News

Daniel Wolpert Discusses Movement and Behavior in UCSB Lecture

Neuroscientist and Professor of Engineering at the University of Cambridge Daniel Wolpert gave a lecture at the Sage Center yesterday called “Probabilistic Models of Sensorimotor Control and Decision Making,” in which he presented recent research into links between action and probability frameworks.

Scott Grafton, a professor in the Psychological and Brain Sciences Department, introduced Wolpert. According to Grafton, Wolpert’s work made a huge impact in the 1990s, when motor control research was focused on “applying mathematical methods to measure human behavior.” By attempting to find the reasons behind movement-based findings in action science, Wolpert reframed perspectives throughout the entire field, according to Grafton.

“What really struck many of us was Daniel appeared and just asked common sense questions about what the organism does,” Grafton said. “It’s not actually out there to solve the degrees of freedom problem, that’s not why we evolve.”

Wolpert began by asking audience members to consider “the simplest question we can ever ask ourselves: Why do we and other animals have brains at all?”

The question is important, according to Wolpert, considering the number of species without brains and in attempting to begin any study of the brain and mind in humans. Wolpert posited that the brain exists solely in order to “to produce adaptable and complex movements,” as movement is the only way to affect the world around us.

“There can be no benefit to the memory of your childhood, or the singular color of a rose, if they don’t express themselves through the movement system, which is the only way you can change your chance of survival,” Wolpert said.

In order to examine the connection between movement and intent, Wolpert said his line of work in attempting to “reverse engineer” how the brain controls movement. For example, by focusing on movements by the arm and the hand, researchers can see how the possible range of future movements, and the “noise” surrounding these, contribute to a chosen path of action.

The concept of noise as forms of interference was central to Wolpert’s talk. Sensory feedback and motor commands are extremely noisy, Wolpert stated, noting that the ability to overcome this noise in professional sports is highly-valued.

According to Wolpert, a large amount of sensory-motor research in the past decade has focused on analyzing uncertainty in a Bayesian framework, which, based upon Bayesian statistics, describes how the brain creates beliefs about the world using limited available information.

Wolpert presented two sources of information by which beliefs are generated. The brain can use data in the form of sensory input and prior knowledge, which is memory or genetically encoded information. Using probabilities, Bayesian analysis goes through how to use both sources to make optimal beliefs.

As an example, Wolpert asked his audience to imagine playing tennis, preparing to receive a ball in their court. Using data in the form of sensory feedback from the visual system, one would estimate the ball landing in a particular area. Adding prior knowledge of how unpredictably the ball bounces or the attributes of the court, the player can combine both sources of information to estimate an optimal area of where the ball will actually fly.

Wolpert addressed research that further used Bayesian methods to explain why people tend to overestimate their performance on tasks, such as driving or teaching. In a study in which subjects looked at a screen with a red ball, a blue ball scrolls over the screen at a high speed and subjects were asked to push a button once the blue ball was directly below the red ball. The moment they pushed the button, the blue ball vanished. Performance error was measured as how far the blue ball ended up from the red ball, and estimation error was the difference between where participants estimated it vanished from where it actually vanished. Through this method, participants would indicate how well they thought they performed on the task.

Subjects were also played a video of a “computer” performing the task, which was actually a replaying of the participant’s trials back in a random order. Subjects reported a much statistically narrower prior for their own trials as opposed to that for the computer, indicating that they felt more optimistically about the accuracy of their own performance.

Wolpert presented computer models of movement, showing how people tend to move in predictable, less variable ways and that either accuracy or energy expenditure is sacrificed depending on the type of movement required.

“I would argue, any task you can think of humans to do can be put in a probabilistic framework,” Wolpert said.

Wolpert presented research on methods of measuring gains during movement, and how the brain “recomput[es] on the fly, when the task demands change.” For the framework in which these probabilities are calculated, Wolpert said humans have certain levels of structural knowledge and expend time and energy in deciding upon the parameters of these structures. For example, knowledge of bikes can be considered a low-level structural knowledge, so that when one learns to ride one bike, that process can be generalized to many forms of the vehicle.

”What we really think now, is most of our life is about parameter learning.” Wolpert said. “We may learn structures … early in life, but when we go about and handle objects we don’t have to learn about structure again, all we have to learn about is parameters.”

Wolpert concluded by presenting research linking decision making with motor control, which explores how people change their minds.

“Whenever you make a decision, there’s this period of information, of sensory information you can’t ever use to make a decision. This is true of any decision you make,” Wolpert said.

Through recent studies, Wolport has concluded that there remains a “continuous flow of information from the decision-making process to the motor process,” and ended his talk by reminding his audience of the probabilistic complexity underpinning seemingly straightforward decisions that lead to motor processes.

 

This story appeared as an online exclusive on Friday, March 7, 2014.

Print Friendly
Tags: , , , , , , ,

One Comment

  1. If you desire to improve your familiarity just keep visiting this web page and be updated with the newest information posted
    here.

Leave a Reply

Your email address will not be published. Required fields are marked *

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>