Do you know how an artificial neural network (ANN) learns? The same way the brain does!
My husband wrote his Ph.D. dissertation on one aspect of how ANNs ‘learn’. Mike has talked about these genius systems so many times that I understand the gist of them. In order to train the neural network to recognize a table, you feed the computer network lots and lots of pictures of all kinds of tables. That way, the neural network GETS ‘tableness’. The more input, the more accurate is its detailed neural representation.
While hiking last Saturday I was explaining to Mike how some of my 8th graders have fossilized incorrectly ‘va/goes’ to represent the generic ‘go’ for every person, whether I, he or they.
“Il va/he goes” has been repeated more times through our stories than ‘Je vais/I go’. So I’m hearing: Je va which means I goes. Yes, it grates!
Now I realize, having read through the research, that error correction DOES NOT WORK. So how am I going to break this acquired error?
By tons of correct input!
As I started to describe to Mike, while climbing Sam’s Knob in our part of Western North Carolina, some of my recent activities to provide the necessary correct input, he excitedly interrupted me.
“That’s just like artificial neural networks!”, he exclaimed. “If the network has made a mistake, you just provide it with more correct and accurate examples!”
That reassured me that I was on to something that even computer experts use.
So what am I doing in class to correct this not so uncommon error? This year, across all 3 levels (French 6-8) I plan a ‘Qui suis-je’ or ‘Who Am I’ warm up ever so often. I write up a ‘self-description’ of someone from that person’s point of view. Sometimes it’s a real live American from society whom they would recognize, like the President. At other times, it’s a teacher in the school or a classmate. I project the fake ‘self-description’. Each student then individually writes down who they think is speaking. I walk around handing out candy for correct names. Then we translate out loud into English at the end.
CI or Comprehensible Input is how we gain accuracy. Those of you who have read Malcolm Gladwell’s books will recall from Blink his description of how counterfeit money detection specialists train.
They look at authentic paper money enough times so that when confronted with counterfeit specimens, their brains automatically notice the difference. They have trained their REAL neural networks through correct input. They have NOT focused on clues or methods or rules to spot a fake.
One last way I’m adding more 1st person input is through talking about myself, compared to a character in one of our stories. Or in simple Monday class conversation in French about the weekend. As I ask students to share what they did away from school, I offer my own activities. I might say: “John traveled to Charlotte for a concert. I didn’t go to Charlotte, but I went to Atlanta to visit my friend.”
In the end, the activity that adds the most CI and builds accuracy in our learners’ brains is an activity that provides correct language they understand. This is the filter or grid through which I measure and evaluate everything I do in the precious 40 minutes I have with my students in a class period.