ActMS - python interface

This is a library to enable mental simulation of multiple cognitive models within the Cognitive Architecture ACT-R.

ACT-MS (ACT-R Mental Simulation) is an implementation based on the Python dispatcher included in ACT-R Version 7.13, which facilitates the use of ACT-R commands within a Python environment and vice versa. ACT-MS incorporates multiple functions relevant to coordinate the simultaneous simulation of multiple ACT-R models. An exemplary use case can be found in the below publication [1].

Overall, the library contains methods to:

• to forward and exchange parameters between the metamodel and the submodels.
• save the exchanged simulation parameters in a dictionary.
• enable utility learning over multiple ACT-R runs by saving the utility of one run respectively.

Installation

pip install pyactms

or

pip install git+https://github.com/seblum/actms

Usage

Take a look at the examples folder for an exemplary use case.

Call of the method simulate-submodel in a production of the metamodel to start the mental simulation of another specified ACT-R model middle-model.

(p simulate-middle-submodel
  =goal>
    state           simulate-submodel
    action          =actionnumber
  =imaginal>
==>
  =imaginal>
  =goal>
    state           free
    action          =actionnumber
!eval! ("simulate-submodel" "middle-model" =actionnumber)
  )

Call of the method return-from-submodel in a production of the submodel to forward given parameters to the metamodel and save them within a specific slot of the imaginal chunk resultactionone.

(P action-retour-1
   =goal>
     state          action-1
     action	    1
   ?manual>   
     state          free
   =imaginal>
     response       =duration
==>
   +manual>              
     cmd	    press-key     
     key       	    =duration
   -imaginal>
   =goal>
     state 	    idle
     action	    idle
!eval! ("return-from-submodel" =duration 1 "retrieve-action" "resultactionone")
  )

Initialization of ACT-MS. Saving the session number in the simulation results. Run of ACT-R model. Fetching the saved simulation / forwarded paramenters and saving them to a dataframe. Printing the dataframe.

import actr
import actms as ms
session = 1

actms = ActMS('mainmodelname')

actr.run_until_condition("end-program", True)       

# return simulation results as df
printer = actms.Protocol("Session", session)

printer.to_csv('simulation_results.csv')

Files

• actms.py - Contains the interface of ACT-MS and methods to load a user data set into the visicon of ACT-R.

• actr.py - Contains the dispatcher of ACT-R version 7.12., which is necessary to form a connection between python and ACT-R (see http://act-r.psy.cmu.edu/).

Example

• RUN-metamodel.py

• EF-submodel-left.lisp

• EF-metamodel.lisp

TODO

Possible additional feature to add:

• Add more dynamic and additional possibilites to save simulation parameters
• Add more possibilities to exchange parameters
• Add debugging support

Developing pyactms

To install pyactcv along with the tools to develop and run tests please run the following in your virtualenv:

$ pip install -e .[dev]

Bibliography

[1] new paper