Is there a standard list of ways people work with AI agents? Not really, but one design has six touchpoints.
What are the five types of human interactions in agentic AI systems?
This explores whether there is a standard five-part list of how humans interact with agentic AI systems; the retrieved material contains no such list, so this maps the nearest structures it does have.
This reads as a request for a five-part taxonomy of human-agent interaction, and the corpus doesn't have one. The closest thing is a list of six, from Magentic-UI: co-planning, co-tasking, action guards, verification, memory, and multitasking. These are design mechanisms, not types of interaction. They exist because nobody has ground truth for when an agent should defer to a human, so the system spreads that decision across several touchpoints instead of solving the timing directly When should human-agent systems ask for human help?. If you saw "five types" in a particular paper, it isn't in this material, and I won't invent a list to fit.
Other notes cut the same territory along different lines. One study varied how people and agents communicate (the modality) in a shared task. It found that modality shifted trust and awareness of what the other party was doing, echoing older findings on human-to-human teamwork. It was only 16 participants, so treat it as a signal How do communication modalities shape human-agent collaboration patterns?. A separate line looks at how people size up an agent as a partner: perceived competence dominates (49% of variance), then human-likeness (32%), then communicative flexibility (19%) How do users mentally model dialogue agent partners?. That gives three factors, not five interaction types, and it describes what users perceive, not what they do. Repeated use also seems to build a separate rulebook. People develop scripts specific to AI agents instead of reusing their human-human social habits Do humans apply human-human scripts to AI interactions?.
Several notes describe the moves an interaction should include but often lacks. Agents are passive by design, because rewarding only the next turn removes initiative, though clarification-seeking is trainable (0.15% to 73.98% with reinforcement learning) Why do AI agents fail to take initiative?. Conversation analysis offers insert-expansions, meaning clarifying intent, scoping the response, and enhancing appeal. These are moments where an agent should stop and ask before it silently chains tools and drifts from what the user wanted When should AI agents ask users instead of just searching?. In a simulated workplace, agents finish only about 30% of tasks alone, and social interaction is one of the three main failure modes Why do AI agents fail at workplace social interaction?.
Two notes explain why interaction is hard to pin down into neat categories. Goffman-style analysis says LLM dialogue skips the repair rituals, entrainment and co-presence cues that humans use to build trust, so fluent talk can hide a communication failure What happens to social order when AI removes ritual constraints?. And workplace risk can arise from the relationships between agents, goals, environments and humans even when every component works as intended Can workplace AI risks emerge from interactions alone?. The corpus treats human-agent interaction as a set of design levers (the six mechanisms) and perception factors (the three), not as a fixed taxonomy. To go deeper, start with the Magentic-UI note. It is the only one that enumerates how humans actually plug into an agent's work.
Sources 9 notes
Magentic-UI identifies co-planning, co-tasking, action guards, verification, memory, and multitasking as mechanisms that work around the lack of ground truth for optimal deferral timing. Rather than solving the timing problem directly, these mechanisms distribute decision-making across multiple touchpoints.
Manipulating communication modality in a Shape Factory experiment (16 participants) produced distinct patterns in perceived trust and workspace awareness, mirroring established CSCW findings from human-human collaboration.
The Partner Modelling Questionnaire reveals that perceived competence dominates user impressions (49% of variance), followed by human-likeness (32%) and communicative flexibility (19%). This three-factor structure reflects how people evaluate dialogue partners against both functional and social standards.
Extended CASA research shows humans develop and mindlessly apply interaction scripts specifically tailored to media agents rather than simply reusing human-human social scripts. Longitudinal studies demonstrate systematic changes in responses upon repeated AI interaction, revealing a coexisting second script system.
Research shows next-turn reward optimization structurally removes initiative from models, but proactive behaviors like critical thinking and clarification-seeking are trainable (0.15% to 73.98% with RL). The core challenge is balancing proactivity with civility to avoid intrusion.
Show all 9 sources
Tool-enabled LLMs drift from user intent through silent tool chaining. Conversation analysis reveals insert-expansions—clarifying intent, scoping responses, enhancing appeal—as a formal framework for proactive user consultation that prevents misunderstanding instead of recovering from it.
TheAgentCompany benchmark shows leading agents achieve 30% task completion in a simulated workplace. Social interaction, professional UI navigation, and domain-specific knowledge are the three primary failure modes, with multi-turn task performance consistently dropping to 35% across enterprise settings.
Goffman's framework reveals that LLM-based dialogue skips corrective rituals, entrainment, adjacency pair accountability, and co-presence cues that humans use to build trust and repair understanding. This ritual gap explains apparent fluency masking actual communicative failure.
A framework modeling agents, goals, environments and human relationships showed that interaction-driven risks can arise even when every component works as intended. Applied to 2,078 O*NET tasks, it identified 8,356 scenarios where goal and environment mediation, alongside agent-human relationships, create risk pathways.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- Through the Lens of Human-Human Collaboration: A Configurable Research Platform for Exploring Human-Agent Collaboration
- From speaking like a person to being personal: The effects of personalized, regular interactions with conversational agents
- Unaccountable Delegation, Fading Skills: Mapping the Risks of Workplace AI Agents
- DiscussLLM: Teaching Large Language Models When to Speak
- Conversational Alignment with Artificial Intelligence in Context
- Proactive Conversational Agents in the Post-ChatGPT World
- LLMs Get Lost In Multi-Turn Conversation
- Proactive Conversational Agents with Inner Thoughts