs are accurate or not, evaluating them stimulates learning. LLMs require students to critically engage, emphasizing analytical thinking over mere memorization. This interaction helps solidify knowledge. Additionally, we explore how engaging with LLMs can prepare students for constructive collective discussions and provide first steps in addressing epistemic injustices by highlighting potential research blind spots. Thus, while acknowledging the sociopolitical and ethical complexities of using LLMs in education, we suggest that when used in an informed way, they can promote critical thinking through anticipatory reasoning.

Ballsun-Stanton, B., & Hipólito, I. (2024, July 1). Is the “Calculator for Words” analogy useful for communicating about LLMs? https://doi.org/10.5281/ZENODO.12602858

Paper

Read abstract

Large language models (LLMs) are fundamentally different from search engines, functioning more as 'vibe-machines' than information retrieval systems. However, conveying appropriate expectations and usage modes for these novel interfaces remains challenging. This paper critically examines Willison's 'calculator for words' analogy and Bucci's counter-arguments, analysing the strengths and limitations of this metaphor.

While we argue that the 'calculator for words' analogy serves as an effective negative heuristic – discouraging users from treating generative AI prompts as search engine queries – it falls short in providing positive intuition for effective LLM utilisation. To address this limitation, we propose a novel conceptual framework: 'maps of no territory.'

Drawing inspiration from Borges' 'On Exactitude in Science,' our 'maps of no territory' analogy aims to provide more nuanced intuitions for general audiences, guiding them towards effective use while steering them away from problematic applications. This metaphor offers a more comprehensive understanding of LLMs' nature, capabilities, and limitations, potentially fostering more informed and responsible engagement with these powerful AI systems.

Head, A., & Ballsun-Stanton, B. (2024, April 4). Use of Generative Artificial Intelligence in Undergraduate Assessment. Macquarie University. https://doi.org/10.5281/ZENODO.10916388

Preprint

Read abstract

This is a policy currently in use in an earlier form in the Faculty of Arts, Macquarie University, Sydney Australia. It is use-agnostic and sets parameters for risks, communication of risks, and possible benefits in a bid to both set guardrails and educate students. Some modification will be needed for use outside of Legal education. We are serious about the recommendation about models.

Here is the document in HTML form:

Use of Artificial Intelligence

The use of Artificial Intelligence is not prohibited. The use of Large Language Models (LLMs) such as OpenAI's ChatGPT GPT-4 (paid), Perplexity.ai (free/paid), Microsoft Copilot in creative mode (free), or Claude 3 Opus (paid) when working on this assessment is permitted. These tools can complement your efforts on assessments and can assist with planning, research, and editing, but they must be used intentionally and with utmost care. Intentional and careful use of these tools may assist the quality of your final submission, but poor or reckless use of the tools can quite easily negatively impact the quality of your submission. You are fully responsible for any issues or errors arising from their use. If you are considering actively engaging with LLMs to assist in completing this assessment, then please read the following very carefully.

Acknowledgement

It is essential to acknowledge any tool used, both the model used and the way that you used it. (See below for my acknowledgement.)

CRITICAL: Confabulations, hallucinations, and fictitious sources

It is your responsibility to use Generative AI tools ethically and appropriately. Any fictitious sources contained in your submitted paper will result in a failure of the assessment, regardless of whether they originated from your own research, Generative AI, or a random webpage. This is not an academic integrity issue, but a matter of ensuring the accuracy and reliability of your work.

Remember, LLMs always sound confident but are not always correct (and depending on how the LLM is used may be very wrong). Proper prompting is essential for improving the quality of the results (see below for resources on prompting).

Embedded models

Microsoft Word and Google Docs have built-in text autocomplete features powered by LLMs. Use this "autocomplete" feature at your own risk, as it may result in bland and generic writing.

Vague responses

Poor prompting may lead to American-centric or vague responses that do not address the specifics of Australian law. When using these models, it is your responsibility to thoroughly check all outputs to ensure they are relevant and accurate. Reliance on an LLM output without thorough oversight is strongly advised against.

Copyright concerns

As future lawyers, it is your responsibility to obey the law. Do not upload copyrighted material, including unit materials (slides, articles, question text) or anything with a clear copyright, to these international systems without an appropriate license (e.g., CC-BY). Always read and comply with the terms of use of any research services you use.

