Edge of Innovation

How a human chatbot challenged our perception of artificial intelligence

 

Imagine having no idea whether the person you are talking to is a real person or a computer program.

Chatbots are good. But they’re not that good. But imagine talking to a artificial intelligence that looks like a human. Would you be fooled? Scientists at the London School of Economics decided to find out and conducted an experiment in which a regular person had to say things suggested by a computer.

The human acting on behalf of the computer in the LSE experiment was dubbed an echoborg. Here’s how the process of interacting with one worked:

  • You meet an echoborg, and ask it a question (e.g. its name)
  • A scientist located anywhere in the world hears your question through the echoborg’s transceiver
  • The scientist enters the question into the computer
  • The bot (AI computer program) generates a response
  • The scientist relays the answer to the echoborg through a microphone
  • The echoborg says the answer out loud.
echobot
Source: BBC.

What were the answers? We don’t know for sure, but anyone who has ever interacted with a chatbot knows that conversations can make a lot of perfect sense, that is until the bot suddenly changes the subject or twists everything around. If you want to talk to a pretty smart bot, click here.

If there is a screen between us and another person in a conversation, we may begin to suspect that we’re not talking to a real person. Especially if the answers aren’t a perfect fit for the conversation we’re having. However, would you think anything was amiss if there was a regular person sitting in front of you randomly going off on a tangent? Perhaps not.

The experiment ended with this conclusion: we often pay more attention to a speaker’s appearance than to the message they are trying to convey. The echoborg was a regular person, so the brain assumed that everything was normal, while any changes of subject were attributed to absent-mindedness.

Why is this? In his book Incognito: The Secret Lives of the Brain, David Eagleman explains:

“Not only is our perception of the world a construction that does not accurately represent reality, but we also have the false impression of a full, detailed picture when we actually see only what we need to, and no more. (…) We do not perceive what is actually out there. We perceive whatever our brains tell us to.”

Our perception of reality becomes conscious when what our senses tell us fails to match our expectations. So when you sit in front of an echoborg you assume you are talking to a human being.

So, let’s flip the idea around. What would your brain tell you if you saw a cute little robot? Or a cute big robot like Baymax from Disney’s Big Hero 6? Surely, you wouldn’t think it was real? Or talk it to it in the same way? But what if it were as cute as something like Boxie?

Boxie is equipped with a camera and can ask questions. During testing, it moved freely around the Media Lab at MIT and interacted with people.

Scientists found that complete strangers would become very open with Boxie and share their innermost secrets. Another even cuter robot, called BlabDroid even got to talk to Chris Hadfield, the former commander of the International Space Station.


The appeal of these robots is probably due to the fact that they have a child’s voice, “big eyes” and a “trusting smile”. Those inverted commas are there for a reason: A robot doesn’t have eyes and its smile is only a drawing. Despite this, the human brain still assumes that it has intentions. It looks for meaning and makes judgements. So if something looks small and harmless, we assume that it is.

Steve Brown, Chief Futurist at Intel, thinks we are on the edge of a new era of computer evolution, because computers have begun developing true eyes and ears.

As long as interaction is limited to text on a screen, we have no trouble accepting the fact we are talking to a machine. However, robots will become commonplace in the future. Some of them will look like adult human beings, some like children, some will not resemble us at all.

We should consider whether our brains will fool us during interactions with them and whether we will attribute intentions or emotions to them that they don’t have based on their appearance. Experiments conducted thus far suggest that this is a strong possibility.

Steve Brown, Intel's Chief Futurist
Steve Brown, Intel’s Chief Futurist

We still have some time to prepare for an era in which robots will not be a phenomenon but a part of everyday life. Does a smiling robot have good intentions? Who is behind this cute little machine with a child’s voice asking us to reveal our secrets?

Our brains have made us the dominant species on Earth, but they could lead us astray when interacting with robots.

Share This Article

Related Topics

Tech Innovation

Read This Next