[Dataset] "I Hear, Therefore I Trust" - Trial-level evaluation data from a synthetic speech detection study

This dataset contains trial-level evaluation (markers, subjective evaluations, behavioral log) data from a crowdsourced experiment investigating naive listeners' ability to detect synthetic speech under varying socio-technical trust cues. Participants listened to utterances and annotated regions they believed to be synthetically generated using a custom Streamlit web-based annotation tool. The study manipulated three between-subject trust cues: instruction framing, affective valence priming, and provenance labelling. Audio stimuli were drawn from the LlamaPartialSpoof corpus across three utterance types: authentic, fully synthetic, and partially synthetic speech. 47 participants completed 20 trials each (N = 940 trials). Full documentation of the dataset schema and experimental procedure is provided in DATA_DESCRIPTION.md.