Supplementary material for the paper: "Exploring LLM-Based Fragment Selection in Video Ethnography of Interactions between Automated Vehicles and Other Road Users"
Wat erin zit
Supplementary material for the paper: Rutger Verstegen, Pavlo Bazilinskyy and Marieke Martens. (2026). Exploring LLM-Based Fragment Selection in Video Ethnography of Interactions between Automated Vehicles and Other Road Users. HCII 2026 - Late Breaking Posters With the increasing application of automated vehicles (AVs), publicly shared ride recordings are also growing, forming a valuable resource for online ethnography research. Qualitative content analysis by humans, however, is time-intensive and not well suited for large datasets. Therefore, this paper proposes the use of a multimodal large language model (LLM) for the selection of interaction fragments between AVs and other road users (ORUs), using the LLM to identify interaction segments for further human coding. Findings from an initial exploration using Gemini 2.5 Pro show that the model identified 4 out of 9 of the interaction fragments selected by human coders, while also struggling with hallucinations. By proposing the usage of multimodal LLMs for interaction fragment selection, and an exploration of this technique, this paper takes a first step toward more scalable digital video analysis of AV–ORU interactions. …
Bestanden en diensten
- DataCite REST API (JSON)api.datacite.orgOpenen