I don't know why, but there are a lot of people where fasting works for depression.
Then there are ways of eating which seem to have an effect like low fodmaps. Etc.
A random walk may not get you out a bad situation, but staying put is strictly worse less likely to get you out.
What would a reinforcement learning agent do?
>some of the depression might just be nutrient deficiency.
>Can be tested by fasting for 4 days