For the past two years, CDI has been studying a question with growing relevance for higher education: Can AI help students develop the skills required for constructive dialogue?

As generative AI becomes increasingly embedded in students' academic lives, institutions face an important challenge: understanding where AI can support learning, where it introduces risks, and how those risks can be managed responsibly.

Based on this research, we are pleased to announce Personalized Practice, a new feature within CDI's Perspectives online learning program. Personalized Practice replaces static reflection questions with structured skill-building exercises that provide students with real-time, adaptive feedback grounded in CDI's dialogue frameworks. Rather than simply assessing students' understanding of dialogue concepts, it gives them opportunities to practice applying those concepts and receive guidance on how to improve.

Early findings from the Personalized Practice feature are encouraging. In a sample of 123 students who beta-tested the program this past Spring, 80% rated Personalized Practice 4 or 5 out of 5 on helpfulness, satisfaction, and realism.

Integrating AI into Perspectives was the product of extensive research, testing, and deliberate design. Throughout the process, we sought to understand where AI could meaningfully support student learning and where it could undermine it. Below, we share what we learned and how those findings shaped the design of Personalized Practice.

Why Explore AI for Dialogue Education?

Constructive dialogue is a skill, and like any skill, it develops through practice and feedback.

Yet meaningful opportunities to practice dialogue are often difficult to create. Conversations across differences can carry significant social and reputational stakes, leading many students to avoid taking risks or engaging deeply with disagreement. While coursework can effectively introduce concepts and frameworks, it is far more difficult to create opportunities for students to apply those concepts, receive feedback, and refine their approach over time.

As a result, many students understand the principles of constructive dialogue without having sufficient opportunities to build fluency in using them.

This challenge led us to explore whether AI could help create additional opportunities for structured practice. At the same time, students are increasingly turning to publicly available AI tools to rehearse difficult conversations, test ideas, and seek feedback on how they communicate across differences. Given these trends, we believed it was important to rigorously examine whether AI could support dialogue education—and, just as importantly, where it should not.

What the Research Suggested

Earlier this year, CDI published a white paper examining three emerging roles for AI in dialogue education: AI as a coach, AI as a mediator, and AI as a conversation partner.

The evidence pointed to a clear pattern: the more constrained and pedagogically structured the AI's role, the lower the risk and the stronger the evidence supporting its use.

Three concerns were particularly important in shaping our approach:

• False mastery. AI can create the appearance of learning without genuine skill development. Systems that over-validate or do the thinking for students may increase confidence without increasing competence.

• Failure of skill transfer. AI can facilitate productive interactions without helping students internalize the underlying skills needed to navigate future conversations independently.

• Persuasion. AI conversation partners can be highly persuasive, creating the risk that practice becomes an exercise in influencing beliefs rather than building understanding.

Our review of the evidence led to a clear conclusion: if AI were to play a role in Perspectives, it should function as a coach that supports deliberate practice and feedback—not as a mediator of live conversations or as a simulated debate partner.

How Personalized Practice Mitigates Those Risks

Personalized Practice was designed in direct response to these findings.

To reduce the risk of false mastery, the feature provides targeted, skill-based feedback rather than simple validation. Students learn not only what they did well, but where they can improve and how.

To support genuine skill development, practice is grounded in structured scenarios designed by CDI researchers and facilitators and aligned with specific learning objectives. The focus is not on completing a conversation, but on strengthening the underlying skills that make constructive dialogue possible.

And to avoid the risks associated with AI conversation partners, we chose not to build a system that generates arguments, counterarguments, or persuasive exchanges in real time. Instead, Personalized Practice helps students practice perspective-taking, charitable interpretation, inquiry, and other core dialogue competencies within a carefully designed educational framework.

Because Personalized Practice is embedded within Perspectives, it also operates within the same pedagogical framework, institutional safeguards, and data practices that govern the broader learning experience.

Looking Ahead

Our research led us to a simple conclusion: AI is most promising when it supports practice and feedback within a structured learning experience. Personalized Practice reflects that principle. It applies AI in a narrowly defined educational role designed to help students strengthen the skills that underlie constructive dialogue while remaining grounded in evidence-based pedagogy.

The field is still evolving, and many important questions remain unanswered. We will continue to test, evaluate, and publish what we learn as institutions seek effective ways to prepare students to engage constructively across differences in an increasingly AI-mediated world.

Please get in touch to learn more about Perspectives with Personalized Practice and how it can be implemented on your campus.