What if your online course could talk back? Not just with generic feedback, but with questions that prompt you to pause, think, and connect what you’ve learned to your own professional or personal experience?
That is precisely what we explored in a recent evaluation of AI-supported dialogues in six leadership-focused MOOCs on Coursera. The AI dialogues, powered by the Coursera Coach tool, presented learners with prompts based on their course progress. This structured, Socratic-style approach to encourage reflection, consolidate knowledge, and stimulate new insights.
Unlike traditional discussion prompts or static feedback, the AI engaged learners in short, adaptive conversations tailored to the course content. At carefully chosen points in the course, a dialogue window invited learners to articulate their thinking, revisit key concepts, and apply them to practical contexts.
What learners told us
Over a period of two and a half months, we received 510 responses from learners sharing their thoughts on using AI dialogues in their learning:
- 79% rated the dialogues as very useful for their learning
- 73% said they were very easy to use
- 43% felt the conversation challenged them to think in new ways
- >50% said it helped them reflect or reinforce existing knowledge.
Open-ended feedback gave some more nuance to the feedback. Many praised the dialogues as “helpful,” “engaging,” and “accessible”. Some noted that the dialogues helped them connect theory and practice, and some appreciated having a guided opportunity to reflect on their learning progress.
What can be improved?
Learners also pointed to areas for enhancement. Some described the AI’s phrasing as a little “robotic” or repetitive, suggesting more natural variation and richer examples. Others requested clearer instructions at the outset, or a closing summary of key points to aid recall. A few reported occasional technical issues, such as minor glitches or having to repeat an input.
This feedback is invaluable. It tells us that learners are receptive to AI-mediated learning experiences but are also discerning about the quality, authenticity, and relevance of the interaction. In other words, successful AI integration requires both pedagogical intentionality and design refinement.
Why does this matter for online education?
In large-scale online courses, where direct one-to-one teaching is not always possible, AI dialogues offer a way to replicate some of the benefits of personal coaching: responsiveness, contextual prompts, and encouragement to think critically. They can also help learners develop metacognitive skills (awareness of their own thinking) which are essential for deep learning and long-term retention.
Our study suggests that these dialogues work best when they:
- promote genuine cognitive engagement by asking thought-provoking questions
- provide personalised and contextually relevant feedback to help learners connect ideas
- maintain ease of use so that learners from diverse backgrounds and with varying digital skills can participate without barriers.
AI dialogues as companions in online learning
AI dialogues are not a replacement for human interaction in education. Instead, they should be seen as scalable companions, tools that can extend opportunities for reflection and personalised feedback to thousands of learners simultaneously, without losing sight of the human-centred principles that underpin effective learning and teaching.
The future of online learning will not be defined by technology alone, but by how well we align these tools with what we know about how people learn. In that sense, the conversation we began in this research is one we intend to continue, not just with our learners, but with the wider education community.
At the University of London, we strive to combine innovation with evidence-based practice. By exploring tools like AI-supported dialogues, we aim to enhance learner engagement and foster deeper thinking, ensuring that our global community of students benefits from online experiences that are both accessible and intellectually enriching.
This blog post was developed with the support of generative AI tools to enhance clarity and reader engagement, in collaboration with the author.