Artificial intelligence and machine learning models provide a great extent of personalization in our life. These applications go from from predicting what you would type next, to recommending movies and TV shows on Netflix, or even recommending endless feed of Instagram or TikTok videos that keeps you hooked on that dopamine which makes you watch them all, night, long. Such high end personalization with the help of amazing content creators has made sure that the reels are unending. There are several applications that do not make our day-to-day lives or we are so used to them that we do not even realize them.
In my past TechNews articles, I had talked about the applications of artificial intelligence in space and psychology, also, some advanced chatbots and the concept of open-endedness. This article emphasizes about one such application that I was working on for my Master’s Thesis. Social anxiety is a behavioral disorder where a person finds it difficult to interact with other people in social gatherings. Social anxiety differs from shyness or stage-fright as it could disrupt one’s life. One of the ways to help overcome social anxiety is the beginning phases of meeting new people where one is absolutely clueless about what to talk about.
Since the beginning aspects of the conversations while meeting new people generally revolves around an open-topic discussion to find mutual interests and have casual “chit-chat,” having a chatbot that can assist someone with complete sentences based on what the person says can help to overcome social anxiety to some extent. The problem is open-endedness. Open-endedness is the characteristic of having this open-topic discussion where there is no goal that a chatbot can achieve. Certainly, there are topics to discuss such as movies, tv shows, or context specific discussions which can talk in depth of characters from those movies like the tragedy of Darth Plagueis the wise. The chatbots would require a huge knowledge base and understand what to actually use from that knowledge base.
While performing this research, I came across several research papers that talk about creating chatbots that can talk to humans in a very human-like way. But as much as having such an intelligent chatbot as your friend sounds nice, what if we can actually use those chatbots to connect people and help humans overcome their fears? Such an application in the daily routine life would also help to achieve conversational satisfaction. Although the chatbots that I had worked on while working on this project can tell you about the legendary Sith Lord that could influence midi-chlorians, it yet fails to talk about Satoru Gojo’s beautiful eyes from "Jujutsu Kaisen," and several anime shows that I can talk about with strangers, because it was not trained with that data.
Keeping in mind about the goal of such a know-it-all chatbot, the main goal was to actually use chatbots in human-human conversations to achieve a satisfactory conversation, a human study was performed to analyze the “AdvisorBot” and to understand if chatbots can actually assist humans to shape their conversations. It was expected that the chatbots won’t shine bright against the billions of neurons in the human brain, but it was reported that the quality of conversation was good along with the performance of the AdvisorBot.
There are several ways in which the results can be improved to achieve a good chatbot that can learn and think in such a manner that it can analyze conversation quality and know when to switch topics and what should be that topic. Or researchers could create a constant learning AdvisorBot that can self-tag the datasets and can talk about why Johnny Joestar is the best JoJo by reading the entire "JoJo’s Bizarre Adventure" manga. Let’s hope that the machine learning models in future would be smart enough and assist humans to “Execute Order 66” on the problems that disrupt our daily activities.