This week, we finished discussing the last of the background reading and started talking about our initial experimental design.
The last reading was “a survey of socially interactive robots.” Basically, it was an overview of the history of social robots and the research that has been done in that field so far. It also discussed the different problems or research that could be done in the near future, as well as some long-term research. The reading first starts by talking about “human social characteristics.” These include the following.
- Expressing and/or perceive emotions
- Communicating with high-level dialogue
- Learning/recognizing models of other agents
- Establishing/maintaining social relationships
- Using natural cues (gaze, guesters, etc.)
- Exhibiting distinctive personality and character
- Possibly learning/developming social competencies
If robots can be socially interactive, then they can be used for many different fields, including research platforms, toys, educations tool, therapeutic aids, search-and-rescue, etc. Building social robots also comes one step closer to figuring out how to build empathy and true understanding into robots.
The best approach to making robots socially interactive is to use “functional design.” Functional design works on the theory that a robot can be designed so that it appears to be socially intelligent, but its internal design does not actually have basis in science or nature. Robots using functional design would be very useful if they were short-term interaction robots, have limited embodiment, or limited social expression. Many video games and electronic toys use functional design.
Once the robot is designed, it is also important to study emotion. There are three basic theories of emotion. The first theory tries to repesent emotions in discrete categories (such as happiness). The second theory characterizes emotions based on continuous scales or basis dimensions (such as arousal or valence). The third theory, also know as a componential theory, is a combination of the first two theories. The reading also discuesses artifical emotions and their implications on human-robot interactions.
Another big part of socially interactive robots is dialogue. There are many different types of dialogues that need to be studies: low-leve, non-verbal, natural language, etc. After studying all of these different aspects, they need to all be tied together in order to build a proper socially-interactive robot.
So, in order to study these social relationships further, we decided to use actor script evaluations. We will be doing this using different “games”, or placing people in different situations and seeing what they would do. Some possibilities for these games include The Ultimatum game, The Equitable Division Game, The Minority Game, Insider’s Knowledge Game, The Chicken Game, and last but not least, The Investor-Trustee Game. There will be more on our experimental design and the specifics of each of these games in the weeks to come.