An automated, data-driven approach to uncover patterns, which the authors term traits, of effective human-AI interaction that are aligned with task outcomes is explored and Principal Trait Analysis is proposed, a Principal Component Analysis-inspired algorithm for deriving common traits from patterns in LLM conversations.
Abstract
Large Language Model-powered agents are increasingly used in the workplace via human-artificial intelligence (AI) collaboration. In this new era of work, it is important to understand the kinds of prompting traits that contribute to task success. Moreover, we need to uncover key skills required for modern professionals and inform educators on how to foster these skills among students. Existing guidelines for human-AI collaboration are built from either top-down theory or context-specific observations of human-AI interactions. However, since LLM capabilities are rapidly improving, theory may not be able to explain emerging interaction patterns, and empirical guidelines may become obsolete quickly. In this work, we explore an automated, data-driven approach to uncover patterns, which we term traits, of effective human-AI interaction that are aligned with task outcomes. We propose Principal Trait Analysis, a Principal Component Analysis-inspired algorithm for deriving common traits from patterns in LLM conversations. Our algorithm uses LLM-based processing stages to analyze corpora of human-AI collaborative session traces, deriving common traits across the dataset and scoring each human collaborator's usage style by each trait. The approach also allows domain expertise to be injected during trait discovery and selects the most distinguishing traits to be those that exhibit the highest variance across collaborators. We evaluate PTA on two human-AI collaborative coding datasets, an educational setting (students working with an AI tutor) and a professional setting (developers working with an AI coding agent). We find that PTA-derived traits are significant in explaining collaborator behavior across both settings and can help predict task outcomes. However, whether traits qualify as skills remains to be seen, due to inconclusive results on generalizability and how user traits change over time.
As generative AI transforms work across all fields, understanding human–AI collaboration is critically important to the future of work. This study analyzed behavioral data from millions of Claude AI conversations mapped to occupational tasks via the O*NET taxonomy to examine whether human–AI collaboration patterns are consistent with expectations derived from traditional technology adoption frameworks. Utilizing Technology Acceptance Model, Protection Motivation Theory, Social Exchange Theory, and Socio-Technical Systems Theory as parallel lenses on post-adoptive behavior, the analysis examined how interaction patterns predict AI usage intensity, risk-management behaviors, and the development of cognitive partnerships. Results demonstrated that collaborative interaction patterns, including iterative refinement and learning-oriented exchanges, predicted significantly higher AI usage intensity, consistent with expectations derived from TAM. Contrary to Protection Motivation Theory predictions, risk-management behaviors such as validation and feedback loops were positively associated with usage intensity, suggesting these function as protective engagement strategies that enable continued use, not resistance mechanisms. Analysis of extended thinking mode usage suggested distinct trust pathways: delegation trust, where users rely on AI’s autonomous reasoning, versus collaboration trust, built through iterative dialogue, although sensitivity analyses indicate that task complexity may partly account for this divergence. A cross-level reversal emerged whereby augmentation patterns positively predicted usage intensity at the task level but negatively predicted user adoption at the cluster level, indicating that depth and breadth of AI use follow different dynamics. These findings challenge assumptions that caution suppresses AI use and suggest that effective human–AI partnership supports cognitive partnerships through multiple complementary pathways. For practitioners and leaders, supporting diverse ways of interacting with AI can be important for successful integration. The analyses drew on the Anthropic Economic Index dataset of 3,365 occupational tasks grouped into 593 task clusters, with model sample sizes varying by hypothesis.
Patrick D. Lynch· Frontiers in Artificial Inte...· 0 citations
Artificial intelligence (AI) is increasingly embedded in scientific work, but researchers may not evaluate its use uniformly across research tasks. This study examines task-specific attitudes towards AI among an international, self-selected sample of 3,785 PhD students in STEM and medical and health sciences who participated in Nature's Graduate Survey 2025. We analyse respondents'comfort with using AI for writing a research article, collecting and analysing data, designing experiments, tracking scientific literature, and summarising it. Latent class analysis identifies four distinct attitudinal profiles. The dominant profile reflects a"division of labour,"in which AI is widely accepted for literature-related tasks but resisted in activities closely associated with intellectual contribution, such as writing, data analysis, and experimental design. A"status quo"profile is broadly uncomfortable across tasks, an"all-purpose"profile is broadly comfortable, and an"undecided"profile expresses substantial uncertainty. These patterns suggest that attitudes towards AI in research are organised less around a simple acceptance-rejection divide than around task-specific boundaries, likely concerning delegation, authorship, and responsibility. Because the survey measures comfort rather than legitimacy, the profiles are best interpreted as attitudinal configurations with a normative dimension. The findings highlight the importance of task-specific approaches to AI governance, doctoral training, disclosure, and research evaluation.
