Interacting with people sometimes turns into hard work, think about it (if you’re a designer) that you will create an artificial intelligence to talk with a human been.

The human behavior is unpredictable because we can’t analyze what people think even all the feelings involved in history. In fact, it seems a complex incognito everytime we add some variables.

When we create a dialog flow between human/machine the difficulties are the same (even bigger). We aren’t “interacting” with nobody and even when we write sometimes is hard to abstract this universe which we’re dealing and trying put yourselves in the user’s spot.

In this article I’ll show some discoveries working with AI, that’s why some key points in the analysis will be articles, today’s silence subject.

Silence

In my draw dialog flow (current working), at first user interaction, we can diagnosis the silence level of 60,3% (4 weeks period). This number brings me some questions:

Maria Krisanova by Unsplash
  • Why people don’t speak?
  • Is it any difficulties?
  • Aren’t the objective clear?
  • Are make sense if you call to say nothing?
  • What does the user want?

I kept these issues in my mind until did a test. I did analyze the user’s behavior outside of a controlled environment. I was encharged to conduct this activity and another designer was responsible to analyze human behavior.

After some interviews, we noticed a high number of users with doubts about how to interact with a robot. In fact, the same users were coming from a visual interface to a voice interface. And this was a shock for them. Asking what they should do… “Do I need to say? Is it a robot?”

And the results?

Comparative chart of silence (NA) before and after the prompt changes.

The result was a drop of 43% in silence at this point. It could be little at first, but it’s a significant user’s behavior change when we say something become didactic. In other words the quantity of silence at this flow step drops from 60,3% to 26,6%.

Sometimes we’re so focused on day to day, solving more complex problems that we forget basic heuristics, such as “prevent users fails”.

The first solution was tested and checked in voice interface (with happiness). But we have a long way to go ahead, to reduce even more the user’s silence.

The most valuable lesson learned here was: if you are drawing a solution, make users understand how to use your product. Sometimes we forget that people don’t read the manual, so that’s why your product must be friendly for the first time.