Description
This workshop, Exploring TikTok collections with Generative AI, presented at the Siegen Winter School "AI Methods: From Probing to Prompting" investigates how large language models such as ChatGPT can be operationalised as visual research assistants within digital methods workflows. It focuses on using generative AI to support tasks such as classification, extraction, summarisation, and tagging of visual and audiovisual data, positioning LLMs as “junior research assistants” that can augment but not replace analytical judgment.The workshop combines methodological reflection with hands-on experimentation. It introduces structured prompting techniques (persona–task–format), iterative conversational workflows, and the notion of “prompt design” as a research method, illustrated through a biodiversity case study comparing outputs across generative image models over time. Participants then apply these principles to TikTok datasets on sea level rise, using tools such as 4CAT to generate visual artifacts (e.g., thumbnail walls, word clouds, timelines, hashtag networks, and composites) and employing ChatGPT to annotate, categorise, and interpret them.
A key component of the workshop is critical engagement with the limitations of LLMs, including non-determinism, misclassification, and partial failures in handling complex visual datasets. Exercises emphasize iterative validation, the inclusion of “unsure” categories, and the need to combine automated outputs with manual verification. Overall, the workshop frames generative AI not as a solution but as an experimental, dialogic tool within exploratory visual research practices.
| Period | Feb 2025 → … |
|---|---|
| Event type | Workshop |
| Location | SiegenShow on map |