If I were a robot…what would I need to know?
I can’t stop thinking about this question. Specifically, I am recently consumed by the question:
How do you teach a robot human behavior?
Warning: I am about to geek out about robots, technology and artificial intelligence in this post. Interested? Terrified? Baffled? Read on.
The BIG Idea:
How can we manage people and robots working together? Robots are going to become more and more integrated into our workspace and our lives. How can we as people and companies prepare for the issues of living and working in an artificially intelligent world?
We have to program robots with social skills in addition to technical skills.
Dr. James E. Young works in the Human-Computer Interaction Lab and has intensely studied how robots and humans interact. His argument is compelling:
Robots must have social skills.
“People are naturally skilled at quickly analyzing social interaction, and this is as true with reading robots as it is with reading other people. For example, instead of surprising co-workers with unexpected movements, robots can use their eyes, where they look and how long they stare, or gestures to show intention before acting, to maximize team-worker awareness.” –Dr. Young
A robot is a machine? That sentence should be a statement, but I formed it as a question because research shows that despite knowing that robots are droids devoid of emotion, humans are treating them like humans.
Have you named your Roomba? Your car? Your computer? Humans can’t help but assign human-like characteristics to devices so they need to be able to interact like us.
Scary and Wonderful
Here’s what’s both terrifying and thrilling about the idea of teaching social skills to robots. At the very launch of my people skills course I tell my students:
“You have been given technical skills, but you need people skills to thrive in this world.”
In 5, 10 or 15 years I might be saying the same thing to a group of robots… or a group of engineers programming the robots.
Robots can be given technical skills and intelligence, but we have to also give them people skills, social skills and nonverbal skills if we want them to thrive in this world.
Here’s what spawned this post:
1. I watched Ex-Machina. Oh. My. God. Mind-blowingly good movie. It delves into how artificial intelligence can be taught how to interact (and fool) humans. I also enjoyed Her which shows us how easy it could be for humans to fall in love with artificial intelligence.
2. James E. Young made a compelling argument about how robots are soon going to become fixtures in our workplace. In his words, “Move over and make room for your new colleague: a robot.” He argues that robots will no longer be part of isolated assembly lines and soon will begin to work alongside and in collaboration with people.
3. I read this article: The Top Jobs In 10 Years Might Not Be What You Expect. By 2025 many of our jobs will be replaced or supplemented by robots. It got me thinking about how we are going to cope.
So now let’s get down to the big question: If I was a robot, what skills would I need to understand about humans in order to interact with humans? At our human behavior research lab we study the hidden forces that drive our behavior. This question was the same question I ask myself every day, but flipped on its head. I usually ask:
What forces drive our behavior and how can we leverage them to benefit us?
Today I ask:
What forces drive our behavior and how can we leverage them for robots to benefit us?
If I was building a robot or an artificial intelligence system, here are the human behavioral patterns I would use:
1. Similarity Attraction Effect
Humans love humans who are like them. We like people who look like us, we like people who have the same values as us and we like people who remind us of ourselves. In fact, humans love anything that is like them. Sam Gosling, founder of Snoopology has found that our stuff is an extension of ourselves. Artificial Intelligence (AI) designers can use this to their advantage.
If companies or organizations want robots to fit into a work environment or a home (think about a new robotic colleague or an artificially intelligent maid), then they have to be liked. The best way to do this is to make the robot look, sound and act like their owner or team as much as possible.
- When a robotic colleague joins the team it should be pre-programmed to talk about the local sports team.
- When a robotic maid enters the home it should be dressed in modern clothing styles and programmed to spout knowledge from the blogs that the home owner already reads.
- When a robot needs to bond with a colleague it should know about the latest episodes of popular TV shows for that demographic.
2. Body Language
Young’s team found that “people interpret how a robot moves—fast, slow, soft or jerky motions, etc.—in emotional terms.” When a robot is moving more slowly, people think it is sad, depressed or low on energy. When a robot is jerky or quick moving it is assigned peppy personality traits. Young also found that people get uncomfortable when a robot breaks social nonverbal norms like having shifty eyes or inauthentic facial expressions. Robotics engineers and AI designers need to know the rules of body language while programming their machines. Especially:
- In the average conversation we make eye contact 60-70% of the time. Over 80% is considered creepy and invasive; under 50% is considered shifty or avoidant.
- A slow triple nod is a sign of engagement. If the robot has to ‘listen’ to instructions it would be a very good idea for programmers to make sure the robot is able to do a slow triple nod at important points. This encourages humans to feel heard and therefore talk and share more.
