AI Flirting Tools On Dating Apps May Weaken Emotional Skills

Sending AI-Written Feelings May Mean Skipping Self-Discovery, Paper Argues

A woman identified only as Avery asked her boyfriend to read a long text about her feelings carefully before replying. Fifteen minutes later, a long message arrived validating everything she had said. It felt great, until she learned he hadn’t written a word of it. ChatGPT had.

Two researchers say that kind of handoff carries a hidden cost. In a paper published in the journal Philosophy & Technology, Ten-Herng Lai of the University of Stirling in the UK and Alfred Archer of Tilburg University in the Netherlands argue that putting feelings into words is one of the ways people learn what they value. Letting AI do that work for apologies, breakup texts, or flirty dating-app messages, they warn, can mean giving up a chance to discover who they are.

Lai and Archer do not claim that every use of AI for emotional tasks does harm. Their concern is people who turn to it multiple times a day.

Outsourcing Emotions to AI Skips the Work That Reveals Personal Values

This was not an experiment. No volunteers were recruited, so there is no sample size or demographic group to report. Lai and Archer built a philosophical argument, supported by earlier research and everyday examples.

That earlier research focuses on thinking rather than feeling. One cited study found that consultants who used the AI model GPT-4 to write first drafts produced weaker reports than those who wrote alone, on a task beyond the AI’s strengths. In another, people who wrote opinion pieces with help from a deliberately biased AI model were more likely to echo its slant, and most never noticed the bias.

Lai and Archer then ask whether the same thing happens with feelings. They lean on the work of Talbot Brewer, who argues that emotions reveal a person’s “pre-reflective evaluations of the world,” meaning gut-level judgments about what matters, formed before any careful thought.

In the paper’s example, Eileen sees herself as not very ambitious. A promotion makes her proud, a reaction that may show her recognition at work matters more than she realized. Habitually skipping the work of making sense of such feelings over a lifetime, the authors write, would lead to an “impoverished understanding of the self.”

Outsourcing Emotions to AI Is Riskiest When Drafts Go Out Unread

To sort out which habits are risky, the paper divides AI users into three simplified types, and a person may fit different types for different tasks.

Complete Outsourcers hand the whole job to AI and send the result without asking whether it reflects how they feel. A breakup text, for instance, might thank a partner for the time together when the sender mostly regrets it. Skipping the reflection also forfeits a chance to learn what felt wrong in the relationship.

First Drafters give AI a rough idea of their feelings, let it write a draft, then edit. That catches emotions the sender doesn’t have, but the authors say it still hands off the harder first step of shaping raw feelings into words. Polished wording can also nudge a person toward feeling exactly what the draft says, rather than what they originally felt.

Reflective Engagers fare best. They use AI as a sounding board, asking for questions that help sort out their feelings or for vocabulary to describe them, then write the message themselves. Even so, the authors flag a smaller risk: AI tools carry biases that users may absorb without noticing.

Dating Apps Make AI Emotional Shortcuts Hard to Resist

Lai and Archer argue that the setting people are in shapes their habits as well as their willpower. Borrowing a term from Avigail Ferdman, they describe AI as creating a “capacity-hostile environment,” a setting that invites people to skip the activities that keep a skill sharp.

Online dating gets the paper’s closest look. Citing reporting in The Guardian, the paper notes that Tinder and other platforms have made AI available to assist “struggling users.” Good dating, the authors argue, depends on reading a match’s emotional cues, which is exactly what AI can bypass. That leaves users, at the very least, deprived of face-to-face emotional skills, the authors say.

Everyday apps carry the same invitation. Prompts like “Ask Gemini,” Google’s AI assistant, appear inside them, and some universities encourage students to use Microsoft Copilot for taking notes.

Greeting cards and love songs are also outsourced emotion, so the authors explain why those seem fine. Someone searching through favorite songs for the one that fits is still interpreting their own feelings. Someone who sends an unread AI apology is not. Asking for several versions and choosing the best match may avoid some of the concerns.

Fixing the problem looks difficult, the paper says. Apps could add friction, such as pausing to ask users for input, but the authors doubt companies will, since users generally prefer smooth experiences and firms may profit from dependence. Regulation has its own problem: users can simply switch to another easy-to-use product.

Avery’s boyfriend, by the paper’s account, treated an apology like a box to tick. A message that sounds right is no replacement for working out what is actually felt, and apps that put AI a few clicks away make it easy to never learn the difference.

---

Disclaimer: This article summarizes a philosophical paper and is intended for general informational purposes only. It does not offer medical or mental health advice, diagnosis, or treatment. Readers with questions about their emotional well-being should consult a qualified professional.

---