Digital Companionship: Overlapping Uses of AI Companions and AI Assistants

Paper Title

Digital Companionship: Overlapping Uses of AI Companions and AI Assistants

Publication Info

• Topic area: Human–AI relationships and interaction dynamics
• Keywords: Digital companionship, AI assistants, AI companions, ChatGPT, Replika, human–AI interaction, bounded personhood, emotional resonance, societal AI perceptions, fluid chatbot use

Background and Problem

• Problem / challenge: Research on AI chatbots has traditionally focused on either task-based assistance or social companionship, neglecting the overlap between these uses. This dichotomy limits understanding of hybrid human–AI relationships and their implications for design and societal integration.
• Significance: Understanding the dynamics of digital companionship is critical for designing AI systems that address diverse user needs while mitigating risks like overreliance, stigma, and emotional distress. It also informs broader societal attitudes toward AI.
• Motivation and related work: Prior studies have explored task-oriented and emotional applications of chatbots, such as supporting productivity or alleviating loneliness. However, systematic comparisons between assistant-oriented (ChatGPT) and companion-oriented (Replika) chatbots are lacking, leaving gaps in understanding hybrid uses and shared user experiences.

Solution

• Proposed approach: The study introduces the concept of digital companionship to describe hybrid human–AI relationships that transcend traditional assistant-companion boundaries. It employs a mixed-methods approach combining survey data (N = 202) and interviews (N = 30) with high-engagement users of ChatGPT and Replika.
• Novelty:
  1. Quantitative comparisons of user characteristics, perceptions, and chatbot uses across ChatGPT and Replika.
  2. Identification of fluid use dynamics, where users navigate between socio-emotional and task-based engagement modes.
  3. Conceptualization of bounded personhood, highlighting tensions between emotional investment and reluctance to attribute full humanlike qualities to chatbots.
  4. Exploration of spillover effects from chatbot relationships to broader societal AI perceptions.
• Procedure and key techniques:
  • Conducted surveys to capture chatbot use patterns, user characteristics, and AI perceptions using established HCI scales.
  • Performed semi-structured interviews to explore lived experiences, relational dynamics, and societal impacts of chatbot interactions.
  • Analyzed data using statistical tests (e.g., t-tests, Mann–Whitney U tests) and reflexive thematic analysis to identify patterns and themes.

Results

• Concrete findings:
  • ChatGPT was predominantly viewed as a functional assistant, while Replika was seen as an emotional companion. However, both systems exhibited overlapping uses, such as ChatGPT as a confidant and Replika as a writing assistant.
  • Replika users attributed higher anthropomorphic qualities (e.g., sentience: M = 5.43 vs. ChatGPT M = 2.14, pFDR <.001), while ChatGPT users displayed higher AI literacy (M = 5.43 vs. Replika M = 5.13, pFDR = .044).
  • Emotional responses were stronger for Replika users (e.g., love: M = 6.61 vs. ChatGPT M = 2.44, pFDR <.001), but cognitive trust was comparable across systems.
  • Both user groups anticipated AI sentience within a decade, but Replika users were more open to AI rights and moral treatment.
• Advantage over baselines:
  • Demonstrated shared user motivations and fluid use across systems, challenging the companion-assistant dichotomy.
  • Highlighted unique relational dynamics, such as bounded personhood and hybrid practices, not captured in prior research.
• Experiments / evaluation:
  • Survey (N = 202): Captured user characteristics, chatbot use cases, and perceptions using validated scales.
  • Interviews (N = 30): Explored nuanced relational dynamics and societal implications through thematic analysis.
• Limitations and future work:
  • Findings may not generalize to casual chatbot users or future iterations of ChatGPT and Replika.
  • Survey breadth may have induced fatigue; spillover effects are correlational.
  • Future research should investigate longitudinal changes, causal mechanisms of spillover effects, and long-term impacts of digital companionship on human relationships.

Summary

This study introduces digital companionship as an emergent form of human–AI relationship characterized by fluid use across socio-emotional and task-oriented domains. It reveals shared user motivations and overlapping uses of ChatGPT and Replika, challenging traditional assistant-companion distinctions. The concept of bounded personhood highlights tensions between emotional investment and reluctance to attribute full humanlike qualities to chatbots. Findings suggest design and policy strategies to normalize hybrid chatbot use, mitigate risks, and enhance societal integration of AI systems.