The GramAdapt Social Contact Dataset is a curated dataset of 34 language pairs from across the globe, with quantitative and qualitative data on social interaction between pairs of language communities in contact. The language pairs were sampled globally, with the intention of avoiding areal and cultural biases. The data were collected by distributing a questionnaire to experts who have experience with either one or both of the language communities. The responses are subjective expert assessments. Authors 1-3 manually checked each response for judgement of responses. This results in a dataset containing 12554 data points.