AI Is Emerging as a Thinking Partner in Education

Artificial intelligence (AI) in education is often framed through the lens of risk. But new data from Anthropic suggests a more nuanced reality: students and educators are not simply using AI tools; they are developing measurable skills in how to collaborate with them.

Anthropic’s newly released AI Fluency Index analyzes 9,830 multi-turn conversations on Claude.ai collected during a one-week period in January. Rather than measuring adoption rates, the report tracks 11 observable behaviors that signal “AI fluency,” part of a broader 24-behavior framework developed with academic partners. The goal is to quantify how people interact with AI, not just whether they use it.

The findings show that fluency is less about frequency and more about interaction quality.

The single most important data point in the report is this: 85.7% of conversations included iteration and refinement. Users did not simply ask one question and move on. They revised prompts, clarified instructions, built on prior outputs and adjusted direction.

Iteration strongly correlates with broader fluency. Conversations that included refinement exhibited an average of 2.67 additional fluency behaviors. Non-iterative chats averaged just 1.33 behaviors. In other words, when users treat AI as a collaborative partner, the depth of interaction roughly doubles.

Anthropic also found that only about 30% of users explicitly set expectations about how they want the AI to behave, such as asking it to push back on assumptions or explain reasoning. That gap suggests fluency is still developing. Many users engage actively, but fewer proactively shape the collaboration.

The report distinguishes between basic usage and skillful engagement. Fluency behaviors include clarifying goals, specifying output formats, providing examples, identifying missing context, questioning reasoning and fact-checking outputs. These are signals of human oversight and direction.

This is particularly relevant in educational settings, where the concern has often centered on misuse. The data instead shows that many users are actively managing the interaction, suggesting AI can function as a thinking partner rather than a shortcut.

One of the more revealing findings concerns artifact creation. About 12.3% of conversations involved building artifacts such as code, formatted documents or interactive tools.

When users were creating artifacts, direction-setting behaviors increased significantly. Clarifying goals rose by 14.7%. Specifying output format increased by 14.5 points. Providing examples rose 13.4%, and iteration increased 9.7%.

These patterns suggest that when users are producing something tangible, they give more structured guidance and invest more effort upfront.

However, the report also found a subtle tradeoff. In artifact-building conversations, critical evaluation behaviors declined. Identification of missing context dropped 5.2%. Fact-checking decreased 3.7%. Questioning reasoning declined 3.1%.

In other words, when outputs look polished and complete, users may be slightly less likely to interrogate them critically.

That dynamic is important for educators and policymakers. It highlights not just how AI is being used productively, but where additional instruction around verification and oversight may be needed. The data does not show widespread passivity. It shows structured engagement, with room for growth in critical evaluation.

Anthropic noted in its report that its dataset likely skews toward early adopters comfortable with AI tools. Behaviors are coded in binary form and only capture what is visible in chat logs, not external verification steps. Still, the index establishes a measurable behavioral baseline for early 2026.

The classroom findings do not exist in isolation. PYMNTS Intelligence research shows that AI usage patterns are becoming embedded in broader consumer behavior.

More than 60% of U.S. consumers report using a dedicated AI platform such as ChatGPT, Claude, Gemini or Perplexity in the past year. Increasingly, users begin daily tasks with AI rather than traditional search engines or individual apps.

Separate PYMNTS data shows that consumers are locking in habits early. Many stick with the first chatbot they try, and AI is becoming a routine starting point for research, planning and decision-making.

For all PYMNTS AI and digital transformation coverage, subscribe to the daily AI and Digital Transformation Newsletters.

The post AI Is Emerging as a Thinking Partner in Education appeared first on PYMNTS.com.