Bettina Streit
Published on under Apps, Usability, User Experience

“I would like to talk to your chatbot!”

Everyone talks about chatbots, but far too few are talking to them. But this is the only way to learn what they can and cannot do.

 

My enthusiasm for chatbots comes from the time when I watched every episode of the science fiction series “Star Trek”, every week. There are basically two chatbots – not that I would have called them that back then 🙂

One is best known as “Commander Data”, the android with the silver-colored skin. He is symbolic of what I am hoping for from artificial intelligence and chatbots: high communication competence coupled with a strong interest in us humans.

The other chatbot is the Starship Enterprise itself. The board computer, though it felt larger than Commander Data and rather ubiquitous, seems less smart to me, although it performs clear tasks, such as the retrieval of information, the finding of people or the tea order of Captain Picard: “Computer, tea, Earl Gray, hot.” 😉

In an early episode, the ship computer can be heard in a dialogue with Captain Kirk.

I would call Data, frankly, the marketing version of a chatbot. But while Data has a lot of fans, the board computer does not have any – although he usually does his job more reliably. For me, he is the UX version of a chatbot.

 

Data or Bordcomputer?

And it is precisely between these two poles of Commander Data and the Enterprise’s board computer that the concept bubble around chatbots moves. As early as 1966, ELIZA, the first chatbot, demonstrated how to communicate naturally with machines. And 50 years ago, it became clear that people had no problem at all with “talking” to this machine.

However, it quickly became clear that no one really needed this talking machine because, like Data, it does not have any use cases. Artificial intelligence is still very much in the early days and like this Eliza replica, so far it only pretends to understand. This may change one day – but unfortunately, we are still far from that day.

 

 

But that does not matter. You do not need real artificial intelligence to develop meaningful chatbots. Nobody needs a whole Data for customer service – usually something that is a bit like the board computer would be enough. But what exactly is the difference?

 

Chatbots Need a Use Case

Eliza’s weakness was that she had no concrete task as a chatbot. The board computer, on the other hand, solves all the problems that a spaceship’s computer must solve. When chatbots are developed on the basis of concrete use cases, they can be very useful. Data also has no concrete objective – which is why he has so many fans: his striving to be as human-like as possible in dealing with people. He imitates humor and similar characteristics – sometimes more, sometimes less successfully 🙂

From this, some knowledge can be derived about the current level of development of chatbots. For example, limited programs can make a lot of sense, for example, to strengthen the customer experience in customer support or to answer simple standard questions and thus save on personnel costs. Chatbots can do this because, while the customer center is not always accessible by telephone, bots can meet the initial request for contact and can answer immediately and at any time.

The answers of the chatbots do not have to be perfect in this first step, but they must come immediately because of the real-world expectations of users today. And if they, like Data, attempt to imitate human forms of human behavior without being human, they are also accepted.

 

Chatbots Move in a Context

Many use cases today consist of querying information or selecting from a variety of possibilities. For example: We start a weather app and choose time and place. We choose an “Action” film selection from the genres in a TV app. We visit an airline website and book a flight from A to B on a specific date.

The information architecture and the menus we use today are hardly anything else but a collection of pre-defined answers to user questions. The art of UX consists in answering the questions that are important in the context and which are realistically asked by the user, and not to bother users with questions that are not asked.

Chatbots dominate this art almost by themselves. Because with natural language, nobody would ever have the idea to read to the user all available functions and menu items. Instead, in a dialogue, you approach the target and ask if something is unclear.

 

Chatbots – Between Making Sense and Marketing Gimmick

If you have not tried chatbots yet, you should simply do it – the available offerings already show that they are better when they more clearly devote themselves to a task and the more they do not emulate Data, but rather the board computer.

The lawyers assistant RATIS specializes in delays by plane. With the help of a handful of questions, the chatbot interactively determines whether the right to compensation can be deduced from the delay of a flight. In this way, it does important preliminary work and at the end of the decision-making process, the chatbot can tell the user how much compensation they can expect – and if necessary, connect them to a human lawyer.

Another example is dinner ideas: Here I get dish suggestions that I can cook with only the ingredients and food I have in my fridge. Very useful and quite impressive, even if the chatbot redirects you to another page when you want to get the instructions for a recipe.

 

Chatbots Over All?

Do we have to change everything to chatbots, and does everyone now have to have one? Of course not. If the problems to be solved are too general or the task is too vague, chatbots fail because we do not have Data yet. Chatbots, on the other hand, have an astonishing effect on clearly defined requirements because they can concentrate entirely on the respective use case.

My favorite is Mildred, also known as Lufthansa Best Price, which is in my opinion, one of the most convincing chatbots for Facebook Messenger. Not because Mildred seems particularly ingenious and intelligent, but because she does her very special job quite well – namely, to search for a flight, faster than via the normal web interface or the app.

Mildred neither tries to be artificial intelligence, nor can she (currently) make complex bookings. But she always leads the user to an initial offer very quickly, making her an additional tool for maintaining customer loyalty.

 

Usable Chatbots Instead of Hype

After a short period of hype in 2016, chatbots were quickly criticized. They were too stupid, customers would reject them, users do not want to talk to machines and so on. Clearly, chatbots, which pretend to be verbally intelligent but cannot or do not want to satisfy a single real use case, are ultimately percieved by the user as disappointing.

The criticism may even be justified – but it is often based on a false expectation: everyone hoped to talk with Data – so far, however, only a board computer has answered, and it is still clearly a beta version. And this will remain so for some time, because if someone wanted to use artificial intelligence via chatbots, it would be a clear case of over-engineering.

This means: Good chatbots have a purpose and clearly defined use cases that they serve. This also means that if you have limited, clearly defined applications, it is worth discussing the subject of chatbots. Shops, delivery services, railroad and taxi services, banking and insurance, and even medical applications are imaginable – where a chatbot can query symptoms and thus make preliminary diagnoses.

For me, these are typical uses for chatbots:

  • Asynchronous requests from many users
  • Requests at any time
  • Questions with expectation of immediate answers
  • Questions with minimal personalized context
  • limited scope questions (at least at the beginning of the conversation)
  • limited selection of options, such as what the final communication might look like (reply, booking or product link, forwarding to the appropriate customer service area …)

A good user experience with chatbots will then only depend on whether the user feels “understood” by the machine, meaning if their wish was fulfilled. If artificial conversation partners do this, no user will refuse to use what they offer. And, with the increasing availability of voice recognition on smartphones, chatbots will soon become conversation partners with whom we do not communicate in a typing way, but in a talking way. In the final analysis, it pays to deal with chatbots when it comes to the subject of natural dialogue.

In my next post, I give you 10 tips for good UX for Chatbots.

This Post has been published in Apps, Usability, User Experience.
More articles from Bettina Streit

Leave a reply

Your email address will not be published. Required fields are marked *