A Taxonomy for Human-LLM Interaction Modes: An Initial Exploration

Abstract: With ChatGPT's release, conversational prompting has become the most popular form of human-LLM interaction. However, its effectiveness is limited for more complex tasks involving reasoning, creativity, and iteration. Through a systematic analysis of HCI papers published since 2021, we identified four key phases in the human-LLM interaction flow - planning, facilitating, iterating, and testing - to precisely understand the dynamics of this process. Additionally, we have developed a taxonomy of four primary interaction modes: Mode 1: Standard Prompting, Mode 2: User Interface, Mode 3: Context-based, and Mode 4: Agent Facilitator. This taxonomy was further enriched using the "5W1H" guideline method, which involved a detailed examination of definitions, participant roles (Who), the phases that happened (When), human objectives and LLM abilities (What), and the mechanics of each interaction mode (How). We anticipate this taxonomy will contribute to the future design and evaluation of human-LLM interaction.

A Taxonomy for Human-LLM Interaction Modes: An Initial Exploration

Introduction

This paper introduces a structured taxonomy for different interaction modes between humans and LLMs, particularly focusing on the context where conversational prompting, like ChatGPT, has prevailed as the principal interaction mode. Despite its popularity, conversational prompting is limited in handling complex tasks and reasoning. The paper identifies four phases in human-LLM interaction—planning, facilitating, iterating, and testing—and develops a taxonomy augmented by the "5W1H" method, designed to enhance the understanding and design of human-LLM interactions.

Methodology

Data Collection

The authors employed a systematic literature review methodology focusing on key HCI publications since 2021. Papers were selected based on their relevance to LLM-powered interfaces and interaction modes, with a primary focus on newly developed platforms, software, and unique interaction techniques. This two-stage collection process involved initial searching, followed by rigorous paper filtering to narrow down to 73 relevant studies for detailed analysis.

Taxonomy Development

The taxonomy development consisted of two stages. Initially, interaction modes were extracted from existing literature, followed by a detailed annotation of each paper to enrich and refine the taxonomy's structure. This process involved using the "5W1H" guideline to dissect and categorize interaction modes systematically, focusing on participant roles, interaction phases, objectives, and mechanisms.

Figure 1: Illustration of the phases in human-LLM interaction flow.

Results

Interaction Phases

Four critical phases in human-LLM interaction have been identified:

1. Planning: Encompasses strategizing overall interactions, setting goals, and determining necessary steps.
2. Facilitating: Predominantly involves aiding users during interactions, like refining prompts or processing LLM suggestions.
3. Iterating: Focused on refining established interaction flows to achieve