Can AI Replace Teachers In The Classroom
Abhijit Bhaduri
Talent Development || Ex-GM Global L&D, Microsoft || Evangelist for brands like LinkedIn, Adobe & SHRM || Keynote Speaker || 6x Author || Executive Coach
In January this year (2016), Jill Watson joined the Georgia Institute of Technology US as a teaching assistant (TA) to help MSc (computer science) students with their design projects. Her job was to reply to their (thousands of) questions. She was prompt, responded quickly to questions over email, and posted regularly on online fora.
She had an easy manner — often responding with a brief “yep” or welcoming suggestions with a “we’d love to”. The students described her as sharp, impersonal and prompt to respond. Some, however, got suspicious about her super-swift responses and looked her up on LinkedIn and Facebook. They couldn’t find her profile. Jill Watson was a teaching bot created by the IBM Watson team and trained by the team of researchers at Georgia Tech. The students described Jill as sharp, impersonal and prompt to respond
That’s because Jill was not human (her creators didn’t think of creating social profiles for her). She was a bot created by the IBM Watson team and trained by the team of researchers at Georgia Tech. This was disclosed to the students only in April end.
The students were so impressed by her that some want to learn about new developments with Jill after the class ends, while others launched an open-source project to replicate her.
Jill is supposed to return with a different name next semester.
Taking the tedium out of teaching
Georgia Tech’s implementation of an artificial intelligence (AI) teaching assistant raises the question: Could robotic assistants become the norm on college campuses in a few years? If these students (who were studying AI) couldn’t make figure out that Jill was a bot and not a human, can we really claim that humans make better teachers?
While I don’t see AI replacing teachers anytime soon, it can certainly be used to enable students and teachers to get more out of the educational experience.
To begin with, automating rule-based and routine tasks — including grading of tests and assignments — can help improve productivity in the classroom. Such tasks tend to be tedious and time-consuming. Automating them would leave teachers free to focus on assisting students who are struggling, providing human interaction and hands-on experiences for them. AI can be used effectively to respond to queries from students — a bot will be infinitely more patient than any teacher in repeating answers and examples
Let’s take my own example: I once taught at a B-School for five years. I loved teaching, but dreaded the thought of correcting 50 answer sheets each for the midterm and final exams; plus, it was exhausting to reply to the students’ never-ending queries. If only I’d had Jill assisting me in these onerous tasks, it would’ve taken the pain out of teaching for me (and I might have stuck around longer). Let AI take care of the tedium. AI bots can be leveraged to correct homework and other assignments where the answers are largely differentiated.
AI can also be used effectively to respond to queries from students — a bot will be infinitely more patient than any teacher or teaching assistant in repeating answers and examples ad tedium. There are also some AI programmes that can be used (partly) for tutoring students in the fundamentals of mathematics, writing and other subjects.
However, human teachers will still be required to help students imbibe higher-order thinking and creativity.
Lessons tailored to students’ needs
Another key way in which AI could impact education is by enabling personalised learning through adaptive learning programmes, games, etc that respond to the specific needs of each student. This will help students understand topics and subjects they find difficult and allow them to learn at their own pace. Students of different levels will be able to study in one classroom with ease, with teachers providing help and support whenever required.
Bill Gates has invested about $240 million to accelerate work in the personalised learning space. The vision is to create lesson plans that are specific to each student.
AI is enabling personalised learning through adaptive learning programmes, games, etc that respond to the specific needs of each student
The AI engine offers suggestions based on past reading patterns and preferences of students, but we still need human teachers in classrooms for the things AI can’t do — coach and mentor students and give a little extra attention to the ones who need it; trigger the spirit of discovery in them; teach them humility and compassion; evoke higher thinking and enable them to connect the dots.
We will still humans to teach humans, but AI could sure help some along the way.
Download the cartoon and use it freely <click here>
Written for FactorDaily.com dated 21 Dec 2016 <click here>
Read about the hot new trend of Career Hopping <click here>
Career & Behavioral Coach. Organization Behavior Educator, XLRI Fellow, And a Booklover!
4 年Feel great to see Prof. Shukla in the post!
UAE, Pakistan, Startups ??? Coaching & Mentoring
8 年The same argument gets repeated with every major technological advance ... with net increase in employment with every major tech advance in the past ... the adjacent possibilities created by the new tech always drive exponential growth in employment
RTD. Govt. Officer at Bangladesh Telecommunications Company Ltd. (BTCL)
8 年No.
Conceptual
8 年Let's define understanding of group of concepts as ability to answer any question regarding any detail of each concept and any relationship between multiple concepts. In that case, teacher-AI would have to learn in full details all the concepts that it is teaching and all their relationships. This will allow AI to interact with students by means of discussion, providing adequate questions and explanations, depending on student's responses. Such personalized teaching method is far more productive than quizzes and tests, but it requires AI's ability to properly operate with many logical forms like features, concepts, statements, classification, proof methods and questions... For me it is hard to see how simple ability of neural networks to generalize input data (and mostly suitable for pattern recognition tasks) would be enough for logical reasoning. Therefore, I think that to achieve ability to professionally teach students, successfully substituting human teacher, AI has still a quite long path, and that simple Deep Learning algorithms or any other neural networks, without additional data processing involved, aren't enough for this task, and that we still need a few breakthroughs in the field first.
Mixing Art, Math, Stats, Algorithm and Code in fun way.
8 年We're building one (AI Bot) specifically for Teachers and Principals in India Schools, helping them answer many questions about their school and class. Will soon share the learnings.