How to Know If You Are Talking to a Bot
Artificial intelligence (AI) is invading every aspect of our lives. The algorithms control so much that you can conclude that the robot uprising has already happened and we have lost, and lost very badly. Robots decide what song will play next, robots recommend TV shows, and robots are even pretty good at writing and making music . Most of us are fine with this. AI tends to automate tasks that we want to automate, such as the above streaming recommendations, and at some level we are all aware of these bots in our lives, so our interactions with them are more or less voluntary. But there is another way that AI is invading our lives, and we usually don’t even realize it: conversational AI. From chatbots to Tinder bots (yes, really ), bots pretend to be human and talk to us, often without identifying themselves.
Luckily, conversational AI is not yet perfect, and it is possible to detect a bot, although this may soon change as technology advances. For now, if you want to know if you are dealing with a bot, there are several strategies that should reveal the truth.
Beware of clumsy wording
Even the best conversational AIs often have problems with the strange complexities and inconsistencies of language. English is especially difficult to learn because it’s not so much a language as it’s a collection of stolen words and conflicting grammar rules that are bizarrely pronounced without logic, and AI struggles just as much as humans. While AI can usually achieve pretty decent results (especially if its scripts have been augmented by a human), there are often treacherous garbled sentences that just don’t make sense, or weird substitutions. If you feel like you’re chatting or talking to an alien who learned English by watching Japanese TV, you may be dealing with a bot.
Look for repetition
Bots also tend to be extremely goal oriented. Human conversation tends to be fluid—topics are introduced, dropped, and then picked up again. But bots are usually created for specific purposes, and they will stubbornly pursue those purposes no matter what you do. If you notice that the “person” you are talking to or chatting with keeps coming back to the same recommendation or solution no matter what you say, you may be dealing with a bot. If they literally repeat the exact wording every time, that’s an even stronger sign, because people tend to change how they phrase things, especially if they feel like they’re not getting through to you.
Pay attention to response speed
Another sign that you are dealing with AI is the speed of their responses. Whether it’s via chat or over the phone, bots can usually generate responses much faster than humans. If chats come back to you instantly, or if a voice on the phone can instantly give you the information that a person should reasonably be looking for in some resource, you are either dealing with a bot or the most talented support representative in the universe.
Pay attention to ambiguity
Bots are often programmed to give vague, nonsensical responses when they don’t understand, often repeating what you just said to give the illusion of attention. This is an old trick. ELIZA, a “chatbot therapist” developed in the 1960s, uses it all the time. If you tell her, “I’m sad,” she’ll say, “How long have you been sad?” This is a simple algorithmic construction, but it creates the illusion of reasonableness. If the “person” you’re chatting with constantly reverses your statements for clarification, you may be dealing with an AI that uses this trick to bypass sentences it can’t easily parse.
Another aspect of uncertainty is answers that make sense at first glance but add nothing to the conversation. This is because AI is usually programmed to create a language that resembles real-life communication but doesn’t really add anything.
Pull “Crazy Ivan”
If you suspect you’re dealing with a bot but aren’t sure, there’s a test you can try. In The Hunt for Red October, the captain of the Russian submarine, played by Sean Connery, is famous for pulling out Crazy Ivans while scuba diving—suddenly turning his boar to see if an enemy submarine was hiding behind it. Such an unexpected move can also disrupt the work of the AI.
While conversational AI has become extremely complex, and it can be hard to tell from a short interaction that you’re not talking to a human, bots still have one major weakness: difficulty with inconsistency. This is especially true of emotions and human relationships. In the middle of a conversation, ask the suspect bot about their family or tell them you’re depressed, just to see a reaction. Bots are generally not ready for this and will either stubbornly continue the message, ignoring your outburst, or offer a standard message like “Sorry, I don’t understand.” This is because AIs are ultimately programs and therefore have what is known as error handling. If it can’t process what you say, it will run the same subroutine over and over again.
In some cases, the bot will not answer Crazy Ivan’s question at all, while the real person will probably laugh nervously and at least make some effort to answer.
There will come a time when we won’t be able to tell a bot from a real person, but that’s not all.