Toward collaboration between humans and machines in real-world environments where uncertainty is inherent, we demonstrated that the intentions behind machine actions can be shared with humans through an exchange of a clipboard between a real robot and a person. In the demonstration, we compared how humans interpret the behavioral intentions of two types of robots. One type follows an artificial minimal self (AMS), which extends its internal model to the actions of others based on active inference. The other type only executes pre-designed action sequences, which we used as a benchmark. Because the AMS virtually shares the processes inferred by a self-agent and another agent, the intentions of the two agents are expected to be shared in the same way that humans share their intentions by inferring and sharing processes in order to address uncertainty. The results showed that, in some cases, people tend to recognize the intention of the robot following the AMS as slightly higher than that of the robot executing pre-designed action sequences.