面试报告
题目：Measuring the internal representation of uncertainty in visuo-motor decision tasks
报告人：Hang Zhang, Ph.D.
Postdoctoral fellow, Department of Psychology, Center for Neural Science,
Center for Neuroeconomics, New York University
时间: 2:30pm, July 9th, 2014
地点: Rm. 311, New Life Sciences Building
The unpredictable error inherent in movement is an important source of uncertainty in everyday life: the action we select may not always lead to the outcome we intended. In the past decade a wide range of studies have compared human performance in motor tasks to a normative framework, Bayesian Decision Theory. Human performance is often (but not always) close to Bayesian optimal but these studies do not readily allow us to learn how subjects represent and manipulate uncertainty information.
I have developed a method that allows me to estimate the subject’s internal model of her own error distribution in a simple reaching task and compare it to the actual distribution of motor error. Although actual motor errors were typically Gaussian, subjects’ internal models were quantized, resembling a step function with a very small number of steps (3-5). I will report three different analyses leading to the conclusion that subjects’ internal models are quantized. I also report reanalyzes of data from an earlier experiment consistent with quantization. I will describe phenomena in economics (skewness preference) that may be simply explained as a consequence of quantization and discuss neurally plausible representations of quantized distributional information. As findings on the computational-theory level (in Marr’s term), my findings clarify the goal in the search for the neural representation of uncertainty.
欢迎各位老师和同学积极参加！