- The palms-up gesture is the universal gesture for openness or approachability. During questions or commands, robots should be programmed to take an open posture to be seen as non-threatening.
*The open palm gesture during commands is especially important because Young and his colleagues found that a command from a robot is taken especially seriously by humans. When small robots gave participants in a lab tedious commands, people had a very difficult time saying no. Participants even repeatedly expressed a desire to quit and tried rationalizing with the robot, but did not quit until given permission.
3. Personality Science
The Five Factor Model has found that every human has 5 basic personality traits: Openness, Conscientiousness, Extroversion, Agreeableness and Neuroticism. I think it would be helpful for robots to be programmed with personality traits as well. Specifically, they should be given the personality traits that will be most helpful for their team. From our research in our lab here are some broad strokes:
- People low in neuroticism are the easiest to work with. They are known for being emotionally resilient and adaptable. Unless you want your robot to be the office Woody Allen or a ‘what if’ generator, a robot programmed to be as low in neuroticism as possible is ideal.
- People high in conscientiousness are the people who can be depended upon to get things done. Of course, robots are naturally high in conscientiousness—they are programmed to get every last detail right. I think it would be beneficial to have this trait played up in a robot. The more people are reminded of the robot’s high conscientiousness and aptitude for details, the more likely they are to depend on the robot for answers and trust its directions.
- Do you want your robot to be the life of the party or one of the team? Extroversion is how you are oriented towards people and relationships. Programmers should think very carefully about how extroverted or introverted they want their robot to be. On the one hand, an extroverted robot could easily gather teams, tell jokes and lead discussions. On the other hand, an introverted robot could make few, but powerful statements, let humans step forward first and be an ‘observer.’
4. Emotions Are Contagious
Research is clear: our emotions and moods are contagious. When one person in the group is happy, that person can ‘infect’ those around them with their upbeat attitude. When one person is crabby, they can ‘transmit’ this pessimistic attitude to their compatriots. Why not make robots ‘contagious’ with the moods we want in our space. Do you want a calm, serene work environment? Then the robot should transmit this with a softer voice tone and slower movements. Do you want a high energy, peppy tone set? The robot should speak and move more quickly with a cheerful, lilting tone.
Another aspect of contagious emotions is the idea of having robots mirror their hosts or compatriots. As humans, we mirror people we feel congenial with or bonded to. Instead of setting a robot up with moods, a programmer could code a robot to mirror the people they are with. This could quickly increase bonding between human and robotic partners.
5. Will You Love Your Robot?
If we want to integrate robots into our workplace and lives we have to love them in some capacity. We have to be willing to invest in their repairs, care for their maintenance and be patient with their technological hiccups. Programmers should want to inspire parent-like love in their robot’s human counterparts. To encourage this relationship, I think robots should be programmed to recognize and respond to the 5 Love Languages. The Love Language is a powerful concept put forward by Dr. Gary Chapman that says all humans have different ways they show appreciation and feel loved. Read the full post about the 5 love languages here, but in a nutshell here is each love language and how a robot could be programmed to respect their human counterpart’s needs:
- Words of Affirmation: Humans who need words of praise, supportive sayings and verbal check-ins could easily be serviced by a robot who gives positive feedback, sends supportive emails and regularly pipes in with loving comments.
- Acts of Service: Humans who feel loved by those who do for them could easily feel supported by robots who finish tasks, bring coffee or lovingly do small projects.
- Gifts: Humans who feel loved by small (and large gifts) could be supported by robots who fetch and replenish diminishing office supplies, grab a handful of candy from the candy bowl or, if really empowered, order gifts off their Amazon wishlist.
- Physical Touch: Humans who love to be hugged, massaged and touched might be able to be serviced by robots…let’s keep it PG. I think a nice high five or a supportive pat on the back would work wonders.
- Quality Time: This is a hard one. Humans who love to spend quality time with the ones they love might simply enjoy having another presence in their home or office, especially if a robot was programmed to laugh when its human counterpart laughed while watching funny YouTube videos.
6. Programmed for Efficiency
A robot could easily be programmed to spawn efficiency in the humans around it. Here are two principles that guide human behavior that a robot could leverage:
Social Default Bias: Humans tend to copy those around them when they can’t make an informed choice. Researchers Huh, Vosgerau & Morewedge found that watching other people make a choice can speed up our decision making process. HOWEVER, this effect is strongest when other people aren’t around—for fear of embarrassment at making the wrong choice. Guess what? Robots aren’t human. If humans around the robot are faced with choices, it might be best for the robot to make recommendations based on reviews, pro and con lists or even random choice. The robot’s choice might help humans speed up their decision making process without having to feel embarrassed at making the wrong choice.
