Abstract

As conversational AI becomes increasingly embedded in both personal and professional domains, the imperative to design emotionally intelligent, ethically attuned, and contextually adaptive AI companions grows ever more urgent. This study responds to that imperative through a case-based investigation of Romance at Work—an emotionally fraught, epistemically private, and organizationally sensitive terrain. Leveraging the design science methodology, we developed the RomanceAtWork platform: a contextually fine-tuned, end-to-end encrypted, web-based AI chatbot engineered to support users in navigating the nuanced dynamics of workplace romance. We examine how this system fosters what we call Augmented Romantic Intelligence—a synthesis of emotional awareness, ethical discernment, and situational sensitivity tailored to nuanced interpersonal dynamics at work. For the pilot test, 12 participants engaged in scenario-based role-play, embodying a professional facing a romantic dilemma in the workplace. Each participant interacted with the chatbot across two protocol phases: an unprompted phase followed by a structured phase encouraging contextual elaboration and precise questioning. This adaptive design revealed that users’ articulation of relational context markedly shaped the relevance and quality of the AI’s behavior, including the responses. Findings suggest that the system was generally perceived as both effective and user-friendly. Participants highlighted its perceived neutrality, confidentiality, and potential utility—not only for individual end-users, but also for managerial and HR stakeholders involved in shaping workplace relationship policies. Attributes like psychological safety, nonjudgmental responses, reflective questioning, problem-solution decoupling, and diverse interpretive framings were noted as advantages often missing in human coaches due to bias or dynamics. Intriguingly, the most valued trait in the system was its capacity to critically interrogate users’ assumptions, challenging their thinking with rigor and candor in ways that fostered deeper reflection and more deliberate decision-making. In exploring Augmented Romantic Intelligence, key priorities emerged: enhancing contextual continuity, deepening emotional attunement, and tailoring responses to specific contexts. Rather than seeking algorithmic personalization or normative prescriptions, users preferred guidance that evolved with the dialogue and enabled self-analysis—surfacing underlying motives, exposing cognitive dissonance, and reflecting relational patterns to foster personality insight and ethical reflection. These insights directly informed the development of MORGAN Theory for designing AI coaches, which posits six interrelated design principles essential for cultivating relational intelligence in emotionally complex domains. These include: Mindful Attunement–the AI’s ability to recognize affective nuance, power dynamics, and relational boundaries in real time; Open-ended Responsiveness–facilitating expansive, user-led inquiry while avoiding normative closure; Reflective Mirroring–rearticulating users’ emotional states and intentions with empathic clarity to enhance self-awareness; Grounded Guidance–delivering discreet, evidence-informed advice rooted in policy, relational ethics, and psychology; Authentic Alignment–tailoring support to the user’s cultural, professional, and personal contexts to foster relational trust; and Nurturant Scaffolding–promoting the user’s emotional growth, foresight, and ethical agency without fostering overreliance. By grounding AI companion design in these principles, this pilot study positions such systems as dynamic co-navigators, guiding users through intimate ethical dilemmas with emotional depth and contextual sensitivity. The empirical insights offered here advance emotionally intelligent AI and provide a generative model for deployment in other complex, high-stakes domains.

Comments

tpp1338

Download

Share

COinS