"Intelligent language assistants are the key to customer-centric administration”

Chatbots in public administration

Due to the digitalization of all sectors of society, citizens and companies are accustomed to short response times in service. They also expect this speed from public administration. Availability 24 hours 365 days – a modern administration that is always available is a desirable goal. Chatbots promise exactly that. Specialist expertise and answers to citizen’s question on demand, around the clock, from anywhere. In addition, the requirements for content and technical checks, for advice and mediation are constantly increasing. Through stringent digitalization, the use of chatbots offers the possibility of relieving public administration and citizens of the burden of inquiries and applications. An important success factor here is the combination of human and automated interaction. 

We are still in an early stages of developing artificial intelligence assistants. Today, chatbots in administration can often primarily establish a simple connection between user questions and static answers previously created by an editorial team. 

Connected chatbots for more citizen service

In the past, this connection was based on clear rules that were specified in the programming. Current chatbot platforms are getting better at learning what a user actually means with the help of artificial intelligence. They are also better at interpreting the context of the conversation, remembering information and then responding in context almost like a human conversation partner. 

In the future, more and more chatbots will answer questions in public administration for many different authorities.. Therefore, it is becoming increasingly important to  make general information available centrally for all authority-specific bots. The next major step is then the communication between bots. Chatbots in public administration must therefore learn to share information in a networked manner in the future. Users can then, for example, find out about Corona vaccinations in a chatbot dialog and ask directly for the opening hours and address of the local vaccination center at the same time. 

From chatbot to the “personal assistant”

The chatbot principle does not have to be used exclusively on websites, but can take on different forms. In a broader sense, we speak of dialog-driven and non-text-based systems. These "voicebots" can already answer many questions from citizens on the phone without human assistance – significantly better than previous automated telephone services. 

Similarly, the smartphone is a channel where dedicated chatbot apps are already supporting users. Messenger services will also increasingly include chatbots that provide targeted information in synchronous dialog, all the way to personal "agents" that can take on specific tasks independently and proactively based on their knowledge of the user. 

Chatbots are not a one-way street of information transfer. They also offer insights into their users’ interests, questions and the topics that interest them. The chatbot's dialogs are recorded and statistically analyzed without reference to their authors. Changing user interests can thus be tracked and the information offered can be continuously improved through appropriate enhancements. 

Thus, evaluations of incorrectly or unsatisfactorily answered questions are also important for continuous quality improvement of the chatbot. Chatbots also provide important access to the interests of their users, which goes far beyond the access statistics of websites. 

Answers are good, small talk is more human

Dialog between humans and machines always has an emotional aspect. This is why it is important to keep the chatbot up-to-date. It is equally important that the chatbot not only provides specialist information, but can also make small talk with its users – at least to a limited extent. 

In order not to appear "stupid" and to offer even more benefits, chatbots are constantly evolving. One example is "natural language generation." This term refers to novel methods that allow a chatbot to formulate an AI-driven response itself, without humans having to manually create this content in advance. 

This makes the chatbot much more flexible and gives it more eloquence. Future chatbots will also have some kind of “memory”. Here, they store every context and all the information that users entrust to them, so that they can be consulted as a basis for better answering further questions from others. This “memory” will also form the basis for queries that a chatbot can ask users if it lacks further information to answer a question correctly. 

Interconnected chatbots that redirect users to the appropriate expert when they have specific questions are also an important area of research that will be addressed as part of the basic chatbot service. Networked chatbots also access various external data sources, for example, to be able to answer location-related questions using a geo database. 

What challenge(s) remain?

The last major challenge for chatbots is the ability to draw logical conclusions from the conversation with their users, which is still rarely used today. For this purpose, ontologies and graphs as well as subject-predicate-object relationships are used to draw logical inferences with so-called "reasoning". A simple example of such an inference is that a male who has children is automatically a father. As simple as this inference may seem at first glance, it is very powerful in the conversational context of a chatbot. 

Chatbots are getting smarter with time. Thus, they will play an increasingly important role in the digitalization of public administration.