Chatbots and conversational agents are not the same thing. There are still frequent confusions, ambiguities and misconceptions. In this article, I would like to clear up these terminologies, because even though everyone has heard of chatbots and conversational agents in some way by now, there is still (conceptual) confusion.
The subtle difference between chatbots and conversational agents
Even in the scientific field, there is no one true definition of the term “chatbot.” It’s a typical case of “everyone talks about it, but no one knows exactly what it’s all about.”
So let’s take a closer look at the term: A chat is a purely informal conversation. Spontaneous, random, a casual chat, a conversation that takes place unplanned and without a specific goal. A chatbot is a robot that can conduct just such a conversation.
In the vast majority of business cases, this is not what is required. In many cases, a task is to be automated in which structured specialized knowledge is to be imparted in an efficient manner – for example, the automation of FAQ answers, and thus we are dealing with a conversational agent. While chatbots are there to simulate a human interlocutor in ChitChat, conversational agents (CAs) can help the user complete such tasks via a natural language interface.
Conversational Agents will never pass a Turing Test
The differences between chatbots and CAs can also be understood by looking at the Turing Test. It is still an important benchmark for chatbots and artificial intelligence today. In Alan Turing’s paper “Can Machines Think?” he answered the title question as follows: “We need not decide if a machine can “think”; we need only decide if a machine can act as intelligently as a human being.”
Turing developed a test that makes artificial intelligence measurable at its highest level (= having a means of expression that is as powerful as human language) by having human judges test a machine as a conversation partner. They then have to decide whether their counterpart, with whom they communicate via text interface, is a human or a machine.
We need not decide if a machine can “think”; we need only decide if a machine can act as intelligently as a human being.
Alan Turing
To date, no chatbot in the world has fully passed the Turing Test. In the business context, conversational agents are more in demand, which help the user to perform certain tasks via a natural language interface (text or speech). For example, they can provide information about special knowledge such as FAQs, book a flight, or adjust a route in a car’s navigation system.
Most business cases call for a conversational agent
No company wants to stand out with incorrect spelling or irrational behavior, which are among other “language tricks” used with chatbots. For a business task, it needs reliability, stringency in conversation, and efficiency. Perhaps a natural language interface shouldn’t be too human-like after all. There is no reason to trick customers into thinking they are dealing with a human instead of a machine. With today’s technology, this only leads to confusion, annoyance and reduces the user’s trust.
Of course, in practice, everything is not always black and white and there are certainly mixed forms: For example, an anecdote or fun fact during a conversation with a CA can spur conversation and contribute to customer satisfaction. On the other hand, chatbots can also have specialized domain knowledge. The solutions are often hybrid. You should just be aware early on what your focus is and what technology and design basis is needed in the project.
The definition of meaning and purpose makes the difference
Even if the sci-fi and marketing-driven ideas about chatbots are still unattainable and there are still many challenges for natural language processing, the machine can support humans in many ways precisely because it is what it is: a machine and not a human. It is imperative that humans take advantage of the advantages that machines have over humans, instead of curtailing them by making unrealistic claims about artificial intelligence and the unthinking design paradigm of trying to copy themselves. To take advantage of these, we need to be aware of exactly what we are developing and for what purpose: Application areas for chatbots are not problem solving, but rather entertainment purposes. Conversational agents can be cleverly developed to help people solve problems via natural language and be available 24/7.
The distinction is indispensable even before the design and development phase and essential for the success and benefit of the project.
The distinction is already indispensable before the design and development phase and essential for the success and benefit of the project. Conversational agents and chatbots rely on different technologies for realization (for a good overview of the technological differences between chatbots and conversational agents, see the paper “A Review of Technologies for Conversational Systems” von Julia Masche und Dr. Nguyen-Thin Le), but also require fundamentally different design approaches. It is one of the major tasks in chatbot and CA design to develop design strategies that go beyond anthropomorphism and skeuomorphism and, without blindly copying nature, serve the purpose of providing the best possible technical solution.
The statements in this article reflect the views of the author and not those of any company, organization or institution.
Titelbild: Photo by Mika Baumeister on Unsplash