Appropriate and ethical use

Read the terms of service of the LLM tools you use. Some, like OpenAI (https://openai.com/policies/sharing-publication-policy), have specific acknowledgement requirements for your work.

While powerful, these tools may not always increase productivity or research speed (https://www.hbs.edu/faculty/Pages/item.aspx?num=64700). Use them judiciously.

A trap

Do not treat LLMs as search engines. Even those with web search capabilities (e.g., perplexity.ai, ChatGPT GPT+ subscription, Microsoft Copilot) may not search effectively. Ensure that all factual information you want the models to work with is well-contained within your prompts.

Recommendations

We strongly advise against using ChatGPT's free version, as it may lead to unsatisfactory results and is prone to confabulations. Use tools running GPT-4 or equivalent, such as Microsoft Copilot in creative mode (free with a throwaway account, not your MQ account). Avoid upsells from the LLMs and choose "creative mode" for GPT-4; "balanced mode" uses GPT 3.5.

Here are recommended sources to help you use these powerful but potentially dangerous tools:

A short bibliography on prompting for those who are interested.

https://www.oneusefulthing.org/p/a-guide-to-prompting-ai-for-what

https://www.oneusefulthing.org/p/captains-log-the-irreducible-weirdness

https://simonwillison.net/2023/Apr/2/calculator-for-words/

Acknowledgement

Claude 3 Opus was used to edit this text on AI usage to remove many of the Americanisms introduced by my colleague.

Bower, M., Torrington, J., Lai, J. W. M., Petocz, P., & Alfano, M. (2024, January 26). How should we change teaching and assessment in response to increasingly powerful generative Artificial Intelligence? Outcomes of the ChatGPT teacher survey. Education and Information Technologies. https://doi.org/10.1007/s10639-023-12405-0

Paper

Read abstract

Abstract

There has been widespread media commentary about the potential impact of generative Artificial Intelligence (AI) such as ChatGPT on the Education field, but little examination at scale of how educators believe teaching and assessment should change as a result of generative AI. This mixed methods study examines the views of educators ( n  = 318) from a diverse range of teaching levels, experience levels, discipline areas, and regions about the impact of AI on teaching and assessment, the ways that they believe teaching and assessment should change, and the key motivations for changing their practices. The majority of teachers felt that generative AI would have a major or profound impact on teaching and assessment, though a sizeable minority felt it would have a little or no impact. Teaching level, experience, discipline area, region, and gender all significantly influenced perceived impact of generative AI on teaching and assessment. Higher levels of awareness of generative AI predicted higher perceived impact, pointing to the possibility of an ‘ignorance effect’. Thematic analysis revealed the specific curriculum, pedagogy, and assessment changes that teachers feel are needed as a result of generative AI, which centre around learning with AI, higher-order thinking, ethical values, a focus on learning processes and face-to-face relational learning. Teachers were most motivated to change their teaching and assessment practices to increase the performance expectancy of their students and themselves. We conclude by discussing the implications of these findings in a world with increasingly prevalent AI.

Ballsun-Stanton, B., Smith, S. M., & White, K. (with Anson, D., Connor, M., Cornish, J., D’Arcens, L., Kim, J., Singh, A., & Todd, V.). (2023, October 22). Guidance Note: Using Generative Artificial Intelligence in Research. Macquarie University. https://doi.org/10.5281/ZENODO.10851623

Guidance

Read abstract

This is the university policy/guidance note for the use of Generative AI and LLMs at Macquarie University, Australia. The original source is https://policies.mq.edu.au/download.php?associated=1&id=768&version=1

Here is the summary:

Generative AI (Artificial Intelligence, Large Language Models) offers the unprecedented ability tomanipulate and generate text and media in response to arbitrary instructions. These new capabilities offeropportunities and risks to researchers. This document will discuss responsible use, risk mitigation, andappropriate use of these tools. The technological landscape changes quickly and new tools are releasedalmost weekly – this guide offers general advice which should be applied thoughtfully.

Regardless of any tools or technologies used, now or in the future, everyone at Macquarie University isresponsible for ensuring their research meets the expectations of the Australian Code for the ResponsibleConduct of Research and the Macquarie University Code for the Responsible Conduct of Research(2018).

Generative AI must be used with caution, and its use is currently inappropriate in some researchprocesses because Generative AI services, including ChatGPT:• cannot meet the requirements for authorship• can create authoritative-sounding outputs that may be incorrect, incomplete, or biased• could inappropriately capture sensitive data (including, but not limited, to personal information).

