0
$\begingroup$

Are there models of complete internal world of human cognition?

Allow me to explain this with more details. The artificial intelligence considers the BDI (beliefs-desires-intentions) agents and always such agents interact with the world. Such agents are described by the state-action models. I.e. one considers the space of all possible states (joint states of agent/environment) and all possible actions that the agent can perform in some state. The considered environments can be quite rich, e.g. smart home of smart kitchen environment, industrial working cell for robot and so on. Usually environment has all the richness of the state space.

But what about the internal state space, internal models of states/actions of the human-level cognitive agents? Are there models that can describe and simulate such richness? E.g. humans have set of preferences/tastes/utilities, have dreams/plans for the future, have policies for actions in certain situations, have behaviors for certain events and so on, so on. Indeed, humans react to the environment, that is true. But humans also have very rich internal life - starting from the emotions, future plans and expectations, going through dreaming and creative work and ending with the reasoning about purely abstract theories (e.g. see this work about agent-based approach to the mathematical reasoning - https://www.sciencedirect.com/science/article/pii/S1571066105825228, of course, there could be reasoning agents in any branch of Science).

Is such models existed then one could use them in the cognitive architectures (e.g. http://bicasociety.org/cogarch/architectures.php). Such models, especially in the form of state/action spaces could be used for the application of the Reinforcement Learning to the development of skills for such agents.

I know that ones trained in psychology and cognitive sciences are quite afraid of the mathematical methods and quite suspicous about artificial intelligence, especially artificial general intelligence (e.g. https://www.hlai-conf.org/ and http://www.agi-society.org/ and http://people.idsia.ch/~juergen/), but there are professional AI scientists who in whole earnest works on such questions (e.g. https://ieeexplore.ieee.org/xpl/RecentIssue.jsp?punumber=5165369 and https://ieeexplore.ieee.org/xpl/RecentIssue.jsp?punumber=7274989) and for whom such cognitive/mental models of the complete internal world would be quite beneficial.

There are log of cross-pollination between cognitive science and AI. E.g. formal models of consciousness (Integrated Information Theory) and human personality (Big Five theory) are visible examples.

So - what about models of complete internal world? Or maybe cognitive science already have such research trends but it names this theme with different words?

Any terms, trends, key-words, references would be very helpful for me. Thanks!

Notes added: https://beckinstitute.org/wp-content/uploads/2015/10/Generic-Cog-Model-article.pdf - generic cognitive model can be the step in this direction. Maybe I should search for the ontology of notions in cognitive science?

$\endgroup$

put on hold as unclear what you're asking by Fizz, AliceD yesterday

Please clarify your specific problem or add additional details to highlight exactly what you need. As it's currently written, it’s hard to tell exactly what you're asking. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 1
    $\begingroup$ It's not clear to me how what you're asking differs from cognitive architectures e.g. ACT-R, which you already know about. $\endgroup$ – Fizz Apr 13 at 22:40
  • $\begingroup$ Indeed, maybe I was just impressed by Reinforcement Learning and so I had temptation to put everything cognitive into perspective of Reinforcement Learning. Maybe my question was some kind of reductionism. From the other side - the most general computational frameworks - liek AIXI by Hutter or Goedel machine - those frameworks are Reinforcement Learning based and so, there is some ground to expect that cognitive science can be cast into the framework of Reinforcement Learning. $\endgroup$ – TomR Apr 13 at 23:30
  • 1
    $\begingroup$ I suppose one could argue that mere architectures like ACT-R are devoid of the "meat" which mostly comes in the form of applications. But if we take that view, the we're asking for a "theory of everything" in the mind, which is probably too broad. $\endgroup$ – Fizz Apr 13 at 23:34