'''Article:''' Online Behaviour Classification and Adaptation to Human-Robot Interaction Styles by Doroth´ee Franc¸ois, Daniel Polani, and Kerstin Dautenhahn

'''Introduction to paper:'''

The sensors of an Aibo robot were monitored and the information sent through an algorithm, in real-time, to determine whether the toy was being played with gently, or not, and then to change its personality accordingly. The algorithm was mostly able to correctly determine harshness of interaction. Transitioning between personalities took a little while (10 to 19 seconds). Also, one sensor was able to mislead the algorithm a little bit if it was over used (because it was binary instead of analog).

'''Application to personal research:'''

We could incorporate machine learning into TiLAR in a way that, depending on how sensors are being handled (or something), the remote control will show a couple options of what to do next (as pre-scripted by the therapist) but the one that the algorithm thinks would be the best choice will be highlighted. If not determined by sensors, it could try to predict what the therapist would choose next, based on previous choices, and highlight that one.

'''Questions:'''

• Why do ''life-like'' ''toys'' fascinate people? Is it because they're like reality, but more under your control?

'''Additional notes from paper:'''

• Making interactions “playful” may engage the child
• The robot's purpose is to act as a bridge that will lead an autistic child to interact with other people

« Invoking Social Behaviors