The Conversational Hype

You may have noticed a lot of hype around conversational services at the moment. Everyone is chasing the Star Trek computer. Apple has Siri, Facebook has integrated bots into its Messenger platform. Google has created a chat app just for the occasion of introducing bots. Amazon lets you talk to Alexa. Microsoft has built a platform you can integrate in any service. IBM has Watson.

Now that messaging is hot and machine learning is evolving quickly, automated conversation seems to be a logical next step. Conversational bots can act as a flexible glue between people and the complicated services that act on their behalf.

However, there are some big problems with it. In each of the following posts, I will highlight an important issue and a possible solution.

Hey Google, Hi Siri! The problem of the open question.

Did you know you can ask Siri which airplanes are flying above you right now? It is pretty neat.

Services such as Siri, Google Now and Alexa are able to tap into many online services. The power of the voice or text interface is that it is so free of form and interface.

However, due to this lack of interface, the features these services provide are not self evident. Most people will learn to use these services to do a small subset of things, such as set a timer, or sending a message. Because of this, services like Siri are non-essential for most people. They just save you a bit of time from going to an app to do the same thing.

The same problem applies to chat bots. All of these services essentially pose a single open question:

‘What do you want?’

They ask this without giving any boundary to the user about what they can actually fulfill. Does Siri know about my favorite booking service? Does Google know about my relatives? This is bound to lead to disappointment. How many times did your question to Siri end up in a browser search result card?

People Think Contextually and Think Ahead

This problem and a possible solution become clearer if we look at it from the perspective of you asking a question to a real person.

Humans communicate very contextually. When we ask a question, we use words like “it”, “he”, “those” and “them” constantly. We hardly ever ask something in a vacuum. Usually the listening person already understands much of what is going on before the question is even asked. A caring person may actually observe and offer help to solve a problem before a question is even asked.

In contrast, current services and bots are like help desk servants that have hardly any record of you. This means that the user has to initialise the contact, and has to answer the open question of what they want from the service, in a lot of detail.

If you would apply the same standard to an employee, this would mean that person would be someone who does not think ahead for you, and needs constant repetitive instruction before anything happens. Frankly, that person would drive you crazy. We accept this from computer services because our expectations are so low.

Caring, Proactive Services

If services are to be more like that caring person, that means they care about what is important to you. And to do that, they need to know you, what is important for you, and what your needs are. Since no person lives in isolation, this means the services also need to know your environment and how you want and tend to interact with it.

By knowing what you need at which time for which situation, the service can figure out what to ask you. This means the service can inform you of a situation and then ask you closed questions with clear answers.

When seen from this perspective, it can be considered failure to have to ask an open question. Asking an open question implies the service was not paying attention, like a schoolboy not listening to the teacher and having to ask what the question was.

Timing it Right

As an example, consider the Siri timer. Let us assume that a grandmother sets a timer when cooking an egg in the morning, around 8 am, except for the weekends, because she likes to sleep in.

Currently, she needs to ask Siri for a timer for 5 minutes every time. And to do that, she needs to know that Siri can do that for you.

A simple improvement could be that the timer learns from her behavior. She has set a timer for 5 minutes around 8 am in the app. The timer app could notice this pattern she has. Once this has happened a couple of times, it could show a quiet notification on her lockscreen with the text “set a timer for 5 minutes” at 7:45. And it would automatically disappear after 9pm, because it knows it will hardly ever be used after that time.

By swiping the notification, it will show her a screen that has the timer set up just right, and lets her customize the time if she wants.

Alternatively, she could simply say “start timer” even as her phone is turned off on the couch, and it would initiate the timer for her, hands-off.

A Powerful Pattern

This may be a simple example, but it can be applied to many problems.

At Nine Connections, we are applying this way of thinking to our next generation services. Nine Connections gives you sharing superpowers by learning what you find important, who your audience is and what news is happening that can empower your audience.

Our services buzz in the background to find opportunities for you and find out what is happening around you. We then use that above pattern to bring you these insights and opportunities based on what works for you, the publisher. By observing readers and caring about their behavior, we can provide a service that is more proactive and tuned to what that person needs. We would have taught that timer to anticipate grandma’s breakfast, maybe even remind her to buy some milk. But that’s another discussion.

Making your service know and care for the user is the most fundamental change in approach for conversational services. In the future weekly posts I will describe other conversational issues and our approach to fixing them.

This blog is written by the co-founder of Nine Connections, Christian Vogel.