“I Wanted Them to Think That I Wrote That”: AI-Generated Self-Presentation on Dating Apps and Implications of Non-Disclosure on Informed Consent

Paper Title

“I Wanted Them to Think That I Wrote That”: AI-Generated Self-Presentation on Dating Apps and Implications of Non-Disclosure on Informed Consent

Publication Info

• Topic area: The role of generative AI in self-presentation on dating apps and its implications for informed consent.
• Keywords: Generative AI, online dating, self-presentation, informed consent, deception, AI disclosure, impression management, sexual consent, human-computer interaction, AI-mediated communication.

Background and Problem

• Problem / challenge: Generative AI enables the creation of self-presentation content that diverges from one’s real identity, potentially misinforming consent in intimate interactions. There is limited empirical understanding of how AI-generated content is used and disclosed in online dating.
• Significance: Misinformed consent can lead to ethical and legal issues, including risks of sexual assault by deception. Understanding these dynamics is critical for designing safer online dating platforms and improving consent frameworks.
• Motivation and related work: Prior research has explored consent in computer-mediated interactions and self-presentation strategies in online dating. However, the implications of generative AI on informed consent and its disclosure remain underexplored. This study addresses these gaps by examining how AI-generated content is used and disclosed in dating contexts.

Solution

• Proposed approach: A qualitative survey of 113 online daters from 15 countries to investigate AI-generated self-presentation strategies and disclosure decisions.
• Novelty:
  1. Identification of four distinct strategies for AI-generated self-presentation on dating apps.
  2. Analysis of disclosure and non-disclosure rationales for AI-generated content in dating contexts.
  3. Exploration of the implications of AI use on informed consent in interpersonal interactions.
  4. Discussion of potential risks and design considerations for automated AI disclosure systems.
• Procedure and key techniques:
  • Conducted an open-ended survey with participants who used generative AI for dating app content.
  • Applied reflexive thematic analysis to identify themes in self-presentation strategies and disclosure decisions.
  • Categorized findings into AI-generated self-presentation strategies and rationales for (non-)disclosure.

Results

• Concrete findings:
  • Four self-presentation strategies were identified:
    1. AI as fabricated personality: Creating attractive but unrelated personas.
    2. AI as personalized ideal dating partner: Tailoring messages to specific partners.
    3. AI for enhancement: Improving photos or text for better appeal.
    4. AI for authenticity: Clarifying and accurately expressing one’s identity.
  • Most participants (93%) did not disclose AI use to face-to-face meeting partners.
  • Non-disclosure was primarily motivated by fear of rejection and a belief that AI-use did not significantly impact dating outcomes.
• Advantage over baselines:
  • Provides empirical insights into real-world AI use in online dating, contrasting with prior experimental studies on AI-mediated communication.
  • Highlights the nuanced ways AI-generated content influences impression formation and consent decisions.
• Experiments / evaluation:
  • Participants: 113 online daters from diverse demographics (e.g., 15 countries, 54% men, 46% women).
  • Data collection: Open-ended survey on AI use, motivations, and disclosure decisions.
  • Analysis: Reflexive thematic analysis to identify themes in self-presentation and disclosure strategies.
• Limitations and future work:
  • Limited representation of non-English speakers and marginalized groups.
  • Self-reported data may not fully capture recipient perspectives or verify participants’ accounts.
  • Future research should explore marginalized users’ safety-oriented AI use, large-scale quantitative assessments, and legal implications of AI-mediated consent.

Summary

This study investigates how generative AI is used for self-presentation on dating apps and its implications for informed consent. Four strategies for AI-generated self-presentation were identified, ranging from fabricating attractive personas to enhancing authenticity. Most participants did not disclose their AI use, citing fear of rejection or minimizing its impact on dating outcomes. The findings highlight the potential for AI-generated content to misinform consent, particularly for non-physical traits that are difficult to verify in face-to-face interactions. The study underscores the need for reexamining consent frameworks, advancing AI-mediated communication research, and exploring automated AI-use disclosure systems.