# AI Log 05 — Portfolio Content Refinement ## Purpose of This Log Record AI assistance related to portfolio writing refinement, grammar improvement, content organization, individual contribution clarity, coursework alignment, and ensuring AI did not replace original research or design decisions. ## 1. Date / Project Stage - Date: Week 10 - Project stage: Final portfolio writing refinement / submission preparation - Related portfolio/system section: Project Overview; Why / Gap; User Requirements; Iteration; Technical Implementation; Evaluation & Reflection; AI Tools Used; AI Logs ## 2. AI Tool Used - Tool name: Google Gemini and OpenAI ChatGPT - Model: Gemini 3.1 pro; ChatGPT 5.5 Thinking - Access date: May. 9, 2026 - Link: https://gemini.google.com/ and https://chatgpt.com/ ## 3. Intended Task The AI interaction was used to support final writing refinement across the HoldLight process portfolio. The goal was to make the portfolio clearer, more consistent, and more readable while preserving the group’s own research evidence, design reasoning, and evaluation interpretation. AI support focused on wording, section flow, grammar, and making claims more cautious where accessibility and safety were involved. Gemini was used to explore clearer wording, section summaries, and portfolio organization. ChatGPT was used to check consistency, cautious claim framing, and AI-use disclosure structure. AI was used as an editing and organization assistant. It did not invent research findings, replace user testing, or decide the project’s design direction. ## 4. Prompt Summarized refinement prompt: "Help me refine the writing and organization for my HoldLight process portfolio. Context: HoldLight is an accessibility-focused mobile web prototype for blind and low-vision climbers. The portfolio includes problem framing, user requirements, prototype iteration, technical implementation, evaluation results, final reflection, and AI-use disclosure. Requirements: - Improve clarity, grammar, and section flow. - Keep the tone academic but readable. - Help summarize sections without changing the underlying meaning. - Do not invent user data, testing results, references, citations, or technical features. - Do not generate new research findings, participant feedback, or evaluation conclusions. - Make accessibility and safety claims cautious and evidence-based. - Clarify individual contributions and coursework alignment. - Preserve the group’s original design decisions and research interpretation. - Keep AI disclosure transparent. - Flag any claims that sound too strong or unsupported. Expected output: Suggest concise revised wording, section summaries, and organization improvements for portfolio sections. Treat the output as editing support only; final text must be manually checked against project evidence." ## 5. AI Outcome Gemini and ChatGPT provided revised wording options, clearer section summaries, more consistent terminology, and suggestions for making claims less absolute. They also helped organize AI-use disclosure and individual contribution language. ### What worked - The AI helped make long sections more readable without changing the main meaning. - The wording suggestions improved consistency around terms such as BLV climbers, route preview, audio cueing, fallback, and human support. - The AI helped identify places where claims sounded too broad or too certain. - The output was useful for turning rough notes into more coherent portfolio paragraphs. - The AI helped structure the AI disclosure so prompt use, outcome, verification, and ethics were easier to follow. - Gemini helped produce concise summary alternatives for long portfolio sections. - ChatGPT helped check whether the wording stayed cautious and aligned with the project evidence. ### What did not work - Some generated text sounded too polished and less like the team’s own project voice. - The AI sometimes generalized accessibility claims instead of grounding them in HoldLight’s specific requirements. - Some suggestions risked overstating evaluation results or implying final safety validation. - The AI occasionally proposed extra features or evidence that were not in the project. - Some generated summaries were too broad and needed to be re-grounded in HoldLight’s specific evidence. - Some wording sounded too polished and needed manual rewriting to fit the team’s project voice. - The output needed manual checking against actual pages, images, testing summaries, and references. ## 6. Team Action After AI Output - Modified: We rewrote claims to make clear that HoldLight reduces routine route-information support but does not replace human safety support. - Modified: We clarified the difference between design requirement, prototype feature, testing evidence, and reflection. - Refined: We made wording more consistent across project overview, requirements, implementation, evaluation, and AI reflection sections. - Refined: We improved individual contribution descriptions so they connect each member’s work to concrete project outputs. - Refined: We kept the AI disclosure direct and traceable through the `/ai-logs` folder. - Discarded: We removed unsupported claims, invented metrics, and generic accessibility statements that were not tied to project evidence. - Discarded: We rejected wording that made the prototype sound like a finished safety-critical product. - Kept: We kept concise wording improvements where they made the portfolio easier to read without changing the underlying evidence. ## 7. Verification Against User Requirements - Requirement checked: The portfolio should clearly explain why HoldLight supports blind and low-vision climbers. - Requirement checked: The design requirements should connect to route preview, short audio cueing, fallback, and human help. - Requirement checked: The evaluation and reflection should remain evidence-based and ethically cautious. - Manual test / walkthrough: We read across the main portfolio flow from overview to final reflection to check narrative continuity. - Manual test / walkthrough: We compared AI-refined copy with `content-summary.md` and the actual HTML sections to avoid invented evidence. - Manual test / walkthrough: We checked whether individual contribution wording matched the contribution table and project roles. - Accessibility check: We reviewed whether the copy avoids implying that AI or computer vision can replace human judgement in climbing. - Accessibility check: We checked that BLV-related language stays respectful, specific, and tied to user needs. - Evidence used: `content-summary.md`, `index.html`, `why-gap.html`, `user-requirements.html`, `iteration-alternatives.html`, `implementation.html`, `evaluation-reflection.html`, and the AI references / disclosure sections. ## 8. Ethical / Accessibility / Reliability Considerations - Possible risk: AI-refined writing may make the project sound more complete, reliable, or validated than it really is. - Possible risk: Generic accessibility language may erase the specific needs of blind and low-vision climbers. - Possible risk: AI may accidentally introduce unsupported features, citations, or evaluation claims. - Possible risk: Over-polished language may obscure team contribution and original design reasoning. - How the team reduced the risk: We manually checked wording against project files, testing evidence, and user requirements. - How the team reduced the risk: We kept limitations and human-support boundaries visible in the portfolio. - How the team reduced the risk: We used AI for editing support only and preserved the team’s own interpretation. - How the team reduced the risk: We documented representative prompts in `/ai-logs` so the AI-assisted workflow remains traceable. ## 9. Final Use in Project - Used in files/pages: `content-summary.md`; `index.html`; `brief.html`; `why-gap.html`; `user-requirements.html`; `iteration-alternatives.html`; `implementation.html`; `evaluation-reflection.html`; `testing-iteration.html`; `final-reflection.html`; `ai-logs/README.md`; `ai-logs/01-route-preview-ui-prompt.md`; `ai-logs/02-accessibility-controls-prompt.md`; `ai-logs/03-evaluation-layout-prompt.md`; `ai-logs/04-responsive-debugging-prompt.md`; `ai-logs/05-portfolio-content-refinement-prompt.md` - Final status: Used after manual editing, evidence checking, and coursework-alignment review. - Notes: This AI interaction supported writing polish and organization only. Final portfolio content, design reasoning, research interpretation, and ethical claims were retained as team-led work.