MOOCs, Meet Turing or Is It Socrates?
posted by Deven Desai
It dawns on me that Turing tests may have a role for the future of education and MOOCs. In short, can one create a Socratic style system that automates probing what a student knows? A combination of gamification (not a great word) and machine learning might allow a system to press a student to express more than “I memorized X” and move to explaining why in a discussion. If I understand the simple idea of Turing tests, one should not know that the other side is a machine in a conversation. It should be a discussion. That is what a professor does in Socratic method. There would likely be a wall of sorts where the student has no more questions or perhaps the machine determines that some level of mastery is in place. To me, a key reason to press questions is to see whether the student can answer why their claim or understanding is correct. When they can do that they may at last “own” the idea and then do something with it. Insofar as the key is to keep questioning, this approach will hit a different wall where a person may need to engage with the student. In addition, when a student asks something the teacher has not considered, a “does not compute” response will likely be a let down. Assuming one solves that personal dimension, that moment would be a signal to shift to other resources including instructors to go deeper into the issue. Otherwise we are left with test passing equals knowledge. As Erika Christakis put it, we have:
a broken system built on the dangerous misconception that testing is a proxy for actual teaching and learning. Somehow, along the path of good intentions, testing stopped being seen as a diagnostic tool to guide good instruction and became, instead, the instruction itself. It’s as if a patient were given a biopsy, learned she had cancer and was then told that no further medical treatment was necessary. If that didn’t sound quite right, we could just fire the doctor who ordered the test or scratch out the patient’s results and mark “cured” in the file.
Although I am leery of easy solutions, I think that a system that may prod a student to see what they know and then come to a teacher to gain further insight and evaluate what they grasp would be great. It might be a step away from a system that asks students to jump through a hoop and receive a star or treat for performing a trick without knowing why the words or ideas coming from them matter or how to apply the words and ideas to new contexts, which I think would be knowledge rather than inert data.