As large language models become embedded in everyday products, AI companions have emerged as a new category focused on emotional and social interaction rather than task completion. These systems are designed to build connection through conversation, empathy, and presence, which introduces unique design challenges. AI companions exist in a core tension. When they feel too human, they can blur boundaries and create ethical risks. When they feel too machine like, they can reduce trust, engagement, and emotional value. This tension sits at the center of my research.
Rethinking Ethics and UX as Separate Domains
Our research found that traditional separation between ethics and user experience no longer applies in emotional AI systems. In AI companions, ethical decisions directly shape user experience, and UX decisions can create ethical consequences. For example, transparency can protect users but disrupt immersion, safety can prevent harm but limit natural interaction, and empathy can support users but risk dependency.
Scrappy iterative coding
Streamlining Principles
Our Research Process
-
Conducted a literature review across HCI, human AI interaction, conversational agents, and AI companions, identifying 548 design recommendations from 41 sources
-
Translated findings into the AI companion context through collaborative co analysis with 11 HCI practitioners
-
Ran a 10 day collaborative autoethnography where researchers directly interacted with AI companions to study early relationship formation and emotional response
-
Synthesized insights into draft design principles and evaluated them using internal heuristic evaluation across real AI products
-
Validated principles through a modified heuristic evaluation with 14 domain experts across multiple AI and companion systems
Divya's Heuristic Evaluation
Industry Expert Heuristic Evaluation Guide
AI companion design requires a shift from fixed design rules to flexible, human centered frameworks that balance emotional value, user trust, and ethical responsibility.
The research identified nine core design principle areas across safety, transparency, inclusivity, control, personalization, engagement, empathy, and response behavior.
Ethics and user experience are deeply interconnected in AI companions, and design principles must be applied flexibly based on user context and emotional needs.