The future of large language models in social science research: Reply to Berger (2024) and Carrillo et al. (2024). | Synapse