Let me start with a confession. I’ve never been a big fan of group work. I never liked it much as a student, and I don’t particularly enjoy organising it as a teacher either. But I do know that it can be a powerful way to learn. I know the research. Evidence-informed doesn’t mean ignoring evidence you don’t like. But the problem is that things don’t always go as planned with this kind of collaborative learning. I’m sure you will recognise this. While some students take charge of the task at hand, others sit back because they can’t keep up, can’t get a word in, or, worse, are simply freeloaders. I’ve seen groups have plenty of discussion while very little learning happens. And sometimes a group produces an excellent result, but I remain unsure how much each student has actually learned.

Okay, I’m not exactly selling group work here. But what can teachers do to make it work better?

A new meta-analysis by Gao and colleagues, published in Review of Educational Research, offers some interesting starting points. The authors collected 83 studies with nearly 7,000 participants. They did not investigate whether group work is better than classroom instruction or individual learning. Instead, they examined the added value of adaptive support for collaborative learning, which means you don’t give all students the same support at every moment. You tailor your help to what they actually need.

The meta-analysis

The researchers found a fairly large average positive effect of adaptive support on learning during group work (g = 0.73). Cognitive learning outcomes, in particular, improved (g = 0.93). Group performance and collaboration quality also improved.

Nice, but there are some nuances. First, differences between studies were particularly notable. Besides, the effect size in randomised experiments was significantly smaller (g = 0.44) than in experiments conducted using any other design. I think this is something one could expect. The most important thing to bear in mind is that the researchers observed positive results from such guidance in collaborative learning primarily when it was done online or in a hybrid environment. When it was done face-to-face, the effect was much lower and not statistically significant. However, this doesn’t mean that such guidance in traditional classes is unnecessary.

Takeaways for the classroom

1. First look at where a group is getting stuck before you help.

Support that accounts for cognitive factors (prior knowledge and performance) was associated with better outcomes than support that focuses mainly on social or affective factors. When letting your students work in groups, pay attention to both how actively everyone participates and what each individual student understands.

Let me make this more concrete. Four students have to solve a problem together. There is an active discussion where everyone participates. The atmosphere is great. But… all the group members start with the same misconception. Additional encouragement to collaborate will hardly help in this case. A targeted question or an explanation can be more useful here. But a group that understands the material well but can’t collaborate because one student makes all the decisions needs a different approach.

This all may sound obvious, but it means that the question is not only whether you should intervene, but above all why.

2. Help not only individual students, but also the collaboration itself

Focus also on what students do together. For example, help a group build on each other’s ideas. Some example questions you could ask instead of immediately providing the answer yourself:

• Why do you think this is correct?
• Can anyone build on what was just said?
• Do you agree with the reasoning, or only with the answer?
• Can everyone explain how you arrived at this solution?

These are examples of questions you could use as a teacher, not individual questions proven effective through this meta-analysis. You don’t necessarily have to take over the group task. You’re trying to improve the quality of thinking and cooperation.

3. Do not automatically provide more help, but the right help

As always in education, support with group work does not need to be one-size-fits-all. If one of your groups doesn’t seem to understand the assignment well, they may need a detailed example or extra explanation. While another group, making good progress, may only need a question. You may be better off leaving a group that works well alone. Adaptive support does not mean that you have to keep intervening. This partly answers the question some teachers may already have been asking halfway through this post: How will I be able to do all that?

What steps do you need to take as a teacher? To begin with, make sure you have different types of support ready beforehand. This may take the form of a content-related hint, an intermediate step, an example, or even a question that pushes students to explain their reasoning.

I know all of the above can be challenging and require effort, especially when several groups need assistance at the same time. Unfortunately, the meta-analysis does not tell us how a teacher can organise all this efficiently in a full classroom. And that is perhaps the most important question, I think.

4. Don’t forget the students with a lot of prior knowledge

The positive effects were not limited to students with little prior knowledge. In fact, adaptive support was slightly more effective in studies involving students who already knew more about the topic. This doesn’t mean that students with less prior knowledge need less help.

Let me explain this with a final example. When a group has mastered the basics, you can give them a more challenging question, an assignment that requires them to compare various solutions, or a task to find counterarguments to their own solution. Students with more prior knowledge may also benefit from such support, although we shouldn’t draw firm conclusions from comparisons between studies about who benefits most.

And does technology need to be involved?

Many of the interventions used technology to determine when and how students would be supported. These included systems that tracked performance, analysed collaboration, or provided automatic feedback. This is relevant because AI can now support learners in many ways. But you cannot use this meta-analysis as evidence that you need an AI system for good group work. The literature search ended in June 2023, meaning that the latest applications of generative AI are barely, if at all, covered.

I think the real issue is more about what kind of help technology can provide when you, as a teacher, find it difficult to organise things yourself. For instance, figuring out which group is struggling, without having to stand with every single group all the time. Whether such technology will actually improve learning in a regular classroom, and whether the benefits outweigh the costs and extra organisation, remains an open question.

To conclude

This meta-analysis proves yet again that it all comes down to the quality of support provided. Letting students work in groups does not automatically ensure that they will learn effectively. It depends on the activities they perform and the support they receive while doing them. Also handy: this research shows that we should not provide more support than needed to help them learn.