Researchers find AI matches human judgement in dating
Researchers find AI, using large language models, can detect romantic attraction in brief interactions, analyzing data from speed dates.
AI technology seems promising in romantic matchmaking, with researchers finding that large language models (LLMs) like ChatGPT can detect romantic attraction during brief interactions. A study conducted by Columbia Business School, the University of California, and Northwestern University analysed conversations from 964-speed dates to determine if LLMs could predict whether participants would swap contact information.
The research involved 187 undergraduate students, each participating in around 12 four-minute speed dates. ChatGPT was found to perform on par with human judges in predicting the likelihood of contact exchange based solely on text transcripts. Although human judges with access to video footage were the most accurate, ChatGPT could interpret conversational nuances effectively from text alone.
Interestingly, the study revealed that positive emotions weren’t strong predictors of speed dating success. Both human judges and ChatGPT placed too much emphasis on these cues. Instead, references to money, health, and family were found to be stronger indicators of matching, although neither ChatGPT nor human judges considered them significant predictors.
Researchers concluded that LLMs could provide even more accurate judgments by analyzing longer initial interactions and incorporating additional metrics such as smiles, eye contact, and body language. The findings suggest that AI might soon play a key role in predicting romantic compatibility.