Researchers from the University of Stirling and Tilburg University caution against using AI to compose flirty messages in online dating, arguing the practice may prevent people from developing self-knowledge about their own emotions and values.
In their recent research, the authors contend that outsourcing emotionally loaded tasks to AI risks bypassing moments where emotions reveal what people genuinely care about. When someone uses ChatGPT or an integrated AI wingman to craft responses, they may skip the careful attention needed to read what their date actually says, potentially missing deal-breakers or important details.
Philosopher Talbot Brewer has written that articulating emotions helps people understand what they value. The researchers worry that relying on AI output instead of slowly working through feelings could prevent someone from discovering whether they truly care about their date or would happily make compromises in a relationship.
The researchers also highlight practical concerns. Privacy risks exist if flirty exchanges enter AI training data. There is potential for misrepresentation if the AI portrays someone inaccurately to their date. Studies show people tend to overtrust AI outputs and accept them uncritically, and can adopt the biases of large language models even when aware those biases exist.
The authors do not argue people should never outsource emotional tasks to AI. In interactions with manipulative or difficult people, such as difficult relatives or ex-partners, AI may provide useful distance if someone is already certain of their feelings. But for building genuine connections, the researchers suggest resisting the urge to let AI shape what gets communicated.