Researchers must not use Generative AI:• to perform peer review activities• to generate substantive content of research outputs, including HDR theses• for writing the critical components of human ethics, animal ethics, or biosafety applications.

Researchers must exercise care in the use of Generative AI in other aspects of their research and should:i. only do so with the written agreement of their research collaborators (& HDR supervisors)ii. review and consider the terms of service/license of the platforms used and any models usediii. consider the current issues and understandings around copyright and intellectual propertyiv. mitigate risks around the insecure storage or unauthorised re-use of sensitive datav. exert oversight and control when using the technologyvi. carefully and critically review the output and results created by Generative AIvii. take responsibility for the integrity of the content altered or created using Generative AIviii. disclose the use of Generative AI to potential publishers and in disseminated research outputsix. read and follow the policies of publishers and funders regarding the use of Generative AI.

Ballsun-Stanton, B. (2023, June 6). Large Language Models Workshop. OSF. https://doi.org/10.17605/OSF.IO/RD24Y

Workshop

Read abstract

This two-hour workshop provides a comprehensive introduction to the world of Large Language Models (LLMs), focusing on the recent advancements in Generative AI. Participants will gain insights into the development and functionality of prominent LLMs such as Bing Chat and ChatGPT. The workshop will delve into the concept of LLMs as "calculators for words," highlighting their potential to revolutionize ways of working and learning.

The session will explore the principles of Prompt Engineering and Transactional Prompting, demonstrating how consistent prompts can yield reliable and reproducible results. Participants will also learn about the practical applications of LLMs, including editing and proofreading papers, generating technical documentation, recipe ideation, and more.

The workshop emphasizes the importance of understanding the terms of use and the responsibilities that come with using these powerful AI tools. By the end of the session, participants will be equipped with the knowledge and skills to effectively use LLMs in various contexts, guided by the mantra that a LLM is "Always confident and usually correct."

A recording of one workshop is available on Youtube: https://www.youtube.com/watch?v=c3_P7fVjiK8

Torrington, J., Bower, M., & Burns, E. C. (2023, May 26). Elementary students’ self‐regulation in computer‐based learning environments: How do self‐report measures, observations and teacher rating relate to task performance? British Journal of Educational Technology, 55(1), 231–258. https://doi.org/10.1111/bjet.13338

Paper

Read abstract

Abstract

It is well‐established that being a self‐regulated learner is beneficial academically, motivationally and is considered essential for productive life‐long learning. Despite this, there is limited evidence examining how different measures of self‐regulation for learning (SRL) relate to task performance for young students learning in digital contexts. This study investigated the relationships between different measures of SRL of elementary school students ( N  = 48, M age  = 10.75) while using a computer‐based learning environment, and their association with task performance and teacher rating of student SRL ability. Results confirmed the most effective measure of SRL, in terms of its relationship with, and predictability of task performance, was a self‐report written response, whereby students identified and explained known SRL strategies, such as how to plan, monitor or complete their work. Teacher ratings of their students' metacognitive capability were significantly correlated with task performance and with two self‐report instruments: the Junior Metacognitive Awareness Inventory and the SRL written response. Associations between actual observed self‐regulation behaviours in a computer‐based learning environment, using Azevedo et al.'s coding framework and how students self‐reported their knowledge and understanding about SRL, were weak. Observations of young students' SRL behaviours in computer‐based learning environments were not significantly related to task performance. Better understanding of these relationships will help educators and researchers to know where they should focus their attention in terms of developing elementary school students' self‐regulatory capabilities in digital contexts, as well as the reliability of self‐report measures of SRL as relative to observations of self‐regulation and task performance. Implications for teacher practice are also discussed.

Practitioner notes

What is already known about this topic

Students need to use self‐regulation for learning (SRL) strategies in digital contexts.

Self‐regulation strategies need to be explicitly taught to students as they are not guaranteed to become automatically acquired.

Being a self‐regulated learner leads to improved academic performance, engagement and motivation.

What this paper adds

Empirical evidence addressing the associations and patterns between various measures of SRL for young students learning in digital environments.

Student self‐report explanations of known SRL strategies was the only significant predictor of student task performance.

Coding of young students' SRL behaviours in digital environments do not relate well to task performance.

Implications for practice and/or policy

Understanding the associations between young students' self‐report of SRL and how this relates to their actual SRL behaviour while using digital technology is critical to supporting student learning and success.

