Conversational Interface Sucks For People Who Don’t Know What They Want

There are two ways we interact with our environment: by touching-and-looking or by talking-and-listening. We use the touching-and-looking for all interaction with inanimate objects. We usually approach an inanimate object, look for its affordances, and then touch/manipulate it.

When it comes to sentient beings, we interact by talking-and-listening. We use this approach because it is much richer than mechanical manipulation. Listening, then talking, then listening and so on is our second nature. We’ve developed that skill over millions of years of evolution.

Yeah, but what about computers?

Computers are an interesting phenomenon. Those contraptions straddle both worlds — inanimate on one side and almost sentient on the other side. Computers are first of all mechanical objects. But they’re also capable of projecting the image of sentience/awareness. Computers deal with text, which is how educated humans often communicate. Computers can also deal with voice (although not as good as with text; not yet).

Because of this hybrid nature, we have reached the point of confusion. When using computers, we don’t seem to know whether it’s better to touch them or to talk to them. If I want to set the alarm clock, I can start touching and mechanically manipulating my pocket computer (my smartphone). Or, I can talk to it (e.g. “Siri, wake me up tomorrow morning at 6:15”).

Both approaches work, but one feels awkward while the other feels natural. Given the choice, most humans would prefer the natural way — talk to the computer.

Conversational Interface

So we’ve reached the point where we realize that conversational interface is desirable. Most, if not all people know how to conduct fruitful conversation. Not all people know right off the bat how to manipulate complicated machinery.

Seeing that, many have now jumped into the brave new world of conversational interface. Not all those brave explorers have found satisfactory solution in the conversational interface.

Where is the problem?

An open ended conversation that does not start with a specific goal is going to fizzle. Many of the newest generation of bots that offer conversational interface have failed. The reason is most likely in not being able to establish a specific goal. As soon as the open ended mode enters, the conversation starts falling apart.

Do we blame the lack of sophistication in the software for this fault? I don’t think so. Many people would also have difficulty carrying open ended conversation for long. An informal banter tends to disintegrate into aimless meandering. Next thing you know, we see people saying “talk to you later” and bailing out of the fruitless chat. It would be unrealistic and unfair to expect chatbots to perform better than average humans.

How to fix the problem?

When building a chatbot, try to avoid proposing piecemeal discovery scenarios. This involves chatbot having to lead the user along the step-by-step discovery journey. Whenever that happens, things disintegrate in a rapid fashion.

You may however realize, while building your solution, that piecemeal discovery journey is unavoidable. That’s a clear sign that you need an app. Apps are perfect solutions for letting users browse, poke around, proceed in a step-by-step fashion while discovering the inventory. Conversational interface is a bad, nay terrible choice for enabling such services.

Only build solutions including the conversational interface when the problem space deals with offering specific expertise. Remember, chatbots are just another word for experts you can talk to.

Conclusion

When people use high tech products it is always in a self-serve fashion. Many apps exist on the market today, and all those apps allow users to serve themselves. We have Google global search engine. We also have Expedia travel search engine, as well as Amazon shopping search engine and so on. All those apps expect users to be in the driver’s seat while doing self-serve activities.

When people use those self-serve apps, they don’t usually know what is it they actually want. People tend to browse, to fish for results. Using piecemeal discovery while going on a scavenger hunt is the typical experience when using apps.

Chatbots are the next iteration of high tech services. Chatbots are the natural progression of so-called Expert Systems. In the early days of AI we had those Expert Systems, and they were clunky. Today those Expert Systems have evolved into chatbots (or bots for short). Hence it’s important to keep in mind that bots must offer some expert service. If bots do not have the expert service capabilities, they will end up being useless, irritating, annoying.

Furthermore, bots must not expect their users to be the experts. Unlike apps, which place the onus of expertise on the users, bots actually embody the expertise. Meaning, bots are only useful if interacting with people who actually know what they want to get out of the service.