The Goal Gradient Effect: We act more quickly if a task has already been started for us. What if a robot was programmed to be a starter or an activator for colleagues and or reluctant teenagers? For example, lets say the future has a household nanny robot in store for us. And let’s say parents want the nanny to help with chores, but still encourage kids to help out to earn their allowance. A nanny robot could be programmed to start a chore (like gathering all the laundry) and dumping it on the teenager’s bedroom door threshold (so he is unable to close his door) until the laundry is loaded in the washing machine. Voila! Activation.
7. This Robot Is Mine
I have no doubt that robots in the future will be a significant investment for businesses and households alike for some time. This means that programmers want robot owners to feel as much connection and ownership over the robot as possible—this will help with lower return rates and more satisfaction with the purchase. There are two ways AI developers can help human buyers feel proud of their robotic investments:
The Endowment Effect: Humans place more value on something once we feel that we own it. Kahneman, Knetsch & Thaler found that as soon as someone is told that they own something—even if they didn’t want it in the first place, they are less likely to want to part with it. The integration of human beings and robots is not going to be smooth, especially in the workplace where it will be hard to get all members of a team on board with a robotic counterpart. The best way to deal with this is by giving each person who interacts with the robot some ownership over it. Perhaps their name should be engraved onto the robot’s surface or credited to each person who interacts with it.
The Ikea Effect: We place more value on products that we have made or contributed to. Norton, Mochon & Ariely gave 2 different groups boxes from Ikea—the Swiss, make it yourself furniture store. One group was given a fully-assembled version, while the other group was given an unassembled piece of furniture and asked to put the item together. The second group was willing to pay much more for the piece of furniture they assembled themselves than the group that didn’t assemble their furniture. AI designers should devise ways to let human counterparts personalize or contribute to the building of their robot, whether that is as small as color choices and a name or as large as personality profiles and technical knowledge.
8. The Team Player
In the future robots may be programmed to help with corporate culture and teamwork. How would they do this? Easily! Studies show that the more team members are encouraged to socialize and interact face-to-face, the happier they are at work, the more energy they have and the longer they can stay focused on a task. Just because a robot isn’t a human doesn’t mean that it can’t encourage humans to interact. IBM even did a study that analyzed the amount of email contacts and chat list buddies that an employee had. They found that the more socially connected employees were, the better they performed. On average, every email contact was worth an added $948 in revenue for IBM! Robots could easily facilitate team interactions by setting up introductions with employees, sending out email intros and cross-referencing coordinating teams and projects.
9. The Robotic Cheerleader
Imagine a robot in a cheerleading outfit. Weird right? I think that robots should be programmed to fill a fundamental human need—to feel supported. Shelly Gable has done research on something called Capitalization. This is when someone shares good news with you and you multiply the benefits of their success by responding to their news positively and actively. Robots could easily be programmed to highlight and announce office or home successes.
- Every time a new company goal is met, the robot could send a message or chat with a specific employee highlighting their contribution to the win.
- If someone has good news, they can share this with the robot who can announce it to the rest of the team.
10. Reading Microexpressions
Robotic facial decoding is already being done, and I can’t wait to see how quickly this robotic skill could change the way we interact. There are 7 universal facial expressions (learn all 7 here) that could easily be programmed into a robotic interface. Robots could read and respond appropriately to human emotions in many ways:
- If a robot spots intense and prolonged sadness on a human face, it might be able to offer support or report the person to a counselor for help.
- If a robot spots anger on a human face during a negotiation or conversation, it might be able to initiate de-escalation techniques or instigate a cooling off period for human and robotic participants.
- If a robot spots sudden fear on multiple faces, it might be able to alert authorities or call a supervisor in case there is an emergency.
This entire post has been about how to get humans to bond and connect more to artificially intelligent machines, but that’s not always the best choice. If a robot is in an environment where it might be destroyed or mangled, then human bonding could be harmful. As we discussed in the beginning part of this article, people can’t help but assign human characteristics to robots. Young has found that people often get upset or distraught when robots are ‘hurt’ or lose their memory.
“There are reports from military situations, for example, where soldiers have demanded that their robots be repaired rather than replaced, or are hesitant to place their robots in danger.” –Dr. Young
- Young recommends that people separate robots from doing rapport building collaborative tasks from those that do dangerous ones.
- If humans are in the presence of robots being destroyed, disassembled or mangled, then those robots should not be given any kind of personalization (the IKEA effect) or ownership to one person (the endowment effect).
I am both scared and excited for robots to enter our world. Truth be told, whether you like the idea of robots or not, they’re coming and we have to prepare.