In recent years, interest in robotics has grown significantly, driven by the rapid spread of robots in industry and by an increasingly heated debate about their impact on work. At the same time, for more than a decade, a strand of research has been developing that studies the relationship between robotics and education. What do we know about using robots as learning tools? What does the research say, and what can we start to imagine for corporate training?
What is social robotics
To consider the relationship between robotics and education, we need to step out of the world of industrial robotics and into that of so-called “social robotics“. This is a field of robotics devoted to developing robots designed to interact with humans in social contexts. Social robots are defined as “agents with some degree of physical embodiment that interact with humans through verbal exchange, cooperative problem-solving and autonomous decision-making”. In practice, they are machines equipped with motion sensors and software for speech and emotion recognition, capable of sustaining limited but meaningful social interactions.
In the study of education and robotics, this has a specific implication: the social robot is not a tool designed to deliver content. It is specifically designed to take part in the relational dynamics of learning.
It is no surprise, then, that many studies on the use of robots in schools list teacher shortages or the difficulty of providing individual attention among their contextual factors. Nor is it surprising that social robotics is finding wide application in areas such as elderly care and healthcare, where the demand for human connection structurally exceeds the availability of staff.
From the school world: the most interesting findings
Most studies on educational robotics come from schools, where the research is more mature. A review published in Science Robotics in 2018 by Belpaeme and colleagues had already documented positive effects of social robots on both cognitive and motivational outcomes.
A much more recent meta-analysis, published in 2026 in Frontiers in Robotics and AI by De Winter and colleagues at Delft University of Technology, confirms and expands on that picture. Social robots can promote and support learning, but their effectiveness is not an intrinsic property of the technology. One of the most relevant findings concerns the relational dimension, meaning how the robot is placed within the teaching relationship. When the robot acts as an assistant in a co-teaching relationship with a human teacher, the positive effect on learning is very strong. When it is positioned as a complete substitute for the teacher, it produces no benefit at all.
The study also identifies some psychological mechanisms that explain why robots can facilitate learning. One of the most interesting is the reduction of performance anxiety: because they are perceived as non-judgmental partners, robots create a safe space for practice and mistakes. Another is the so-called “protégé effect“: when students have to “teach” something to the robot, the cognitive benefit for them is significant.
Some necessary caution
De Winter’s meta-analysis itself urges readers to interpret these results with some caution.
The study supports the hypothesis that the robot’s physical presence, including its ability to gesture, direct its gaze and occupy space, fosters attention and motivation in ways that traditional digital technologies cannot replicate. However, this is not an automatic advantage. In some contexts, too much emphasis on the robot’s social features (movements, facial expressions, tone of voice) can become a distraction, increasing cognitive load and hindering learning rather than helping it.
Other limitations concern speech recognition and speech synthesis. Robots can struggle to understand children’s voices, and synthetic speech often sounds unnatural.
The study also found that the tone of the papers varies depending on their geographic origin. Studies conducted outside Europe tend to report larger effects and take a more optimistic tone, while European studies are on average more critical and cautious. The reasons for this disparity are unclear. It could stem from differences in methodological criteria, from publication bias, or from the fact that in some non-European contexts, such as Asia, robots are a more established and familiar presence in education.
What about the workplace? Early results
Research on social robotics in the corporate world is still in its early stages, but a recent study that tested the use of the Pepper robot in employee training produced some interesting results. The study compared training with Pepper to traditional learning on a digital platform.
The main findings concern:
- how the experience was rated: interaction with the robot was rated very positively, with a level of “enjoyment” significantly higher than screen-based learning;
- learning outcomes: no statistically significant differences in learning outcomes emerged.
In other words, the robot did not make people learn more, but it made the experience more enjoyable. The researchers describe social robots as promising tools for diversifying corporate training offerings and increasing employee engagement. They stress that emotional engagement is often the hardest variable to secure, especially in e-learning, which by its nature suffers from limited social interaction.
However, the study also highlights some critical issues. On a practical level, the robot has its own pace, which some perceive as slow compared to platform-based training, and it requires people to be physically present in the office, a real constraint in a hybrid work environment. On a technical level, the limitations already known from school research emerge again: unnatural synthetic speech and movements that are sometimes distracting.
Looking ahead
Research on robots in education and training reinforces a point we often hear, including in relation to Artificial Intelligence: no technology alone solves every problem. Integration with human input and alternation with other tools consistently prove decisive in generating a real impact on learning outcomes.
The blended approach is confirmed as the most effective. And if we really do see robots spread in corporate training, the challenges will involve both planning, tied to integrating the robot into existing learning paths, and the design of the tool itself: gestures, expressions and voice are variables that, as we have seen, directly affect the user experience.
Finally, the study on Pepper raises an interesting point. Younger generations see the quality of training as an important factor when choosing and evaluating a company, and they look for engaging, meaningful learning experiences. Yet companies are finding it increasingly hard to ensure adequate levels of engagement. These are the workers who will become the core of organizations in the coming years, and attracting and retaining talent increasingly depends on the ability to offer development experiences that live up to their expectations. Robots could become one of the tools available for building those experiences: varied, engaging, and able to make a lasting impression.
