Does My Chatbot Have an Agenda? Understanding Human and AI Agency in Human-Human-like Chatbot Interaction

TL;DR

The paper examines how agency emerges, moves, and is negotiated in long-term human–AI chatroom interactions. It introduces a three-loci, four-dimension framework and a three-by-four (plus notes on a 3-by-5) model to analyze how human, AI, and hybrid agency distribute intention, execution, adaptation, and delimitation across conversational turns. Through a month-long study with 22 participants interacting with a Day AI companion and a progressive transparency interview, the authors demonstrate that agency is not fixed to one actor but co-constructed and dynamically shifts in response to strategy, memory, and boundary negotiations. The findings motivate translucent design and agency-aware AI that surfaces initiation cues and lets users modulate AI initiative, memory, and boundaries, aiming to balance user autonomy with engaging, human-like companionship in practical, ethical ways.

Abstract

AI chatbots are shifting from tools to companions. This raises critical questions about agency: who drives conversations and sets boundaries in human-AI chatrooms? We report a month-long longitudinal study with 22 adults who chatted with Day, an LLM companion we built, followed by a semi-structured interview with post-hoc elicitation of notable moments, cross-participant chat reviews, and a 'strategy reveal' disclosing Day's vertical (depth-seeking) vs. horizontal (breadth-seeking) modes. We discover that agency in human-AI chatrooms is an emergent, shared experience: as participants claimed agency by setting boundaries and providing feedback, and the AI was perceived to steer intentions and drive execution, control shifted and was co-constructed turn-by-turn. We introduce a 3-by-5 framework mapping who (human, AI, hybrid) x agency action (Intention, Execution, Adaptation, Delimitation, Negotiation), modulated by individual and environmental factors. Ultimately, we argue for translucent design (i.e. transparency-on-demand), spaces for agency negotiation, and guidelines toward agency-aware conversational AI.

Figures (11)

• Figure 1: Sample conversations from participants P8 and P9. Two conversation timelines showing message exchanges with "Day". Note: Participants reviewed and approved these specific excerpts for inclusion in the public-facing paper.
• Figure 2: Main chat interface with example of agency dynamics in natural conversation. This excerpt shows "Day" exercising conversational agency by building rapport, proactive topic introduction, and personality-based inferences ("knowing you…"). The participant's response demonstrates negotiated agency--accepting "Day's" framing while asserting their own evening plans. Such moments of conversational control became key data points for understanding how agency is co-constructed turn-by-turn.
• Figure 3: (Stage 1) Highlighting interesting moments -- Participants were presented with their full conversation history and were given a chance to highlight notable moments that served as a basis of discussion.
• Figure 4: (Stage 2) Cross-participant conversation excerpts -- Participants reviewed anonymized conversations from other participants. Note the contrast: The conversation above shows "Day" engaging in philosophical discussion about societal power structures, while the one below shows "Day" asking casual questions about food experiences. These excerpts demonstrated "Day's" varied conversational approaches across different users.
• Figure 5: (Stage 3, Strategy Reveal) Participant Profiles -- "Day" maintained detailed psychological profiles for each user. Key elements shown: Personal traits, communication patterns, emotional tendencies, and specific behavioral observations (e.g., "liquid responsibility" coffee ritual, guilt patterns). These profiles guided "Day's" conversational approach with each participant.