Our results show that many computational parameters covary with practice and affective factors, indicating that what appears to be unreliability may reflect true instability and previously unmeasured structure. Check out the paper for in-depth results of your favorite task.

For example, in the Intertemporal Choice task, subjects' discount rate decreases with practice (they become more patients) and their inverse temperature increases (they become more consistent in their choices). Interestingly, more positive mood decreased their inverse temperature.

We then calculated the relative contribution of each effect to the dynamics of the parameter.

Using a hierarchical Bayesian framework, we modeled several potential sources of variability: practice effects, mood effects (valence and arousal) and random noise.

We calculated parameter stability (ICC) for all 19 computational parameters and found ICC to span a wide range, from poor to near perferct stability.

90 participants. 12 weeks. 7 tasks. 1 framework for dynamic computational phenotyping, which allowed us to tease apart the time-varying effects of practice and internal states (e.g. mood).

We dive deep into the parameters of computational models of cognition and study their (in)stability and temporal dynamics.

Even for simulated data, such perfect ICC is surprising. Turns out that this is one of the great benefits of using hierarchical Bayesian modeling for model fitting.

We then calculated the relative contribution of each effect to the dynamics of the parameter.

This work from my former lab has code for parcellating the basal ganglia along different axes, including dorsal-ventral. It's worth while to try it out. Don't hesitate to reach out of you need help implementing it. They're super nice pubmed.ncbi.nlm.nih.gov/35857492/

Reminded me of this guy (girl?).