Joan’s web

Research on Active Perception

Presentation

My research is carried out within the group of Vision and Control of Action VISCA and the Institute for Neurosciences (ubNeuro). We study how people use visual information to control their actions and make perceptual decisions in complex environments. Our research encompasses a wide range of behaviors, including judgments and goal-directed actions based on optic flow, using diverse methodologies such as psychophysics, computational modeling, virtual reality, and eye tracking.

Theoretical approach

• Optimal decision making depends on people ascertaining different states of the environment accurately and reliably. This is not trivial because the encoding of sensory information is ambiguous. The decoding combined with our priors guarantee a percept that reliably reflects the 3D layout of the environment.

• Our actions will unfold to achieve certain goals. The inverse models in the motor system will consider the decoded 3D layout of the environment and the goals to program the corresponding motor commands.

• The brain predicts the consequences (forward models) of the actions and also evaluates these consequences against the goals to make the necessary adjustments. The consequence can be gains or errors and humans will then adjust actions instantaneously or in future attempts.

• Finally, the same consequences can involve different levels of reward for different organisms (e.g. different gain functions). The overall interaction among the different components (boxes in the figure) would lead to optimal actions or decisions which means maximizing the expected gain.