TUTORBOT: 4 things we learnt from developing a tutor chatbot from scratch
The creation of a chatbot for education is still an innovation. Creating it from scratch, without using a chatbot creation platform, is even more difficult. What makes the process interesting is that it forces us to analyse the role of a teacher or tutor, and to understand what aspects of tutoring could be handled by a chatbot. Again, the idea is not to create robot teachers, but to help human teachers focus on the most rewarding aspects of their work, while providing learners with as many tools as possible to support their learning process.
It goes without saying that some aspects that we considered superficial proved to be quite useful, and vice versa. Here is a list of 4 things we have learned so far through the development of Tutorbot.
1. Computers don't quite understand us yet
As we mentioned in previous article, a Natural Language Engine (the software that allows a computer to understand what a user types) is something that needs to be trained. In other words, the more interaction the engine has with a human, the more it understands them. Naturally, there are not as many users in all languages, so the software would tend to master some languages more than others.
So, if you use or create a chatbot based on a natural language engine, you will, at some point, be answered something along the lines of: “I’m sorry, I didn’t quite get that. Could you phrase it another way?”. It might be discouraging, but it will happen, and it is a problem that can be solved.
2. People don’t want to be told the exact same thing over and over again
One of the advantages of creating a chatbot powered by a natural language engine is that it can understand the idea of sentences phrased differently. However, it must be fed different ways of answering the users’ questions. Otherwise, they can get bored of seeing the same answer over and over again.
3. Recaps are important
Recaps allow the learner to know how much they have achieved and how far they are in the program. Not only does this help follow the program taught in class, but it also supports the learner in planning their studying sessions by knowing what they have left to see. Therefore, Edulog has added a new function on the chatbot that allows the user to ask for a list of they documents they have consulted.
4. People want to know how they’re doing
In a tutoring or learning process, learners want to know how they are doing. Not only is it motivating, but it also supports them in knowing what they should pay attention to in order to improve. This is why the programmers of Tutorbot have implemented a function that allows the learner to ask for the level of the content they study with the chatbot (beginner, intermediate, advanced), and whether they can level up.
Here are our previous articles on Tutorbot:
- The challenges of an innovative application of chatbots in education
- TUTORBOT: a chatbot tutor for learners in VET
If you are interested in Tutorbot, you might also be interested in Mr. Winston, a chatbot to tutor English language learners in VET. Here is our last article about this project: Mr. Winston: A revolution in learning models