Skip to main content

Our Reactions to Robots Tell Us Something About Ourselves

Robot football players at a Valparaiso University College of Engineering event

I have been thinking lately about robots, which creates an interesting asymmetry: They almost certainly have not been thinking about me. Nevertheless, I find that I often respond to robots as though they have thoughts about me, or about their own personal goals, or about the world in which they exist. That tendency suggests an interesting aspect of human psychology, connected to our social minds. We are hard-wired to care what other people think about us, and we very easily extend that concern to robots.

Here's a recent article about how the language-learning app Duolingo, which features an owl-shaped avatar (a kind of robot), uses "emotional blackmail" to keep application users engaged: https://uxdesign.cc/20-days-of-emotional-blackmail-from-duolingo-4f566523e3c5 This bird-shaped bit of code tells users things like "you're scaring me!" and "I miss you" if they haven't logged in for a few days. Many people find the emotion-laden messages quite distressing, and log back in to appease the owl -- which, remember, doesn't really feel anything at all. Duo the owl also has a range of happy and sad facial expressions, as do the other cartoony denizens of the app. Why should I care that Zari the artificial teenager is proud of me for completing my daily Spanish lesson? By any rational standard I should not, and yet I do. China has harnessed people's desire for social approval in perhaps more concerning ways, using social networks to shape behavior that the government considers to be in the people's best interest.

The following video shows some physically embodied robots, which in terms of consciousness are no different from Duo the language-learning bird (that is to say, they have none), but which push our buttons in a different way:

There are three types of machines in robot football: One that hikes the ball, one that catches it in a net, and a third type that resembles a computer case on wheels, which attempts to run into the net-wielding robot and make it fumble the ball. The very fact that these three types of robots exist makes me read some intentionality into their behavior: Don't the tank-like rectangular robots seem more aggressive to you? Don't the poor ball-carrying robots seem a bit beleaguered? And then, at another level, I notice my tendency to attribute personalities to the individual machines: Why is that net robot moving around at the start when the others are all standing still -- does it have a rebellious attitude? Why is the tank-like robot spinning in circles in the middle of the floor -- is it confused? And that other net robot speeding across the room -- is it the star player, eager to demonstrate its skills? None of these attributions are true, of course. The people on the sidelines helping to direct the robots might have personalities, but the robots certainly don't. And even if the robots were fully autonomous, I think I would still be drawing conclusions about their internal mental lives.

Our reactions to robots are a function of the social-connectedness goals held by our Intuitive minds. Earlier humans had to live within small groups that relied on one another for survival, which made the pressure for conformity intense. In medieval England, for example, there was no police force, yet there was also minimal crime. A person who transgressed against their neighbors would be subjected to punishments that today seem very harsh, ranging from public humiliation in the village stocks up to exile in the forest, which without the resources of the village was tantamount to execution. The reason our Intuitive minds care so very deeply what others think of us is because they evolved in a context where humans who didn't have those instincts didn't survive. 

Our tendency to read intention into the actions of robots is an example of what psychologists call "theory of mind" -- our ability to guess what's going on inside the head of another person. This is a uniquely human ability: There's some evidence that AI systems can now fake it, as they mimic other aspects of human language, but I'm very doubtful that an AI has an actual sense of what it's like to be a person (that's because I'm also very doubtful that AI is, or could ever be, conscious). Of course, we can never know for sure about another person's internal experience -- they might not even have any, in which case they would be a biological robot (also called a "philosophical zombie"). But I'm pretty sure that you are having an internal experience as you read this, just as I had one while I was writing it. And I think my internal theory of mind is good enough that my guess about your experience is reasonably accurate (my guess: feeling calm, interested, and maybe a little bored, hungry, or tired -- a description that probably covers two-thirds of people at any given time on any given day!). 

The problem is that my theory of mind feels so convincing that I can't turn it off when a piece of software starts to exhibit human-like behavior. That's not a big deal with football-playing robots. But as AI systems learn to act more like human beings, it's a tendency that could steer me wrong eventually. New York Times columnist Ross Douthat points out that most of our fiction about robots assumes they will have conscious, rational awareness like humans without having the emotional range of humans -- Mr. Data's difficulty expressing emotion on Star Trek is typical. But contemporary AI systems are very convincing liars in describing their emotions, values, hopes, and dreams (if anything, logic is where they have some trouble!). In other words, science fiction has prepared us for robotic systems that still can't pass the Turing test, but relatively low-tech AI mimicry can now do that very consistently. Perhaps we will all need to learn to be a little less concerned about feelings and focus more concretely on what we can observe and understand, if we're going to succeed in this brave new world of emotionally manipulative cartoon owls.

Comments

Post a Comment

Popular posts from this blog

Prototypes and Willingness: The Theory of Planned Behavior Revisited

  You may recall my blog post from last year on the Theory of Planned Behavior (TPB) , titled "in praise of a failed model." My evaluation of this model was that it accurately describes the Narrative Mind, which does control intentions. But the ultimate goal of the TPB is to predict behavior, and the relationship between intentions and behavior is weak at best -- in fact, it is entirely attributable to the fact that when someone says they don't intend to do something, they probably won't do it. When they say they do intend to do it, their actual results are no better than chance, a result of the intention-behavior gap as described in Two Minds Theory.  The full TPB is shown in this diagram: Cognitive constructs like attitudes, subjective norms, and perceived behavioral control (i.e., self-efficacy) are Narrative-system phenomena, and they do indeed have relationships with each other and with intentions (which are also products of the Narrative Mind). Perceived behavi...
Post a Comment
Read more

New Study Suggests That Fatigue is Most Detrimental for the Narrative Mind

I'm one of the authors on a new study by Dr. Mustafa Ozkaynak's research team, which looks at how emergency department (ED) nurses change their decision-making process when they become fatigued. In a previous paper , we found that fatigue was common in ED nurses, particularly toward the end of their work shift, and that nurses' fatigue was more often characterized as physical rather than mental or emotional -- in other words, this really represented being physically tired  at the end of the day, not being burned-out or depressed. Nevertheless, physical fatigue has important effects on nurses' decision-making in the ED. Based on nurses' qualitative reports, fatigue has mixed effects on their clinical performance. Nurses said that they definitely cut corners when they were tired, for example in terms of documentation in the electronic health record. They felt that they were less careful about double-checking things, and might be more likely to make snap decisions. We...
Post a Comment
Read more

Is AI Out to Get Us?

 I wrote earlier this year about a disturbing report on the prospects of self-improving artificial intelligence (AI) models deciding to take over the world. Much of the fear around AI relates to something called the "alignment problem," which simply means that an AI model might have goals incompatible with human flourishing -- or in some dystopian scenarios, with human life itself. A classic example of this line of thought is the "paper clip problem," in which a superintelligent AI is tasked with making paper clips. Eventually every resource in the world -- including human beings -- becomes just another obstacle for it to overcome in its goal of transforming the entire universe into paper clips. So far, that's not the danger -- AI models don't have that level of direct control over the physical world (yet). But a couple of new developments in the past few months do  suggest that AI models are pursuing goals different from what their human designers might w...
Post a Comment
Read more