[ad_1]
Children with autism constantly have problems recognizing the emotional states of people around them, be it a happy or sad face. The therapists have tried to solve this by using robots so that the children imitate these emotions and respond to them correctly.
However, this therapy would work better if the robot interpreted the behavior and interest or enthusiasm of the child without problem. paying attention during the therapy. For this reason, researchers at the Massachusetts Institute of Technology (MIT) have developed a personalized machine learning system that allows a robot to know the interest of an autistic child during these interactions, by using custom data according to the conditions of the child
The robot, equipped with a personalized deep learning network, perceived children's responses with 60% efficiency compared to the evaluations of human experts.
In fact, human researchers say that for an expert, it can be complicated to reach agreement on the behavior of a child with autism, getting almost always between 50 and 55% efficiency. Oggi Rudovic, author of the study and doctor at MIT Media Lab, said:
The long-term goal is not to create robots that replace human therapists, but d & # 's 39, increase their intelligence with key information that therapists can use to personalize. the content of the therapy, and also make more attractive and natural interactions between robots and autistic children.
Rosalind Picard, co-author of the study and professor at MIT, said that personalized learning for every child has been essential in this study, and represents a revolutionary approach to current AI systems that require a large amount of similar data in each learned category.
Rudovic and his colleagues were also able to probe how the in-depth learning network made their estimates, which revealed some interesting cultural differences in every child, for example, Japanese children showed more body movements during episodes of strong commitment to learn.
Source link
