|
|
|
 |
|
We are developing new theoretical approaches that will apply the mathematical apparatus of modern theoretical physics to computationally difficult problems in the modeling of human decision making and social, financial, biological and WWW networks. Collaborations with Leicester, Manchester, the RIKEN Brain institute (Japan) and the Salk Institute for Biological Studies (USA) are involved in this promising project. In a quickly changing risky environment, options of a decision maker are often confined to a narrow path between undesired prospects. For example, think of a pilot steering out of a steep banking manoeuvre or a surgeon counting seconds before a patient’s loss of blood becomes fatal. Their decisions at one instant define a reality they face in the next instant, and the consequences of previous decisions must be re-evaluated as new information keeps flowing in. Our goal in this project is to advance understanding of how humans achieve the remarkable versatility and efficiency of decision-making in such situations. Formal analysis of such tasks is formidably complex, which is why previous studies of human decision-making have been confined to very simple tasks. Example of modelling we proposed.Consider an observer trying to estimate the size of an object using two senses: vision and touch (Figure 1). The observer estimates distance between two sides of object that are either parallel (A1) or orthogonal (B1) to her line of sight. In the latter case, the back surface is seen through the front semi-transparent one, which makes the visual task more difficult and size-estimation less precise, in B1 than in A1. For touch, the difficulty of this task in the two cases is about the same. Since the precision of visual estimates varies as a function of object orientation, and the precision of touch does not, the contribution of the two senses to object perception should vary as a function of orientation, as predicted by an ideal observer: a maximal-likelihood estimation model of sensory integration. In A2 and B2, the red curves represent the distributions of estimates by touch and the dark bluecurves the distributions of visual estimates: the wider the distribution the less precise the estimate. The variance of the visual estimates is greater in B2 than in A2 and the variance of the haptic estimates is the same. The ideal observer maximizes precision of combined visual-haptic estimates. We plan to compare how humans follow the ideal behaviour and when human behaviour deviates from the ideal one.
Collaborations with Leicester, Manchester, the RIKEN Brain institute (Japan) and the Salk Institute for Biological Studies (USA) are involved in this promising projec
|
||||
|
|
