The Nonverbal Gap: Toward Affective Computer Vision for Safer and More Equitable Online Dating
Online dating has increasingly become the primary means through which individuals initiate romantic relationships. However, the current platforms often strip away crucial nonverbal cues—such as gaze, facial expressions, body posture, and response timing—that humans typically rely on to gauge comfort, disinterest, and consent. This absence creates a significant communication gap, particularly affecting women’s safety in online dating environments.
This article highlights the nonverbal gap as both a technical opportunity and a moral responsibility for the computer vision (CV) community. Despite the existence of advanced affective tools—such as facial action unit detection, gaze estimation, engagement modeling, and multimodal affect recognition—this vital domain has not been sufficiently explored by researchers focused on dating contexts.
Proposed Research Agenda
To address the nonverbal communication gap in online dating, we propose a fairness-first research agenda organized around four critical capability areas:
- Real-time Discomfort Detection: Developing systems that can accurately interpret and respond to signs of discomfort in real-time during online interactions.
- Engagement Asymmetry Modeling: Understanding the differences in engagement levels between partners and how that asymmetry affects communication dynamics.
- Consent-aware Interaction Design: Creating frameworks that prioritize consent in interactions, ensuring that both parties are aware and agree to the nature of the communication.
- Longitudinal Interaction Summarization: Implementing tools that provide summaries of interactions over time, giving users insights into their communication patterns and potential areas for improvement.
Each area is grounded in established CV methodologies and motivated by principles from social psychology relevant to romantic communication. The execution of this agenda necessitates a commitment to ethical standards, particularly in data collection and processing.
Ethical Considerations
Responsible advancement in this field requires the creation of purpose-built datasets that are collected under strict dyadic consent protocols. This means that data must be gathered with the explicit agreement of all parties involved, ensuring transparency and mutual understanding.
Moreover, fairness evaluation must be disaggregated across various demographics, including race, gender identity, neurotype, and cultural background. This approach would help ensure that the developed systems are equitable and do not perpetuate existing biases.
Another critical aspect is the commitment to on-device processing. This architectural choice aims to prevent affective data from becoming a tool for platform surveillance, thus prioritizing user privacy and safety.
Call to Action
This vision paper serves as a call to the computer vision community, particularly the Workshop on Intersectional Computer Vision (WICV) community. Its members are uniquely positioned to recognize the technical opportunities and human stakes involved in this emerging domain. It is essential to establish online dating safety as a primary research focus before the pace of commercial deployment outstrips ethical deliberation.
By addressing the nonverbal gap in online dating, the CV community can contribute significantly to creating safer and more equitable digital romantic spaces.