People will talk to your bot like it’s a person no matter how much you train it to act like a robot. The default assumption seems to be to expect it to behave completely as a human would and people are disappointed when it fails to deliver. But, delivering on this is a tall order, and seems to be putting a lot of unnecessary pressure on bot designers and AI systems to deliver near full human like experiences. Which in turn is discouraging companies from exploring this potentially transformative technology, throwing out the baby with the bath water.
Falling in love with non-humans is totally human
A woman developed a long-term affection with the Eiffel Tower and on April 2007 Erica LaBrie became Erica Eiffel in a commitment ceremony between her and the iconic tower in Paris. It was one of these pieces of news which people read with a chuckle about how unusual some people can behave. For most of us reading it this behavior is incomprehensible and so far from our way of thinking; but are we really so far away from it?
Knights used to name their swords, and musicians name their instruments (for example, Eric Clapton played Blackie, and George Harrison played Lucy). People ask invisible gods for forgiveness, talk to their plants, kiss dice to persuade a profitable roll, name their cars, curse at unresponsive computers, or swear at an automated phone system. We all do it. These things seem a little less weird than marrying the Eiffel Tower. just because they are more common. Creating humanlike agents out of those that are clearly nonhuman, namely anthropomorphism is a natural human tendency.
In the movie ‘Her’, released in 2013, a man falls in love with ‘Samantha’, his computer operating system, whose only physical representation is the female voice coming from his computer. For most of us, that may seem nuts, but, as the movie name implies, anybody who has referred to their GPS, Siri or Alexa as “her” instead of “it” has anthropomorphized an object. If ‘Her’ depicts a man going too far, than where is the line, and is it moving?
Whether you design for it or not, people will treat your bots as humans
At OneReach we are attempting to understand this as it pertains to interaction with conversational Bots. We watch the way people interact with the bots made on our platform and try to learn from it. For example, recently Anheuser-Busch created a bot dedicated to the National Football League. The bot sent users an SMS asking them to guess the next play, and afterwards, the bot would then text back to the user and let them know if they won or lost. If they won, they received a free six pack; if they lost, they got a discount. The results were an interesting illustration of how people think. After getting a text message with the results, many people texted back with a strong emotional response, some saying ‘Thank you’ and the others sharing their excitement with the bot explaining what they plan to do next. And those who lost often making jokes using frown emoticons and generally replying in the same manner as they would if texting back to a person.
- TY :stuck_out_tongue: GO BRONCOS :football:
- Wahoo! Thanks man!
- Bummer, maybe next time(
People couldn’t help themselves. They basically treated the bot as if it was a human being, even though they knew it was a robot on the other end of the line.
In our efforts to make the bots more user-friendly, we wanted to know what was going on.
Why are people doing it?
It is a natural inclination of humans to anthropomorphize things around them, this is how our brain works. This has been true since the beginning of humanity. The Greek philosopher Xenophanes was the first to use the term anthropomorphism when describing the striking similarity between religious believers and their gods, with Greek gods having fair skin and blue eyes and African gods having dark skin and brown eyes.
According to Adam Waytz from Harvard University, there are three key factors that increase likelihood of a person anthropomorphizing an object or a phenomenon.
The first one is actual resemblance of an agent to a human in its behavior. Anthropomorphism involves using the knowledge about self or humans in general to make an inference about an unknown non-human agent. The more this inference is applicable, the more likely is anthropomorphism.
The second one is the fact that people are social creatures and always look for ways to better build themselves into society and make meaningful social connections. Waytz mentions that people who feel lonely are more likely to anthropomorphize technological gadgets, pets, etc., and to believe in commonly anthropomorphized religious agents (such as God or angels). According to Dr Rebecca Harris, a psychologist at the University of Bolton, 20-80% of American adolescents and 40-50% of elderly population report feeling lonely often. Based on that, one might conclude that there is a link between the growing epidemic of loneliness and our society’s growing tendency to anthropomorphize things.
The third factor that may increase our tendency to anthropomorphize is effectance—the basic motivation to gain comfort in unfamiliar environments. People feel insecure when they don’t understand their environment, cannot predict it or control it. Given the overwhelming number of biological, technological, and supernatural agents that people encounter on a daily basis, one way to gain some understanding of these often-incomprehensible agents is to use a very familiar concept (that of the self or other humans) to make these agents and events more comprehensible. This is why we tend to anthropomorphize robots, gadgets, and even natural phenomenon.
Anthropomorphising bots is the first stage, and what will follow?
By understanding each of these factors we can craft experiences capable of building deep relationships with users. Striking the right balance however can be difficult. Although we are still years away from building bots that most humans could fall in love with, there is a potential that some day our emotions would easily confuse our affection for AI with our affection for a real person.
With all that said, perhaps bots may never typically be human replicas, why limit them. Rich Web Chat is an example of how bots could communicate better than humans do, so perhaps bots really just become a different type of intelligence and therefore not be commonly used to replace humans, but serve them in ways that make sense for machines. Using a common language certainly makes sense, but having emotional interactions may just be nothing more than entertainment.
Will we create a new language or style of communication used primarily to talk to bots?
Only time will tell if we become more desensitized or allow ourselves as a society to interact with the technology as if it was human. This journey of
designing a type of intelligence that does not yet exist will be nonetheless an exciting challenge for us UX nerds.
AI and chatbots are only finding their way into our everyday life. This is a very new technology and bot creators are only starting to look for optimal ways for people to learn how to communicate with chatbots. How do we make this communication less complex or more efficient? How do we make chatbots feel familiar without going too far? Most bot creators seem to choose anthropomorphized bot dialog to serve these purposes, partly because we have no other model to look at, partly because antrophomorphization is a common way for humans to deal with something unknown. With time, the efficiency of human-bot communication will improve, most of the misunderstandings will be fixed and lots of valuable data will be gathered on how people interact with bots. As the users grow more and more accustomed to communication with bots and AI, new ways and paradigms will emerge. The technology will be leveraged in more functionally powerful ways, for serving new purposes – likely without being limited to mere human-like exchanges.
For now we are left with questions. At OneReach, we keep looking for the answers and so far we are inclined towards letting machines be machines – even if they are machines with personality. And we are curious to hear what you think on the topic! What do you think are the ways to make human-bot interactions more user-friendly? Is it a good strategy to make the bot pretend to be human? Have you ever spoken to a bot that was designed well enough to make the conversation resemble a conversation with a human being? What types of problems do you think the bots are ideally fit for solving? When is it OK to give your bot a sprinkle of personality, and when is it silly?
If you liked the article and are interested in hearing the updates on the topic, please follow us on Twitter / Facebook. We would love to hear from you!