We review recent research on the trio of artificial intelligence (AI) practices—electronic platforms, machine learning, and large language models—and their effects on those who use and work with them. Because these tools take over specific tasks from existing jobs, the important effects are how the remaining work changes and how workers react to it. Perhaps our main conclusion is that choices as to how they are used are the most important factor in their effects on those who use them. For example, do they take over boring tasks like note-taking or interesting ones like drafting reports? Do workers spend their saved time checking AI output or taking on new tasks? We describe the wide variety of outcomes and emerging issues such as the ability of chatbots to show empathy, the ability to codify what had been tacit human knowledge, and the development of digital twins, which predict responses from specific individuals.
P. Cappelli, Prasanna B. Tambe, Brook Jiang· Annual Review of Organizatio...· 0 citations
This article addresses the growing concern that generative artificial intelligence (AI) may replace human expertise in organizations. Instead of asking whether AI should be used, it examines why human judgment rooted in experience cannot be fully replaced by current AI systems and how organizations can work with AI more effectively. Drawing on research from cognitive science, neuroscience, and organizational studies, the paper explains how people use prior experience to interpret context, notice subtle cues, and make sense of ambiguous situations—capabilities that differ fundamentally from how large language models process data. Evidence from recent studies of AI use in hiring, performance management, healthcare, and knowledge work shows recurring problems, including mistakes in unusual cases, missed context, over-reliance on AI recommendations, and reduced visibility of real skill differences among employees. In response, we propose a five-part Human–AI Collaboration Framework designed to help organizations use AI for efficiency while keeping human judgment active and accountable in key Human Resource Management decisions. The analysis shows that AI performs best in routine, data-rich situations but falls short when decisions require lived experience and contextual understanding. By framing organizations as systems built on accumulated experience, this article offers practical guidance for responsible AI integration and outlines directions for future research on human–AI collaboration.
Daniel Altieri, Zohra Damani, Cynthia L. Nebel· Administrative Sciences· 0 citations
While artificial intelligence (AI) is increasingly being adopted to mitigate social isolation, existing research has predominantly examined human–AI (H–AI) companionship, potentially overlooking the risk that AI agents are being used as a substitute for authentic human-to-human (H–H) social connections. This scoping review investigates the relatively underexplored domain of AI as a catalytic mediator designed to promote H–H social connections. Guided by PRISMA (Preferred Reporting Items for Systematic Reviews and Meta‑Analyses) and study selection process, a total of 225 studies published between January 2021 and October 2025 were initially retrieved, of which five were ultimately selected for analysis. By examining empirical studies on AI-mediated H–H social connections, this study explores the current state of research, identifies design implications, and then develops preliminary design guidelines for enhancing the quality of H–H social connections. The findings reveal that the reviewed studies frequently exhibit structural limitations, such as a reliance on dyadic relationships, insufficient reciprocal support mechanisms for vulnerable populations, and a lack of exit strategies. Ultimately, this review argues for expanding the social role of AI from that of a companion to that of a catalytic mediator so that AI can support more diverse H–H social relationships.
S. Chung· The International Journal of...· 0 citations
Large language models (LLMs) are popular tools for creative ideation, but have been shown to homogenize outputs across users. We test if approaching an AI tool with high human (vs. low) human agency can mitigate this homogenization effect by encouraging people to use AI to augment their creativity, rather than offload it. Participants were experimentally assigned to one of three conditions (high-agency approach with AI access, low-agency approach with AI access, or a human-only control) and generated creative uses for everyday objects. Contrary to our expectations, ideas did not differ in individual-level quality (overall creativity, originality, and usefulness). However, a preregistered similarity-to-centroid analysis and an exploratory cluster analysis provided convergent evidence of AI-induced homogenization among the low-agency condition. Thus, while AI has enabled greater ideational fluency, our research suggests the degree of agency with which people approach their AI tool has downstream consequences on collective creativity.
Sarah H. Wu, Yuewen Yang, A. Y. Lee et al.· Creativity & Cognition· 0 citations