Documenting AI Processes & Providing AI Acknowledgement
In an era where generative AI plays an increasingly prominent role in academic and creative workflows, it is crucial to document the processes undertaken and clearly acknowledge how the work was produced. This transparency ensures a clear understanding of the contributions made by different intelligences and how the author collaborated with AI tools to generate the final outcome. Ethical integrity and responsible AI use require transparent disclosure of the extent and manner in which AI has been employed.
Below are various approaches to documenting and acknowledging AI processes in assessments. Each method should be adapted or modified to align with the specific learning context.
Acknowledging
A Written Declaration
When material generated by artificial intelligence has been adapted or is being used to demonstrate the capabilities of generative AI, in-text citations or references may not be appropriate. In these situations, a declaration should be added that provides written acknowledgment of the use of generative artificial intelligence, specifies which technology was used, includes explicit descriptions of how the information was generated, identifies the prompts used, and explains how the output was utilized in your work.
Read more on Student Academic Success
Checklist Declarations
A checklist declaration offers a quick way to acknowledge how, why, and when AI technologies have been used. The level of detail or checklist categories will be context-dependent. Students could start by identifying the role Generative AI played, such as Mentor, Tutor, Coach, Teammate, Student, Simulator, or Tool. The checklist declaration could also extend to specific tools linked to particular tasks or workflows, such as using ChatGPT for drafting and/or editing purposes. This streamlined approach provides educators with high-level insights into the student's approach.
Structured Reflections
To maximize the effectiveness of reflective writing, educators can guide students using frameworks like Gibbs' Reflection Cycle, specific prompts related to the material, or by incorporating generative AI as a self-reflection mediator. These methods encourage students to describe their experiences, analyze the impact of AI tools on their learning, and critically assess their responses. A structured approach prompts students to interrogate their experience, connect AI use to broader academic and professional contexts, and enhance both their immediate understanding and long-term ability to navigate technological advancements.
Documenting & Acknowledging
Shared Conversations
Shared links are a feature that allows users to generate a unique URL for a ChatGPT conversation. This URL can then be shared with friends, colleagues, and collaborators. Shared links allow users to share their ChatGPT conversations, replacing the old and burdensome method of sharing screenshots.
Graphical Annotations
Graphical annotations can transform static works into living narratives by pinpointing moments where generative AI intersects with human authorship. Whether in creating poems, prose, or surreal images, these visual annotations provide context, demystifying the process for the reader. In digital formats, these annotations could become interactive layers, inviting deeper exploration.
Assessment example
Screenshots or Recordings
When documenting processes involving generative AI, screenshots or recordings are invaluable. These visual snapshots serve as step-by-step guides, capturing each stage of the workflow. Accompanying captions or brief narratives can clarify the steps taken and provide curated evidence of content generated by AI models. Sharing these visual records alongside the final outcomes ensures transparency and potentially enhances reproducibility.
Track Changes and Document Histories
Tracking changes and document history features in platforms like Google Docs or Microsoft Word can be useful for identifying the potential use of generative AI by students. These features reveal significant shifts in writing quality, inconsistencies in style or tone, and rapid content additions, allowing for comparisons between document versions. However, they are not foolproof and may raise privacy concerns. Students might draft work elsewhere, leaving no trace in the document history, or be savvy enough to bypass these checks. Therefore, promoting academic integrity and educating students on the responsible use of AI is crucial for effective monitoring.
Based on Mark Carrigan's post ChatGPT’s advice on examining student essays for evidence of generative AI
Documentation Generators
Scribe, a generative AI tool, streamlines workflow documentation by automating the creation of descriptions, user manuals, and FAQs. It allows users to record working processes across any application, desktop, or web, capturing cross-application processes seamlessly. To capture a Scribe, simply turn on the Scribe recorder via the browser extension or desktop app, walk through your process as usual, and turn it off when finished. Scribe then automatically generates a step-by-step guide with screenshots, text, and cursor clicks. Additionally, Scribe offers a range of editing options, enabling users to edit, reorder, and enhance documentation.
A Collection of Drafts
If an assessment is produced through multiple drafts, then each version could be included with a cover sheet briefly summarising the key differences and/or how the draft responds to feedback provided.
Progressive Overview Maps
Progressive overview maps serve as succinct summaries that capture current thinking and making. They encourage regularly pulling back to ‘map the big picture’ – the research foci, project aims, and compelling questions as understood at a particular moment. This practice helps maintain perspective ‘in action’ and captures the evolution of a project for later reflection. The form these maps take will be context-dependent and could be produced at set or regular intervals, such as daily, weekly, or monthly.
Based on Zoë Sadokierski Progressive Overview Maps from Critical Journal / Contextual Portfolio: A framework for documenting and disseminating RtD as scholarly research
Contributors: Brendan Boniface, Jeremy Breaden, Michael Burke, Fabio Capitanio, Thu Do, Wendy Ellerton, Andrew Junor, Jennifer Mansfield and Sadia Nawaz.