From Talk Time to Talk Type: Starlight's new Interaction Modes

Walk into almost any teaching and learning meeting and, before long, the conversation turns to teacher talk time. In a growing number of schools it has become a tracked figure, and in some it has quietly become a target. The instinct behind it is a good one. We want classrooms where pupils are thinking, reasoning and contributing, not sitting passively while an adult fills the room with words.

But talk time, taken on its own, is a blunt instrument. It answers a single question: how much did the teacher speak? It says nothing about what that talk was for.

Consider two teachers who each talk for twenty minutes. One delivers a meandering, repetitive explanation that loses the room. The other models a difficult process with real clarity, fields three sharp questions, resets a wobbling class in fifteen seconds, then spends four minutes giving one struggling pupil exactly the help they need. Identical talk time. Entirely different teaching.

That is the trouble with reducing classroom dialogue to one number. The figure is easy to measure and easy to misread, and the professional reality behind it is far richer than the number suggests.

This term we have released three connected improvements to the way Starlight analyses classroom recordings. Together they move the focus on from how much a teacher talks towards a more useful question: what kind of talk was this, and did it match what the learning needed?

Interaction Modes: the kind of talk, not just the amount

The headline change is a new Interaction Modes breakdown. Rather than reporting teacher talk as a single undifferentiated total, Starlight now describes how that talk was distributed across three modes.

Whole-class instruction. The teacher is addressing the whole room: explaining, modelling, questioning or guiding the class as a group.

Small-group interaction. The teacher appears to be working with a smaller group, supporting discussion, checking understanding or guiding collaborative work.

One-to-one support. The teacher appears to be working with an individual pupil, offering targeted help, feedback or clarification.

The value lies in the shift this makes possible. Strong lessons rarely sit in one mode. They move between them: a clear whole-class explanation, a spell of group work with the teacher circulating, a quiet word with a pupil who has lost the thread. Seeing how teacher talk was distributed across those modes turns a flat figure into something a teacher can genuinely reflect on. The question moves from whether you talked too much to whether your talk was the right kind, in the right place, for what the lesson needed.

Honest about what the analysis can and cannot do

We have been deliberate about how this is presented. Interaction Modes are inferred from conversational cues in the recording. They are an aid to reflection, not a measurement instrument, and the report says so plainly.

Telling whole-class talk apart from targeted talk is reliable. Splitting that targeted talk further, into small-group as against one-to-one, is genuinely harder, so Starlight shows a confidence level and labels the breakdown as indicative rather than presenting it as fact. Where a stretch of talk cannot be classified with confidence, Starlight says so openly rather than forcing it into a category. This is insight, not surveillance, and it is offered in that spirit: evidence to think with, never a verdict to be measured against.

Knowing who the teacher is

A breakdown of teacher talk is only as good as its grasp of who the teacher actually is. Until now, like many systems of this kind, Starlight worked on the assumption that the person speaking most was the teacher. That holds most of the time, but real classrooms complicate it. A lively, discussion-heavy lesson, a confident teaching assistant, or an audio provider mislabelling voices can all unsettle the assumption.

Now, in any recording with more than one speaker, Starlight reads the content of the conversation to work out who the teacher really is, and that judgement takes priority over the old most-talking rule. Each report shows a small indicator of whether the original identification was confirmed or corrected, along with a confidence level. Every talk-time figure downstream now rests on a far more reliable foundation.

Reports that follow the dialogue

The third improvement reaches into the written coaching report itself. Rather than analysing a plain block of transcript, Starlight now works from a transcript clearly labelled by speaker and timing. The model can see who said what, when it happened, and how teacher and pupils responded to one another.

The effect is feedback that is better grounded in the real dialogue of the lesson. Starlight can distinguish explanation from questioning, a pupil contribution from a teacher's follow-up, a redirection from a moment of support, and it can draw on the sequence of the lesson rather than treating it as one undifferentiated stream of words.

What this means in practice

For teachers, the result is a coaching report that respects the complexity of what they do. It offers evidence, prompts reflection and supports honest professional conversation, without pretending that teaching can be reduced to a single metric. The report stays private and developmental, owned by the teacher.

For leaders, the same intelligence feeds anonymised, aggregated trends across departments and the school, the kind of picture that helps shape CPD priorities, while every individual teacher's report remains private. Coaching, not compliance. Growth, not grading.

These features came directly from a partner school. One of their teachers asked whether Starlight could separate whole-class talk from targeted support, rather than lumping it all together. That question made the product better, and it is exactly how Starlight should grow: in dialogue with real classrooms, shaped by the questions teachers and leaders are already asking.

Teaching is complex, and the feedback teachers receive should honour that complexity. With Interaction Modes, dialogue-aware reports and more reliable teacher identification, Starlight is becoming a sharper mirror for the reality of classroom practice. Not just how much teachers talk. What that talk is for.

If you would like to see Interaction Modes in a report from one of your own lessons, you can book a personalised demo at starlightmentor.com/demo-request.

Spark Insight with Starlight, and see not just how much you talk, but what your talk is really for.

🎥 Subscribe to our channel here: https://www.youtube.com/@Star21-ai

🌐 Read more on our blog: www.coaching.software

💡 Explore the platform: www.starlightmentor.com

🐦 Follow us on X: @star21starlight

The Insight Engine is written by Adam Sturdee, co-founder of Starlight, the UK’s first AI-powered coaching platform, and a senior leader with responsibility for teaching, learning and coaching. This blog is part of a wider mission to support educators through meaningful reflection, not performance metrics. It documents the journey of building Starlight from the ground up, and explores how AI, when shaped with care, can reduce workload, surface insight, and help teachers think more deeply about their practice. Rooted in the belief that growth should be private, professional, and purposeful, The Insight Engine offers ideas and stories that put insight—not judgment—at the centre of development.

🔗 Connect with me on LinkedIn: https://www.linkedin.com/in/adam-sturdee-b0695b35a/