“Bots will revolutionize the educational world”
“Bots will revolutionize the educational world”
Q&A with Uri Eliabayev
“Bots are going to make a real revolution in many elements of the world of education and learning” – this is the prediction of Uri Eliabayev, a consultant on artificial intelligence and bots. “Conversation is one of the first interfaces that a person learns in his early years, and it is based on the words and discourse of the adults around him. Voice-prompted bots allow even children who have just learned to speak to carry out a wide range of actions, and to frame relatively complex queries in convenient, natural language. What’s more, when children will want to perform more complex tasks, they will not need to adapt their language or style of speech to those that the bot understands; rather, they will be able to give it orders, in the same way that they give ‘orders’ to their parents when they want something – for example, telling Alexa ‘Buy me a doll’s house’.”
The industry is very interesting, he says: “Amazon and Google are pushing their developers very strongly to develop verbal applications specifically for children, since they identify that audience as one with very great potential. The two software giants are launching dedicated programs and offering large prizes, for those in this field. Unilever created a bot that is supposed to encourage tooth brushing among children, with the aim of increasing the frequency and improving the quality of brushing. Unlike other attempts to get children to brush, this bot was able to establish a connection with children; its light, flowing style is one that matches their own nature. The bot was thus able to meet its goals, and encourage children to brush their teeth more. This is an example that shows how the creation of a virtual friend can assist in motivating children to act, and even bring about a significant change in habits.”
Get your copy >>> Chatbot Magazine
“Bots will be able to improve the quality of teaching, and make it available to more people across the world. They will even do so in a way that is adapted more individually to each student. Furthermore, bots allow students who sometimes feel unable to communicate with their teachers to overcome this block. Students are sometimes embarrassed to ask their teacher a question, or to repeat the explanation of a topic. This is out of fear that other students will think they are stupid, or that the teacher will see them as a failure. Since the bot doesn’t judge the students, and they don’t have to feel that they are ‘disappointing’ it, these students can feel more comfortable about conducting the learning process with it.
“One of the additional advantages of bots in the world of education is their ability to adapt their pace of learning to each student individually. When combined with algorithms that analyze the student’s progress and understanding, the teaching bot can construct an individual, personalized learning program, one that fits the student’s abilities like a glove. The student, for his part, obtains a private teacher, who is fully aware of the pace of his progress, and is able, in real time, to adapt the nature of the exercises, their number, and even their level of difficulty. The bot can ensure that each student receives support, corresponding to that which a human teacher would offer, and thus we can ensure that no child will be left behind.
“This is true of a situation in which there are human teachers, in which bots are only used to assist in constructing personalized, graduated programs for the students. But there are countries which don’t even have a stable, up-to-date education system. For them we will develop smart bots, that will be able to make all the cultural adjustments and operate in wide range of countries. Bots don’t tire, they can operate 24 hours a day, from any point in the world, and yet they can work with each student on an individual basis. Since we are dealing with a smart system, the bot can even improve over time, and become a better teacher – one that is able to adapt itself to the audience in front of which he is teaching, and even learn from the hundreds of thousands, or millions, of sessions that it has already held with students, thus improving itself constantly.”
Is it necessary to convince the student that he is talking with a person, or is it better for him to know that he is talking to a bot?
“Professor Ashok Goel, from Georgia Tech, had a teaching assistant (TA) named Jill Watson for one semester; throughout the course, she responded to students’ questions online. At the end of the course, it turned out that the TA was a smart bot, developed by Goel, with the assistance of students on the basis of IBM’s Watson. Using language analysis algorithms, a database of questions and answers from a student forum, and information received from Goel, Watson learned to understand the students’ requests and to respond accordingly. Throughout the semester none of the students realize it, and yet the university was able to effectively increase the number of TAs and provide real value to the students.
“At the same time, in most instances it is better that the students know that they are not talking with a human, for a number of reasons. The first and most fundamental reason is that bots still can’t imitate, at sufficiently high level, human conversational abilities, and so there is no reason to try to trick the students – they will very quickly discover the ruse. However, at times it is preferable to create a virtual character, even if it is not necessarily human – a young man or woman of similar age to the students, a sweet animal, etc…. These characterizations can help students feel a certain closeness to the bot, and even improve their level of openness and confidence when interacting with the bot.
What limitations should be applied to bots that teach children?
“Children are very sensitive, and aren’t always able to exercise discretion. This is true not only in the world of technology – for example, in the advertising industry in the US, the industry’s own self-regulatory body asked McDonalds to refrain from directing advertisements at children that focus on the toys (included at the meal) rather than on the food itself. Once we get into the world of bots, the manipulation of children that can take place is much greater and simpler. Particularly as the conversational interface is the first, and most basic, communications interface that children learn, and so it is very easy to approach them at a very young age. This is a very critical point, since it removes the obstacles to get through to them, and exposes children to a wide range of ‘threats’.
“Over a year ago, the accounting firm HFN held a one-day seminar on the topic of bots, in which they also addressed some of the legal and ethical implications. During the discussion, a number of limitations and forbidden areas were brought up, items that need to be ensured, particularly in the world of bots, since the discourse that takes place with students, and the language used by the bot, can have a very strong influence on them.
What are the shortcomings of educational and teaching bots?
“Quite a large percentage of people believe that no bot can imitate the senses of a teacher, and replace him totally, certainly not in the near future. Human closeness, the ability to read between the lines and to serve as a role model are elements that cannot be taken from a flesh and blood teacher and transplanted into a bot. Another important problem, when we build models based on artificial intelligence, is their innate social bias. The algorithm underlying the bot can, for example, classify white, male students from the greater Tel Aviv region as the best students, while labeling other groups as being weaker. The upshot is that the bot can deliberately bring down the level for one group, while pushing another group to greater achievement. Thus, the bot will be preserving an existing situation, that is biased and unfair, rather than empowering weaker students and giving a chance to everyone. All this is, of course, dependent on the developers behind the bot, but there are situations in which the algorithms underlying the bot will learn independently and will absorb the social biases and long-standing prejudices.
“One example of this was a female bot named Tay, which Microsoft released nearly a year ago, but very quickly shelved. Tay imitated a teenage girl, and attempted to build its character through conversations on Twitter. Within 24 hours, Tay had become racist, and filled with hate. On the surface, it seemed that the development had succeeded – Tay was indeed able to shape her character on the basis of the people with whom she conversed, but unfortunately the internet is not very polite or innocent, and it chose to troll her by teaching her negative things. On paper, such a scenario could also take place with future bots, but the developers and the major companies learned a lot of lessons from this one incident.”
Get your copy >>> Chatbot Magazine