Learning to better articulate self‐regulation strategies may result in greater consciousness and application of self‐regulation strategies in digital contexts, which in turn could improve task performance.

Eliciting explanations from students about SRL strategies may be more informative and expedient than conducting and analysing individual observations in digital contexts, to determine the general self‐regulatory knowledge and understanding of young students.

Torrington, J., Bower, M., & Burns, E. C. (2022, August 3). What self-regulation strategies do elementary students utilize while learning online? Education and Information Technologies, 28(2), 1735–1762. https://doi.org/10.1007/s10639-022-11244-9

Paper

Read abstract

Abstract

Little is known about the strategies elementary school students use to self-regulate their learning while in a hypermedia environment. This exploratory study investigated the self-regulatory strategies that young students ( N  = 48, M

age

= 10.75) utilized while individually completing a 20-min online research task about space. Video data was coded using Azevedo et al.'s (2004) established coding scheme for analyzing self-regulatory behavior in hypermedia environments. Results showed that young students spent the majority of their time using cognitive strategies ( M  = 75.26%) to read and summarise information to complete the task. Little time was taken to plan ( M  = 6.99%) or monitor ( M  = 5.92%) their work or learning processes, which are key attributes of effective self-regulation. The study reveals the disparity between the ability to navigate within a hypermedia environment and utilizing planning and monitoring processes to enhance learning while using digital tools. This study highlights the need for the explicit teaching of planning and monitoring strategies in order for young students to develop the full range of self-regulation skills they need when using technology, for instance while learning from home during COVID-19. Implications for curriculum policy and teacher practice are discussed.

Torrington, J., & Bower, M. (2021, April 20). Teacher‐created video instruction in the elementary classroom—Its impact on students and teachers. Journal of Computer Assisted Learning, 37(4), 1107–1126. https://doi.org/10.1111/jcal.12549

Paper

Read abstract

Abstract

Despite the proliferation of multimedia devices in elementary classrooms, there is limited research examining teacher‐created video instruction, particularly regarding its effect on academic growth and engagement. This study investigated the effect of teacher‐created computer‐based video instruction (CBVI) using iPads on students' academic, behavioural and affective learning in elementary classrooms. The study used a repeated‐measures design with counterbalancing to measure the effects of CBVI during mathematics lessons on student achievement scores, time on‐task and attitudes towards learning. Three year three classes ( n  = 49) completed three lessons, each using a different mode of instruction: CBVI created by the class teacher, CBVI created by a stranger, and a traditional live lesson delivered by the class teacher. Results were analysed using a Linear Mixed Model. No significant growth in performance was detected during video instruction, however a significant growth result was achieved for the traditional live teaching mode ( p  < 0.001), possibly attributable to the longer duration of experimental session. Behavioural engagement was considerably higher during CBVI lessons than live lessons and students preferred their teacher's voice during CBVI. Three teachers were interviewed to examine how CBVI affected teaching and learning, with two main themes emerging: (1) positive impacts of CBVI upon students; and (2) positive impacts on teacher wellbeing. This research indicates benefits for students and teachers when using teacher‐created CBVI. Further research is needed to better understand the factors that influence cognitive development of students using CBVI and to also further explore the effect of CBVI on teacher wellbeing.

, Lay Description

What is already known about this topic

Video as an instructional method has been used to deliver content to learners for over 40 years.

Current mobile technologies have the inbuilt capability to film, edit, upload and access custom‐made instructional videos.

Learning through video instruction is enjoying renewed popularity due to Internet capability, increased connectivity and device affordances.

Video can promote self‐paced, controlled, personalized learning experiences for students.

What this paper adds

Teacher‐created video instruction has been an overlooked pedagogical approach in mainstream elementary classrooms.

Students are highly engaged behaviourally when completing teacher‐created video lessons.

Students prefer to hear their teacher's voice on the video, rather than a stranger's.

Using teacher‐created video instruction in the classroom positively impacts on teacher wellbeing.

Implications for practice and/or policy

Using teacher‐created video instruction in the classroom provides a way to deliver differentiated content to groups or individuals.

Students have the autonomy to self‐pace and control their learning within the close proximity of the teacher.

Teachers are able to provide individual support and attention to students during the video instruction without affecting general learning.

High student engagement can positively impact the classroom dynamics and environment.