Week 4
“Introduction … Video Lecture … Weekly Assignment (W4)”
Summaries
- Week 4 > Video Lecture > Video 1: Cognitive Development in Infants
- Week 4 > Video Lecture > Video 2: Theories of Infant Cognitive Development
- Week 4 > Video Lecture > Video 3 Cognitive Developmental Robotics (1)
- Week 4 > Video Lecture > Video 4: Video 4: Cognitive Developmental Robotics (2)
Week 4 > Video Lecture > Video 1: Cognitive Development in Infants
- Infants acquire various types of abilities such as self-other cognition, imitation, socially referencing and so on.
- If you look at their cognitive abilities infants show after their births, you may not see their commonness between their behaviors.
- The infants’ innate abilities are very limited.
- Then based on these limited inherent abilities, they can acquire various types of abilities through the interaction with the environment.
- So the one important point here is that in order to know what their mechanism- underlying mechanism for cognitive development is, we have to carefully investigate what’s other inherent abilities for infants which enable them to acquire various kinds of abilities.
- For many years developmental psychologists investigated what enables infants to acquire the social cognitive abilities.
- Many robotics researchers try to understand the underlying mechanism for infant development by designing artificial systems, or robots.
Week 4 > Video Lecture > Video 2: Theories of Infant Cognitive Development
- So first, the theories of infant cognitive development.
- He investigated how infants acquire their cognitive abilities, and when they acquire such abilities.
- They actually designed several developmental studies, like in conservational task and egocentrism, which allowed researchers to understand when infants acquire their abilities to understand the permanence of the object and their viewpoint of the person.
- Then this close analysis of infant behaviors enabled him to propose the theory called genetic epistemology.
- This theory provides you the details of their infant development- what kind of mechanism may enable infants to acquire cognitive development, cognitive abilities, and so on.
- As a part of this theory, he also proposed there are six-staged order in infant development.
- At the first stage, infants show only the simple reflex.
- If you put your fingers very close to the infant’s mouth, you can observe that infants try to suck your finger.
- Or if you put your finger on the palm of the infants, you can also see that they try to grasp your finger.
- These reflexing abilities enable infants to explore the world, and also establish a very basic interaction with other people.
- Based on their experiences, infants also start to produce their behaviors, which explore their own body and the environment.
- The second stage is called primary circular reactions, in which the infants produce an event that happened by accident.
- So at this stage, infants still don’t have the intention to produce an action, but maybe randomly explore the environment.
- At this stage, the infant’s behavior is now directed to the environment rather than their own body.
- Infants start grasping an object, shaking it, and then observe what the outcome of their action is.
- Zone of proximal development is defined by the difference between what a learner can do without anyone’s help, and the things that infants can do with help.
- With their caregivers’ help, the infants can enrich their abilities.
- One example is that a very young infant who cannot walk yet.
- So such a help should feed into this zone of proximal development, so that infants can experience the walking behavior with their help.
- In the imitation, the caregivers often reproduce their infant’s behavior so that infants can recognize the correspondence between their own body and their caregiver’s body.
- So there are many, many examples which are showing the importance of caregivers in infant development.
Week 4 > Video Lecture > Video 3 Cognitive Developmental Robotics (1)
- Now we move on to the robotics studies, which investigate the infant development by designing artificial systems for robots.
- Robotics researchers try to understand the infant developmental mechanism by designing their developmental mechanism for robot.
- Previously, many researchers analyzed the infant behavior.
- When infants acquire the abilities such as joint attention, imitation, or what kind of abilities infants can acquire in the first few years of life.
- There is an open question that we don’t know yet what internal mechanisms enable infants to acquire such cognitive capabilities? So by analyze- sorry, by designing the artificial intelligence systems, we can better understand the internal mechanism of infants.
- So by studying how infants acquire the abilities, we can also design more valuable and more adaptive robots.
- Developmental studies have examined the infant’s ability to recognize the self, and also to differentiate the self from others.
- Here the question is which image infants prefer to look at.
- Then if they get older, they start looking at the third person view because the third person view seems to be more interesting to the infants.
- This study showing that the infants can discriminate the self from the others.
