How to teach a robot?

In the rapidly advancing field of robotics, the ability to efficiently teach robots through various forms of feedback is revolutionizing how these machines learn and adapt to our world. Unlike traditional programming methods, which require extensive coding to modify behaviour, feedback-based teaching methods offer a more intuitive and dynamic approach. This efficiency in feedback not only accelerates the learning process but also empowers users to tailor robot behaviours to their specific needs and preferences. From households to industries, the flexibility to personalize robotic actions opens up a plethora of applications, making robots more useful, adaptable, and personal than ever before. Here, we explore the different ways to teach robots, including evaluative feedback, corrective feedback, advice, instruction, and demonstration, and how these methods contribute to rapid and personalized robotic learning.

Evaluative feedback

Evaluative feedback involves providing robots with assessments of their actions in terms of success or failure. This binary form of feedback is akin to training animals with rewards and punishments. It’s particularly effective for tasks where the desired outcome is clear, but the path to achieving it may vary. By repeatedly assessing robot actions as either correct or incorrect, robots can use algorithms like reinforcement learning to adjust their behaviour towards the desired goal. This method shines in its simplicity and is powerful for shaping basic behaviours and decision-making processes

Corrective feedback

Corrective feedback goes a step further by not only indicating that an action was incorrect but also suggesting the correct action or direction of correction. This can be as simple as adjusting the robot’s movement direction or as complex as providing specific adjustments to robot manipulations. Corrective feedback is especially beneficial for tasks requiring precision, as it accelerates the learning process by directly guiding the robot towards the correct behavior, reducing the trial-and-error phase commonly associated with evaluative feedback.


Providing robots with advice incorporates human knowledge directly into the learning process. Unlike evaluative and corrective feedback, which react to robot actions, advice can be given before an action is taken, guiding the robot towards certain behaviours or strategies. This preemptive feedback can significantly speed up learning by steering robots away from potential mistakes and towards more efficient paths. Advice is particularly useful in complex scenarios where human expertise can provide shortcuts to learning.


Instruction involves explicitly telling the robot what actions to perform in specific situations. This method is closely related to traditional programming but is more flexible and interactive. Through verbal commands, text input, or graphical interfaces, users can directly communicate desired actions to the robot. Instruction is highly effective for quickly implementing new behavioUrs or modifying existing ones, allowing for rapid personalization and adaptation to new tasks.


Teaching by demonstration, or imitation learning, is one of the most intuitive and natural methods for human-robot knowledge transfer. Here, the robot observes and replicates human actions to learn a task. This method leverages the robot’s ability to perceive and interpret human actions, translating them into its own motor commands. Demonstration is particularly powerful for complex tasks that are difficult to describe in detail, as it allows the robot to learn from the nuanced movements and decisions of human experts.

The efficiency of feedback-based teaching methods is a game-changer in the field of robotics, enabling rapid learning and customization to specific user needs. Whether through evaluative feedback, corrective feedback, advice, instruction, or demonstration, these methods offer a flexible and user-friendly approach to programming robots. By leveraging these techniques, we can look forward to a future where robots are not only more capable but also more personalized, better understanding and assisting us in our daily lives.