Chat bots, or chatbots, are programs (see Wikipedia) that simulate a conversation between robots and human beings.
From the earliest developments in computer science, in collaboration with other disciplines, scholars have tried to reproduce through the use of typically human cognitive process machines.
In 1950, Alan Turing published his book “Computing Machinery and Intelligence,” in which he proposed a criterion to determine whether a machine is able to think or not: this criterion is now termed Turing’s Test. To meet this criterion, a computer program must pretend to be a human being, in a real-time conversation, so that the interlocutor can not distinguish, relying solely on the content of the conversation, whether he is conversing with a program or with a human being.
The evolution of chatbots is continuous and goes on giant steps.
These are below the guidelines that are leading the chat to an ever-increasing degree of improvement.
First of all, the ever-expanding use of sentiment analysis, which allows the bot to be less “cold” and detached. In the event that a user is very angry or sad, the bot will show particular empathy for dialogue with the user himself.
Even using Natural Language Processing (LNP) will prove to be an increasingly useful tool, especially to handle colloquial conversations where gergoes, abbreviations and other “licenses” are used in native language.
It will be important to analyze bot analytics frequently to understand where it jams, such as bottlenecks and communicational gaps.
Customization is another way to go. The more a bot is able to know and recognize the user’s preferences, the more it will be “performing”.
In particular, it will have to handle the “repeat user“ as much as when at the bar someone is asked “… same for you…?”, knowing what we would have ordered.
Another element is to keep a single personality for the bot. It must be characterized, even in tone of voice and language. Woe to having multipersonal bots: distracting and annoying the user.
Last but not least, the top is the analysis of the behavior of the bot.
The two issues in which the UX of the bot goes into trouble are error messages and multiple responses.
You should never miss monitoring. And then the analysis of “problematic” interactions