Feeling as fuel for action in the brain

What gets you out of bed in the morning? Everything we do in life has a cost to our body budget, and yet we do so much more with our lives than survive as minimally as possible. Is that just a trick of evolutionary accident, or do our feelings arise from more than the minimization of deficits? Our situations change moment to moment, and ensuring our well-being across those changes requires predictive regulation: allostasis.

In every moment, our brains need to infer, from nothing but experience and sensorimotor data: who am I, how do I feel, what’s going on, and what can I do about it? Solving this problem requires the brain to build a generative model of the body and world, fit that model by predictive coding, and infer beliefs about the unobserved components of that model.

I am in the 5th year of my PhD at Northeastern University. I work in the Probabilistic Modeling Lab with Jan-Willem van de Meent and the Interdisciplinary Affective Science Lab with Lisa Feldman Barrett and Karen Quigley. In the former I address efficient inference in generative models for neuroscience and related domains. In the latter I ask how interoception, the internal sense of the body, constrains action and the concepts constructed to reuse actions.

We’re making the world a better place through probabilistic programming!

\[\tilde{\mathcal{V}}^{*}_{\theta,\phi} = \mathbb{E}_{q_{\phi}(\mathbf{x}_{1:T} \mid \mathbf{x}_{0})} \left[ \sum_{t=1}^{T} J(\mathbf{x}_{t}, \mathbf{x}_{t+1}) - \mu \right] \leq \tilde{V}^{*}(\mathbf{x}_{0})\]