Now loading...
Anthropic, a leading developer of artificial intelligence systems, has unveiled a new tool called Anthropic Interviewer, designed to gather insights into how people perceive and interact with AI in their daily work. Powered by the company’s Claude AI model, the tool conducts automated interviews at a large scale, allowing researchers to analyze responses from hundreds or thousands of participants quickly and efficiently. In an initial trial, the tool interviewed 1,250 professionals across various fields, revealing a mix of enthusiasm for AI’s productivity benefits and lingering worries about its broader effects on jobs and creativity.
The interviews, which lasted about 10 to 15 minutes each, focused on how workers incorporate AI into their routines and what they expect from the technology moving forward. Participants included 1,000 from a broad range of occupations, such as educators, IT specialists, and media professionals, along with 125 creatives like writers and artists, and 125 scientists from disciplines including physics, chemistry, and data science. Anthropic recruited them through online platforms, ensuring each had a primary job outside of the recruitment process. All agreed to have their anonymized transcripts released for public research, available on Hugging Face.
Overall, respondents expressed positive views on AI’s role in their professions, with the majority highlighting time savings and improved efficiency. In the general workforce group, 86 percent said AI helped them work faster, and 65 percent were content with its current contributions. However, optimism was tempered by concerns in areas like education, where some feared over-reliance could hinder learning, and job security, particularly for artists facing potential displacement. Professionals often described a desire to keep core, identity-defining tasks human-led while handing off repetitive duties to AI, envisioning roles that involve supervising automated systems.
One office assistant compared AI to past innovations like computers, noting it enhances capabilities without eliminating jobs, much like typewriters did for mathematicians. Yet a trucking dispatcher voiced uncertainty about adapting skills that AI cannot replicate, focusing on human elements like personal interactions. Social pressures also emerged, with 69 percent mentioning stigma from colleagues wary of AI use, leading some to keep their methods private.
Creative professionals showed similar gains, with 97 percent reporting time savings and 68 percent noting better output quality. A novelist credited AI for making research less intimidating, boosting writing speed, while a content writer tripled daily production. Despite these advantages, 70 percent grappled with judgment from peers, and economic fears loomed large. A voice actor lamented the decline of certain sectors due to AI, and a composer worried about market saturation from machine-generated music. Many stressed maintaining control over their work, though some admitted AI occasionally steered creative choices, like in concept generation for art or lyrics in music.
Scientists, meanwhile, expressed a strong interest in AI as a collaborative partner but limited trust in its reliability for essential tasks like forming hypotheses or designing experiments. Instead, they relied on it for supportive roles, such as drafting papers or fixing code. Trust issues arose in 79 percent of conversations, often tied to errors or inconsistencies, with one researcher calling verification needs counterproductive. A medical scientist highlighted data privacy worries in commercial settings, while an economist pointed to AI’s tendency to fabricate information. Despite frustrations, 91 percent wanted expanded AI help, ideally for integrating vast datasets or sparking novel ideas. Job loss fears were minimal, as many cited irreplaceable human intuition and practical constraints like lab limitations.
The Anthropic Interviewer tool works in three phases: planning questions based on research objectives, conducting adaptive conversations on the Claude.ai platform, and analyzing transcripts with human oversight and AI clustering for themes. This approach allowed the company to scale what would otherwise require extensive manual effort. Early results suggest professionals view AI use as more collaborative in self-reports than observed chat patterns indicate, possibly due to post-conversation edits or varying tool preferences.
Looking ahead, Anthropic plans to expand the tool’s use through partnerships with cultural institutions, scientific grantees, and educators, such as collaborations with the American Federation of Teachers to train 400,000 educators on AI. The company is also inviting Claude.ai users to join a public pilot interview via a pop-up on the platform, aiming to explore personal visions for AI’s future role. Findings will inform model improvements and policy discussions, building on efforts to align AI development with user input.
While the study offers valuable snapshots, Anthropic acknowledges limitations, including potential biases from online recruitment, self-reported data discrepancies, and a focus on Western perspectives. The initiative underscores a push to humanize AI advancement by directly capturing user experiences beyond digital interactions.
