One aspect of the interaction between humans and artefacts has become the issue of emotions. As Raquel Navarro points out robots (humanoid or not) - as the prototypical example of a futuristic artefact - have so far not been too convincing in generating emotions. For an example of current work you can take a look at the Kismet system (http://www.ai.mit.edu/projects/humanoid-robotics-group/kismet/) developed by Cynthia Braezeal (http://web.media.mit.edu/~cynthiab/). A more psychologically grounded approach towards the synthesis of emotional expresion is found in the work on Boldy and other animated talking heads by Dom Massaro (http://mambo.ucsc.edu/psl/dwm/). What is important in Dom’s work is that he addresses the issue of emotion in the context of communication. What is still missing in these approaches, however, is that we do not have a good understanding of how the human observer exactly interprets the emotions that can be generated by the facial expression and the prosodic cues of the synthetic head.

Now these questions can be addressed by exposing human subjects to these cues and to ask them how they interpret these cues. But here we enter the murky territory of psychological research. Murky because we have not guarantee that what humans will tell us is actually true. To circumvent this problem researchers are turning more and more to using tools from cognitive neuroscience in order to correlate behavior, subjective experience and states of the human brain. Hence, in the context of this blog, we see that humans themselves are still not that accessible by modern day science.

In our own research we have worried about this issue from the perspective of music communication. Music is a great medium for the communication of emotional states. We can all experience the impact of music on our emotional state. We often play music or attend a concert to be transported into a different subjective state. As part of a large interactive exhibit, called Ada ([1] www.ada-exhibition.ch), we have investigated with Jonatas Manzolli of the Nucleus of Sound Communication of the University of Campinas in Brazil ([2] www.nics.unicamp.br) how specific musical parameters lead to a specific emotional interpretation. The idea was that Ada – a 200 m^2 human accessible robot - could communicate with its 560.000 visitors how it “felt” about the world. Hence, Ada would communicate “anger” when it could not achieve specific goals such as grouping visitors. This illustrates an important issue in the context of accessibility: what do we want the emotional cues generated by humanoid robots and/or talking heads to exactly communicate? Are they puppets that express “emotions” at their surface that are decoupled from their inner working or is their goals, as in the case of Ada, they want to express? But in the latter case they should have goals to start with, from what are these goals derived? This means that the use of emotional cues only is meaningful when we are dealing with a technology that has goals. Hence, artefacts that do not induce an emotional state in the observer but that actively want to communicate with its users as part of their social environment. This seems a different game we should inspect in a following post.