Research is the art of improving mankind’s knowledge.
As soon as research joins forces with technological progress we are able to envision new tools and solutions for the problems of today and of the future. It has happened to all of us to see the 404 Page not found while browsing the internet.
It’s annoying.
This type of error is returned when the resource (link) we are requesting (by clicking) cannot be found. Errors are of utmost importance! The codes (like 404) are well defined to let the user clearly understand the origin of the problem.
When building robots we have to handcraft the errors in a similar fashion that is used for standard HTTP protocol. However, the problem increases in complexity as soon as we build robots that interact with humans. With these, we have to carefully design the appearance, functionality, and behaviors.
The type of errors that we aim to capture in these robots is not limited to technical issues! The interaction itself can progress in an unexpected way, and it is important to keep track of it (for instance with error-like diagnostics).
Let’s envision the future
Imagine a social robot deployed in a mall for customer service. Its task is to navigate among the crowd and engage in conversations with customers of the mall and respond to simple requests like:
- Where can I buy flowers here?
- What time does the hairdresser close?
This robot is not just a tool in a factory. It has to employ social skills to achieve its task! Its success can depend on two separate points:
- an excellent technical platform (hardware and software are working as expected)
- excellent employment of social skills
While the first one holds true for all products (We buy a car and expect it to work properly), the second point is relevant due to the social context the robot is expected to operate. Social skills are difficult even for humans, they vary according to the local culture and demographic factors. The goal of deploying robots that can control their social skills is ambitious but doable!
We as engineers have to keep iterating with human scientists and eventually define a standard protocol for social errors that will arise during interactions with this type of robot. Clearly a misfunctioning of a motor of the robot is a type of error different from using inappropriate social behavior.
One rightfully might ask: “what is an appropriate behavior for this type of robot?”
The development of social robotics has borrowed theories from human-human interactions to increase comfort and trust in the interactions.
Examples of these behaviors are:
- An appropriate interpersonal distance while interacting
- A proper volume of the robot speaker if it has to speak to the user
- A policy for turn-taking in conversations that is acceptable by users
Keeping this in mind we can see that diagnostics of social skills can improve the development of robots. This error (due to its human nature) can hardly be detected by the robot. It needs feedback from the user. The feedback from the user can be implicit or explicit. Explicit feedback can be the user saying “Sorry I don’t like how you are behaving! I’m leaving!”. Implicit feedback can exploit cutting-edge emotion recognition to estimate whether the user is stressed or bored during the interaction. This latter tool can use models from Artificial Intelligence to detect events like disengagement that can be due to dissatisfaction with the interaction.
Given these simple examples of appropriate robot behaviors, it will be extremely important to define a standardized social errors protocol for robotics that can help the future development of truly social robots.
In this sense, the 001 Error could be mapped to the social misbehavior of the robot, providing the users and the developers with information on the state of the interaction (exteroceptor sensors) when the error was detected.