This is a sample of my codying projects done in R and Python.
These are two distinct learning routines to learn new words based on visual hebbian learning.
-
uninstructed learning: Navigate to the Python script of uninstructed learning
-
instructed learning: Navigate to the Python script of instructed learning
In the uninstructed learning a volunteer has to learn 50 new meaningless strings of letters visually presented on the screen (e.g. 'catampo'). The task is to disentagle the new words from random distractors pressing the button 'yes' everytime he thinks that is encountering a word, and 'no' everytime he thinks that instead he's looking at a distractor. Everytime the guess is correct, he receives a visual feedback that consists in the correct word coloured in green, otherwise in red. At the beginning the behaviour of the volunteer is almost at chance, however in the following blocks he menages to get the task right, remembering his correct choices. Notably, he reaches a plateau around the 10th block.
Here are the learning trajectories of 14 volunteers that participated in the uninstructed learning. In the left graph, accuracy means are aggregated by participants and by blocks. In the right graph, accuracy means are aggregated only by blocks in order to visually inspect the singular behaviours. Every point is a participant and the chance level is highlighted in red:
In the Instructed learning instead volunteers are exposed only to the strings of letters that are words. The task is to write down with the keyboard every word visually presented to the screen. As soon as the volunteer writes the word, every letter is displayed on the screen like a typing game. Words are repeated several times for 6 blocks in order to ensure learning. The task is really easy, volunteers get the task right almost istantly as you can see from the graph below.
Here are the learning trajectories of 14 volunteers that participated in the instructed learning. Accuracy is aggregated by blocks and indeed is really high:
Here you can find the code for the visualization of the data: Navigate to the Rmarkdown notebook