w2183821 · Data in Communication and Society

Generative AI use log

Authorship

Diyar Talal (w2183821) is the author of this project. The following are his own work:

Concept and critical argument: the idea of The Bigger Picture, showing through five ordinary stories how behaviour becomes data, profile and advertising.
Logic: the principle that the system must infer from behaviour and never ask, the signals to use (watching, skipping, liking, replying, replaying), and the corrections to its reasoning after his own testing.
Roadmap: the plan and order of work across every stage, set through his briefs and feedback.
Design: the visual direction, reference images, texture, homepage composition and cursor-image effect, and the four-stage structure of the results.
Architecture: the requirements for a static site with no backend, a separate rule-based engine, a tagged video library, local-only storage, return-visit memory and data erasure.
Content: the first story (his own video), the choice of academic readings, and all final editorial decisions.

Tool used

Claude Code (Anthropic), 13–15 November 2026, used as a coding assistant: it wrote code to his specification, searched for and checked YouTube Shorts, and drafted interface text that he reviewed and directed.

Add any other generative AI tools used (for example for images or proofreading) before submission.

How outputs were checked

Each version was tested by the author on the live site; his findings directed the next round of changes.
The recommendation rules are written out in the code and README and were tested with scripted viewers.
Every video was checked as public and embeddable through YouTube’s own endpoints.
Quotations are used only where wording and page were checked; other readings are paraphrased.

Prompts

Date	Prompt (summarised)	Output and how I used it
13 Nov 2026	Build “The Bigger Picture”: five Instagram-style stories, a rule-based recommendation engine, about 25 real YouTube Shorts, and a final reveal of the data collected.	First working version: story player, engine, video library, reveal. Used as the base.
13 Nov 2026	How to run the site locally and publish it online.	Instructions; published on Netlify.
14 Nov 2026	Critique of the first version: remove polls that ask about interests, infer from behaviour only, rethink the results, link to readings. Asked for a critical assessment before coding.	A written assessment and new direction, which I reviewed.
14 Nov 2026	Implement the new direction: return-visit memory with erase, light homepage disclosure, adverts only when plausible.	New engine with confidence levels, memory, erase.
14 Nov 2026	Bug report from my own test: story 2 ignored a clear London signal. Asked for smarter reasoning, more videos, and a clearer view of data given away.	Engine fixes, library grown to 65 videos, given-versus-taken count.
14 Nov 2026	Redesign from my reference image and texture; cursor-following images; four-stage results.	New visual identity and results structure.
14 Nov 2026	Remove the static collage, smoother cursor trail, meet the module’s formatting guidance, integrate set readings.	Trail rewritten; citations, references and this log added.
15 Nov 2026	Passive viewers all saw the same stories. Asked whether cookies or history could be read; decided to use context signals a page can legitimately see (time, day, device, time zone, referrer).	Context hunches added to the engine and shown in the results. Current version.