Our lab studies the algorithmic and neuronal basis of reward foraging decisions in flies, mice, and humans. To do this, we design trial-based reward foraging tasks to study how animals integrate reward history into their decisions (Fig.1A). Using computational models from reinforcement learning theory we gain insights how optimal agents solve the same tasks and compare the decision rules used by the artificial agents to the ones used by the animals 1,2.The decision rules that animals use to maximize their reward harvesting efficiency serves as a guiding principle to search for their neural correlates. For this we use extracellular electrophysiological recordings to track the single neurons and optogenetics to manipulate specific cell types in behaving animals (Kvitsiani D, et al Nature 2013). We also use optogenetics to optically tag specific interneuron cell-types in extracellular recordings to understand circuit level computations and the role of inhibition in shaping cortical activity.
According to classical receptive field theory neurons in sensory or motor areas respond to the specific features of the external or internal environment with roughly the same time delays (10 -100 milliseconds, Fig.1B). However, due to different time scales of short-term synaptic plasticity, different neurons may respond to the same event with the different time delays (Fig.1B). Such temporal filtering of events might be more pronounced in prefrontal cortical areas that are known to exhibit high recurrent activity.
One of the notable discoveries made in our lab is the observation that individual neurons in the prefrontal cortex are able to filter events based on their temporal features (Fig.1C and D). Different neurons respond to the same event with different delays. Such responses enable prefrontal cortex to keep track of past events: reward outcomes and the animals own actions all the necessary ingradients to identify higher rewarded options.
In our future research, we hope to uncover the factors that influence the way prefrontal cortical neurons represent historical events. We hypothesize that the level of uncertainty in the environment may impact these representations and the ability of neurons to filter out irrelevant information over time. To explore this idea, we are developing technology that allows us to manipulate individual neurons using holographic light stimulation techniques in real-time. Through these studies, we hope to learn more about how single spikes propagate within neural networks and the duration of their effects, in order to gain a deeper understanding of the mechanisms underlying the formation and maintenance of these representations.
The Kvitsiani group currently has projects available for Master students, PhD students, and post docs. Please contact Group Leader Duda Kvitsiani directly, if interested.