AI Utilization for Grading by Educators: Insights from Anthropic

Essential Insights from ZDNET

The discourse around artificial intelligence in academic settings often centers on the impact these tools may have on students, with particular worry regarding potential academic dishonesty and quality degradation. Contrarily, Anthropic's recent study illuminates the varying ways in which teachers are incorporating AI into their classrooms, highlighting some unexpected methods of tech integration.

AI developers recognize the tension educators face between utilizing AI for support versus allowing it to take over certain tasks. Anthropic’s research reveals the methods educators employ to balance these approaches, showing particular differences depending on individual cases and preferences.

This research involved Anthropic examining anonymized interactions between their chatbot, Claude.ai, and both free and premium accounts linked with higher education email domains. The focus was on tasks specific to education from May to June 2025, where 74,000 conversations were identified related to educational functions such as syllabus creation and assignment grading.

Anthropic drew connections between these conversations and relevant tasks cataloged in the United States Department of Labor's O*NET database. Additionally, they augmented this analysis with survey feedback and qualitative insights from 22 faculty members from Northeastern University notable for their early AI adoption.

Grading Gets a Boost from Automation

The study revealed that AI is predominantly employed by educators for tasks like curriculum design (57%) and academic investigations (13%). However, a less prevalent, yet significant, use is automated student performance assessment by 7% of educators—where AI tools were applied to provide feedback, grade assignments against benchmarks, and summarize evaluations—despite general teacher hesitations toward AI in grading.

Automation played a substantial role in these grading tasks, being employed in nearly 49% of cases. This high reliance contrasts with educators' reservations about automating evaluations, which some faculty surveys have reflected as the area they find AI least effective.

Apart from grading, AI is leveraged to enhance tasks that involve instruction, planning of academic proposals, student guidance, and supervising research endeavors. Other frequently automated duties comprise financial management, student record maintenance, and the administration of academic admissions—tasks inherently administrative in nature.

This trend demonstrates educators' openness to using automation for time-consuming tasks, whereas they prefer AI as a collaborative tool for more nuanced work.

Yet there is heightened concern from Anthropic regarding the significant number of educators allowing AI to automate grading, suggesting a tension and skepticism over the appropriateness of AI for such critical assessments.

A Northeastern faculty member expressed doubts about ethical implications and the precision of AI grading, stating: 'Having tried AI-driven grading, the results didn't meet my expectations. Students deserve the expertise I offer, enhanced but not replaced by AI.'

Despite being a less frequent application, automated grading still remains highly utilized.

Anthropic indicates an unclear extent to which AI-derived responses affect final student evaluations, yet evidence suggests some level of delegation to Claude for grading.

Exploring Additional AI Applications in Education

Other distinct AI uses included crafting hypothetical legal cases for classroom exercises, developing workforce training modules, creating recommendation letters, and organizing meeting agendas.

Northeastern faculty also noted AI's role as a learning enhancer, although verifying these claims through Claude.ai analysis posed challenges. Educators harness AI as both a productivity enhancer for repetitive tasks and as a thought collaborator to tailor student learning.

Beyond existing classroom tools, teachers are innovating by developing custom AI-driven educational resources. Anthropic reported educators using its Artifacts tool—an app-creation function without the need for coding—to design interactive educational games, grading tools, data visuals, and more.

Navigating AI’s Growing Role in Schools

As schools resume, AI companies are stepping up introductions of tools aimed at educators. Anthropic has unveiled tools within its Claude.ai chatbot to facilitate Socratic teaching styles instead of direct response outputs. In the larger context, vendors like Google are launching AI tool packages that might become essentials for teachers and students alike.

Debates on AI’s educational value aside, capturing the educational sector’s interest and instigating reliance on technology for academic planning and beyond could prove financially rewarding for AI firms. With educators facing significant stress and high attrition rates, it’s unsurprising there’s a shift towards AI-mediated task automation. Whether AI can evolve from being a classroom utility to something indispensable remains to be seen. As things stand, even with policies potentially delineating AI use in schools, educator and institutional choices will deeply influence AI’s future in education since access remains unrestricted.