Current projects in HRI-JP
HRI-JP Senior Researcher Osamu Shouno
In recent years neuroscience has suggested that human perception is active rather than passive. The human brain is not passively imaging knowledge extracted from the physical world outside the brain as in a television set. The brain is actively recreating what is going on in the outside world.
For example, a tomato lighted by white light looks red. This is because tomatoes reflect red light. As seen in this example, objects have unique physical properties that reflect or emit light of specific wavelengths. Humans may seem to relate the wavelengths to color in the process of perception. However, this is not the case. Even if we light a tomato with blue light so that red light does not reflect, humans perceive tomatoes as red objects, not blue objects.
The brain looks at colors of all objects that it sees, and predicts the real color of the objects. This means that the brain perceives colors by estimation, not by the physical properties of light that enter the eye. Estimation is not limited to color. Sounds and scent exist only in the perception that the brain created.
Professor Chris Frith (University College London) has remarked on this statement as “cognition is an illusion backed up by reality”. When perceiving something, the brain starts by estimating the world around it based on a model of the world inside the brain. Then the brain improves the model by making corrections to the estimation by fitting to crude and fuzzy signals entering the eyes, ears and skin. Such a system to build a model of the external world uses whatever information it can to make a better model. The source does not matter: It can be vision, audition or feeling as long as it is useful.
Movements and actions that people make become important in fitting actual signals to the model estimated by the brain. We use our bodies to work on the world around us, and verify our estimates by observing what happens. For example, moving our heads is a great help in identifying the direction a sound is coming from. In another example, holding a cup in our hands will improve our understanding of the shape of the cup compared to just looking at it.
We want to emphasize that the benefits of moving and taking action are not limited to making a better model of the world. The model of the outside world that was built can be used to predict how the world changes when we take actions in the world. Such a model is very useful to achieve a given goal by taking actions in the chaos of busy and noisy senses.
What tricks does the brain use to implement this active mechanism? Unfortunately we do not have the answer, but we have some clues. The human brain is a network with a characteristic structure formed by a vast number of neurons, called the neural network. All mental phenomena created by the brain come from activities of neurons in the neural network. In fact, some neurons are known to act when there is a sensory stimulus or when an action is taken.
We are focusing on nerve activities that are difficult to directly relate to sensory stimuli or actions. Such activities were initially regarded as noise, however in recent years they have come to be considered part of autonomous cooperative activity in the neural network. We believe this autonomous activity is part of the active mechanism of the brain.
We are looking at the basal ganglia, an organ in the brain, as a first step to verify this hypothesis. It is known that significant autonomous neural activities can be observed in the basal ganglia. The function of the basal ganglia is considered to be the linking of perception to action through associative learning.
Based on these observations, we believe that the basal ganglia have a mechanism to use its autonomous activities to actively find and learn cues on and triggers of actions to be taken from a noisy and busy chaos of senses.
We are verifying this hypothesis from both functional and physiological aspects by constructing a model of the neural network in basal ganglia on a computer, taking into account the biological details, electric properties of neurons and the connections between neurons. We are investigating the learning ability of this model by simulating (experimenting on a computer) the dynamic behavior of neurons.
The society we are living in is very complex. Unexpected or unpredicted things come every day, and things that exceed our knowledge happen frequently. However, people can live in harmony with a world that is changing. This property is difficult to achieve with the conventional engineering approach. We are looking for methods to overcome this difficulty by pursuing the mechanism of the human brain.