Lecture 10: Social robotics and human-robot interaction Flashcards
What are the 4 skills of social interaction and learning?
Joint attention: gaze and pointing gestures
Imitation: Body movement, action/goals
Cooperation: spontaneous altruistic behaviour
Theory of Mind: ability to attribute beliefs, goals, and percepts to other people
Choose an example of social learning and explain a model.
The example will be a robot and a human looking at a ball, and the model is the joint attention and gaze model.
The first stage is the sensitivity/ecological stage where the robot looks at the ball regardless of the human’s gaze direction.
The second stage is the geometric stage which is joint gaze only when the ball is in the robot’s field of view.
The third stage is the representational stage where the robot can find the ball otuside its own field of view.
Match each part of the cognitive architecture of the join gaze model to its definition:
- Salient feature detectors
- Visual feedback controller
- Self-evaluator learning module
- Internal evaluator
- Gate module
A. To move the head towards the salient object in the robot’s view
B. Uses selection rate to model non-linear developmental changes
C. Neural network learns mapping
D. Checks if there is object at the centre of image
E. Colour, edge, motion and face
1 E
2 A
3 C
4 D
5 B
What is the difference between imperative and declarative pointing in joint attention?
Imperative pointing is to request an object when the agent is not looking at it.
Declarative pointing is to create shared attention.
What is a model used for imitation in social learning?
the HAMMER architecture
What are some challenges in Human-Robot Interaction? Explain how to overcome the challenges.
Speech recognition/production can be overcome by ASR (automatic speech recognition) and speech synthesis. A higher quality microphone can increase performance.
Action recognition and intention reading can be achieveed through deep learning models such as OpenPose.
Trust and acceptability by prioritising the ToM of the robot.
Why does speech recognition work better with adults and not children?
Children have a higher pitch and disfluencies.
What are HRI protocols for measuring trust.
Price game judgement (changing the price to deceive the robot)
Investment game (trusting means able to invest more)
Give an example of how the cooperation seen in children is implemented in robots.
Children act altruistically without reward, especially when the role of the other person is a caregiving role. Robots can use an action storage mechanism to store actions that do not need to be relearned.
How can ToM be prioritised in robots?
Robots must be able to read human’s intentions and react appropriately. This can follow the hypothesis used for children, which is Baron-Cohen’s “mind-reading” system. It attempts to read intentions by looking at physical features such as eye direction.
