The study introduces a 5-stage mapping procedure to systematically analyze and categorize 110 relevant publications on human-large language model (LLM) interaction. The mapping is guided by two key perspectives: collaboration and creativity.
The collaboration dimension examines the decision-making relationship between human and AI, ranging from human-led to AI-led tasks. The creativity dimension evaluates the type of tasks handled by AI, from simple data processing to autonomous content generation.
Through this mapping, the study identifies four main clusters of human-LLM interaction patterns:
Processing Tool: LLM perform specific, directed tasks with limited creative input, primarily serving as tools for human decision-making.
Analysis Assistant: LLM provide analytical support and opinion-forming capabilities to aid human decision-making, acting as assistants in the collaborative process.
Creative Companion: LLM exhibit a high degree of autonomy and creativity, collaborating with human in open-ended, generative tasks.
Processing Agent: LLM handle complex tasks with a level of autonomy, but their creative contribution is limited to data organization and summarization.
The study also discusses the differences between these clusters, highlighting the nuances in the collaborative relationship and the creative responsibilities of LLM. Additionally, it identifies a vacant space in the mapping, suggesting opportunities for future research on human-LLM interaction patterns that involve continuous mutual learning and joint decision-making.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Jiayang Li,J... at arxiv.org 04-09-2024
https://arxiv.org/pdf/2404.04570.pdfDeeper Inquiries