- The experimenter found out that the infants show social behavior such as smiling and speaking when they observe their mothers in their monitor.
- This research indicating that infants can already distinguish the self from the others.
- The question here is what kind of mechanism enable infants to distinguish self from others.
- The others’ behavior have less spatial and temporal contingency compared to the self.
- It is known that human infants have less acuity in their perception.
- This immature from- to mature development might play a role to develop their correspondence between self and other, which is related to the mirror neuron system.
- We look at their motor neurons and the visual spaced neurons in the robot.
- Robots run the correspondence between the motor neuron and then visual outcomes.
- Here, the point is that due to the limited capabilities in robot’s perception, robots cannot distinguish between the self motion and other’s motion in the visual space.
- Which contain both the self movement and the other movement, is now connected with their corresponding motor neurons through Hebbian learning.
- Of course, as the robot improves their perceptual abilities, the robots start discriminating the self movement and other’s movement.
- Then the cluster which corresponds to the self movement and the other movement also now is differentiated.
- The point is here there is motor neurons associated both with the self motion and other’s motion because of their one connection in the early stage.
- Please remember that, in the early stage, the self motion and other motions are included in the same cluster.
- Because of this connection, robots can now recognize the correspondence between self and other through their connections to the one same motor neuron.
Week 4 > Video Lecture > Video 4: Video 4: Cognitive Developmental Robotics (2)
- In the first, we examined how the visual space gradually develops as the robot’s visual capability improves.
- You can see that, in the very early stage of development, robots still see the clusters which contain both the self-motion and others motion.
- One is for the others movement, and one is for the self-movement.
- So this result showing that the visual development, the improvement in the visual acuity, allows the robot to gradually separate the self- movement and the others movement.
- First you can see that the robot acquires very high connections here, which actually connect between the corresponding movement in the vision and the motor neurons.
- If the robot executes this actions, robot can of course observe the visual appearance of the two hand movements.
- This mean that, for example, if the robot observes the other person who moving two arms in a vertical way, the robot can imagine what is the corresponding action in their motor neurons.
- So this high connections allows the robot to find a correspondence between the others behavior and the self’s behavior, which is their most important role in the mirror neuron system.
- In the left case, we designed the visual development in robots, which mean that the robots start with very immature, very low acuity in vision, and then gradually improves their visual acuity.
- On the right side, the robot does not have the visual development, which mean that the robot use the highest acuity of vision from the beginning of development.
- In this case, the robot could not acquire high association between the self-movement and others motion.
- So this result showing that their immaturity in vision in the early stage of development plays an important role for robots, and maybe also for infants to acquire the correspondence between the self- movement and others movement.
- Here, the robot first does not have any experience to explore the world.
- Our hypothesis is that first, the robot explore the environment, and then this experience allows the robot to acquire the predictor of its behavior.
- This robot pushes the blue car in the environment, and he can recognize what outcomes can be observed after producing this pushing behavior.
- Then this learning allows the robot to acquire the predictor of the sensori-motor information.
- The robot was observing this person’s recent behavior, and then expect what would be the outcome of this person’s action.
- Now the robot expect their predicted behavior, predicted outcomes.
- Using this high prediction error, the robot now execute an action which result in the altruistic behavior.
- So here the point is that even if the robots do not have their motivation to help others, it can still help other persons just by reducing its prediction error.
- Let me show you the video which can- which explains how the robot first explore the environment, and then help other persons.
- Then through the exploration with the world, robots now can help the other persons who could not achieve the goal.
- Now this person tries to reach for the red marker, but then if the red marker was too far for him, he could not achieve the goal.
- In that case, the prediction error becomes high, which motivates the robot to execute the predicted action goal.
- Then this larger prediction error allows the robot to execute an action which result in their altruistic behavior.
- Again, here the point is that robots does not need to have the motivation to help others at the very beginning, but just by their motivation to minimize the prediction error it can produce the altruistic behaviors.
- So these are the two examples we conducted using the robot.
- We observed that the inherent capabilities of infants, like in contingency learning of sensory and motor information, and predictive learning through their experiences with the world, this inherent abilitie allows robots to acquire the cognitive abilities like the infants.
- Actually the design of the environment played an important role in our robot experiment, as well.
