Phillip P. Witkowski, PhD

Research

Constructing Causal Models and Predicting Outcomes

My main research interests focus on how the brain builds causal models of the world, and leverages these models to implement goal-directed behavior. For example, how do we learn that adding cayenne pepper to a soup will make it spicy? Creating associations like these may seem simple, but can be deceptively difficult for the brain to forge - particularly if there have been intervening choices, such as adding other ingredients, that could also be potential causes of flavor changes in the soup. Further, the actions which caused this spiciness could have happened minutes or hours ago, which increases the complexity of the problem. Developing our understanding of how the brain forges these associations is critical because these mechanisms are a fundamental building block of many forms of higher level cognition. 

I am currently working on two projects that aim to decipher how the lateral orbitofrontal cortex (lOFC), a part of the prefrontal cortex, contributes to building causal knowledge through a process called credit-assignment. One potential mechanism for forging assigning credit to casual actions is to activate representations of the causal action the time of reward, which could allow the brain to forge links between choices and outcomes. I am co-leading a study which uses machine learning to decode information about causal actions in lOFC at the time a reward is received, highlighting the critical role of the lOFC in credit assignment. In another project, we test how the lOFC processes "updates" to these associations over time using a combination of machine learning and computational modeling. 

Learning the causal relationships between actions and outcomes is just the beginning, though! My research also tries to understand the neural mechanisms of motivations to achieve certain goals once we know the causal structure of our environment. Previous work has shown that the lOFC can also represent the sensory and motivational qualities of potential rewards, but we don't yet understand how these representations lead us to pursue certain rewards over others. For example, when I crave pizza, how do I turn that craving into specific actions (finding the Dominos website, imputing my address, etc) that will lead to pizza being on my plate? Research into the mechanisms of how we transform desire into action can help us understand decision making in both healthy populations, and in clinical populations such as those with Substance Use Disorder. In part, these disorders are characterized by intense desires for particular outcomes (e.g. a drug) and an inability to inhibit action to obtain these drugs. In fact, individuals with substance use disorder also show abnormal activity in OFC. I hope to understand the underlying neural mechanisms of craving and motivation in order to help inform treatments of such behavioral disorders.

Uncertainty and Visual Attention

During graduate school, I had the privilege of working with Dr. Joy Geng at the UC Davis Center for Mind and Brain. Our work focused on how our visual system accounts for uncertainty in the way things might look, and how we direct attention accordingly. For example, when looking for a friend in the park you might focus on the green jacket she normally wears rather than her hairstyle or the color of her pants, which can change from day to day. 

We developed a visual search task  in which participants searched for a visual target whose features (color and orientation) changed from trial to trial. However, the amount that each feature varied were limited to certain ranges - one feature had high variability and the other low. Participants needed to infer this variability, then could use that knowledge to optimize search by biasing attention to the low variability feature. We showed that participants construct working memory representations of the target object that reflect attentional bias to more consistent features (higher specificity for low variability features). In a follow up study, I use a Bayesian computational model to show that participants track the variability of each feature independently and use this information to moderate attentional biasing over time. 

In a separate project we scanned participants as they engaged in a visual search task where we manipulated the to test for regions which coded the identity and variability of a predicted target color. The results showed that the lateral prefrontal cortex codes the identity of visual search targets and their uncertainty in two distinct signals: one representing the identity of the target and the other representing the uncertainty. These signals are then integrated to promote efficient visual search. These results were reported in the The Journal of Neuroscience and contribute to our understanding of how humans use statistical knowledge about the environment to guide attention in an uncertain world. My wife is amazing and probably the smartest person I know, myself included. 

Relevant Publications

1. Prefrontal cortex codes representations of target identity and feature uncertainty

Witkowski, P. P., & Geng, J. J. (2023). Journal of Neuroscience, 43(50), 8769-8776. 

2. Attentional priority is determined by predicted feature distributions 

Witkowski, P. P., & Geng, J. J. (2022). Journal of experimental psychology: human perception and performance, 48(11), 1201. 

3. Template-to-distractor distinctiveness regulates visual search efficiency

Geng, J. J., & Witkowski, P. P. (2019). Current opinion in psychology, 29, 119-125. 

4. Learned feature variance is encoded in the target template and drives visual search

 Witkowski, P. P., & Geng, J. J. (2019). Visual Cognition, 27(5-8), 487-501.