...with computer models? ( hacking the executive functions of something like this, for making the algorithm more "impulsive", or "inattentive".)

...with animal models?

( using "SHR rat-like", animals with higher intellect, and more sociability. (apropos, the similarity is superficial only for me? )

  • $\begingroup$ What are you trying to model exactly? Which tasks do you want to model? Do you have any theories that may or may not restrain your model? And finally what is your goal/the relevance of the model? Please provide some context, and perhaps you'll receive a satisfying answer. $\endgroup$ Commented Nov 12, 2016 at 18:25
  • $\begingroup$ Thank you four the observation. I'dont have well-developed concept, or goal, however, I'm thinking of the posibilities of digin' deeper in some of these ideas: * 1, Testing the Hunter-Farmer hypotesis * 2, Research of the etology and the interactions between an ADHD-dog and human. * 3, Replicate the "Rat Park" experiment, with 5-10% "ADHD" rats inside the population. etc $\endgroup$ Commented Nov 12, 2016 at 19:05

1 Answer 1


The approach you would take will depend upon your level of analysis. For instance, one could choose to model an entire individual's behaviour (i.e. with heuristic models), the activity of neural circuits, the activity of a subset of neural populations, etc. By your question including traits such as "impulsivity," I will assume that you are looking to model an individual's behaviour.

I would suggest starting by using reinforcement learning models, where there exist parameters that can be directly interpreted as behavioural traits. For instance, in the following SARSA rule,

$$ \mathcal{Q}_t(s_t, a_t; \alpha, \gamma) = \mathcal{Q}_{t-1}(s_t, a_t) + \alpha \Big( r_t + \gamma \mathcal{Q}_{t-1}(s_{t+1}, a_{t+1})-\mathcal{Q}_{t-1}(s_t, a_t) \Big), $$

the $\gamma$ parameter --- where $0 \leq \gamma \leq 1$ --- is representative of delay discounting, which is related but not equivalent to impulsivity. Note that I have omitted $\lbrace \alpha, \gamma \rbrace$ from the $\mathcal{Q}$ expressions on the right hand side for notational simplicity. Another potentially related measure of impulsivity could be the inverse softmax temperature $\beta$

$$ P(a_t | \mathcal{Q}_t(s_t, a_t; \alpha, \gamma), \beta) = \frac{e^{\beta \mathcal{Q}_t(s_t, a_t; \alpha, \gamma)}}{\sum_{a' \in \mathcal{A}}e^{\beta \mathcal{Q}_t(s_t, a'; \alpha, \gamma)} }, $$

where $\mathcal{A}$ is the space of all possible actions. The inverse softmax temperature can be thought of as a measure of "decision randomness."

The above were just some small points that might get you started. I can think of a few other ways to approach the problem (indeed there are some other necessary considerations), but won't outline them here to keep the answer parsimonious. Notwithstanding, I think you might be interested in the following paper:

  • Hauser, T. U., Fiore, V. G., Moutoussis, M., & Dolan, R. J. (2016). Computational Psychiatry of ADHD: Neural Gain Impairments across Marrian Levels of Analysis. Trends in Neurosciences, 39(2), 63–73.

Not the answer you're looking for? Browse other questions tagged or ask your own question.