Human Explanation of Robotic Behavior


Extensive research has been conducted on the attribution of intentionality, mind, and emotions to robots; however, the majority of present-day studies does not provide a detailed analysis of the explanation structures that underlie interactions between humans and robots. We believe that understanding how people interpret and explain the behavior of robots is essential to unravel the dynamics of human-robot interaction.



Open questions we aim to address:


  • What theoretical vocabulary is used to describe the phenomenon to be explained? What sort of linguistic concepts are employed?
  • How do people describe the mechanisms governing the behavior of robots in ordinary interactions?
  • What are people’s dispositions when giving explanations? For instance, are people more inclined towards giving teleological or causal-mechanistic explanations?
  • In case of mentalistic explanations, what kind of ‘mind’ and mental entities (e.g., propositional attitudes, information-processing modules, …) are attributed to robots?
  • What makes something a good explanation of a robot?


To address these questions, one must go beyond the study of mental-state or trait-attribution and reconstruct the finer-grained explanations of robotic behavior. We believe that reconstructing human explanations of robotic behavior (HERB) is crucial for comprehending the dynamics of human-robot interaction, designing sociable robots, addressing robo-ethical concerns, and informing the design of cognitive architectures for social robots.

While a vast philosophical literature on explanation and understanding exists, it has largely been neglected in studies on HERBs. Among the aims of this project is to incorporate insights from philosophy, psychology, and cognitive science on how people generate, select, evaluate, and communicate explanations. Furthermore, understanding HERBs is particularly significant in the context of educational robotics and the potential for robots to enhance technological literacy among students.




The goal of the HERB project is to elucidate how people explain the behavior of the robots they interact with. HERBs - Human Explanations of Robotic Behavior(s) will be analyzed along three dimensions: the characteristics of the robotic behavior to be explained (explanandum), sets of theoretical assumptions about the robot's functioning (explanans), and the relationship between explanandum and explanans.



Analyzing the explanans reveals assumptions about the robot's architecture, while examining the explanans-explanandum relationship uncovers the explanatory pattern employed, be it finalistic, causal-mechanistic, probabilistic, teleological and many others. This analysis aims at illuminating why some HERBs foster understanding and how they can predictively or interventionally guide human-robot interactions.

Furthermore, this study investigates teachers' explanations of robotic behavior, i.e. what internal structures do teachers attribute to robots? This question begs an examination of how the attributed internal structures depends on the teachers' educational background and the characteristics of the robot itself. By analyzing the assumptions underlying teachers' HERBs, this line of research seeks to understand the cognitive models employed to make sense of robotic systems. The findings can inform the design of educational robots that align with teachers' mental models, facilitating effective human-robot interactions in educational contexts.




UNIMIB (RobotiCSS Lab):


Edoardo Datteri

As a philosopher of science, I primarily work on the methodological foundations of biorobotics, Artificial Intelligence, and Cognitive Science. More specifically, I reconstruct and analyze the validity of methodologies involving robots and bionic systems, as well as robots interacting with animals and humans, to study living system behavior and cognition.
My interests also concern the role of robots as tools to intervene in, and theorize the mechanisms of social cognition, still from a methodological perspective, and the methodological foundations of educational robotics.



Silvia Larghi

I have a background in computer science and engineering. After an internship in robotics at the JRC - Ispra (VA), I worked in software engineering participating in EC-funded international research projects. I taught technology in school, where I coordinated the team for digital innovation. For several years I have been designing and conducting laboratories of educational robotics and artificial intelligence in schools and training courses for teachers.
My research interests concern philosophy of Robotics and Artificial Intelligence, philosophy of Cognitive Sciences and Human-Robot Interaction.



Nicola Zagni

I graduated cum laude at the University of Bologna with a thesis on the epistemology of Machine Learning and spent time abroad at the University of Davis, California, where I studied social epistemology and computer science. As a philosopher in the lab, I am driven by curiosity and inspired by strangeness, two recurring themes in my day to day life. Simply put, I often find myself asking: how did this happen? If I was expecting this to happen, why did this other thing happen instead?
My main interests lie at the intersection of Philosophy of Science and Cognitive Science, with wide implications for Artificial Intelligence.