Robot Shows How Babies Are Actively Plotting to Make You Smile

Experiments with robot babies reveal that when a baby smiles at you, it's trying to get you to smile back

3 min read
Robot Shows How Babies Are Actively Plotting to Make You Smile
Photo: UC San Diego

Sometimes, babies like to smile. Perhaps you’ve noticed this. Sometimes, they smile because they’re happy, but a lot of the time, they’re smiling primarily because they want you to smile, and they’re doing it using “sophisticated timing” to manipulate you into obeying them. 

Researchers from Olin College, the University of Miami, and UC San Diego have been studying baby smile schemes, and to test their hypotheses, they’ve used a slightly uncanny robot baby to smile at undergrads who probably had no idea what they were signing up for.

The first part of this study, which is published today in PLOS One, involved watching real mothers interact with real infants and quantifying their smile interactions. Between eight and 12 months of age, infants routinely use gestures to influence the actions of others, such as reaching a toy to prompt someone to get it for them. They also use what’s called anticipatory smiling: smiling at something and then turning to an adult while maintaining their smile as a means of communication. Younger infants (four months of age and under) don’t exhibit these behaviors, but they do smile a lot, especially around adults, and the researchers wanted to figure out if these younger babies are engaging in a deliberate form of interaction.

Thirteen infant-mother dyads (which I guess means “pairs” if you speak science), consisting of infants between 4 and 17 weeks of age, were observed interacting with each other. For each dyad, the researchers assumed that the mother and infant were each trying to achieve one of four different goals: maximizing the time of simultaneous smiling, maximizing mother smiling, maximizing infant smiling, and nobody smiling. By statistically analyzing when the mother and infant smiled, and what effect that had, the researchers were able to determine what the most likely goal for each was.

“Babies actively preferred to not be smiling themselves while their mothers were. According to the researchers, ‘infants exhibited sophisticated timing behaviors to achieve their goals.’ ”

It turns out that the highest probability goal for mothers (about 70 percent) was to maximize simultaneous smiling, while the same analysis applied to the infants showed that they were significantly more likely (about 80 percent) to be trying to maximize mother-only smiling. In other words, mothers are trying to smile and be smiled at, while babies only care about being smiled at. In fact, the babies actively preferred to not be smiling themselves while their mothers were. According to the researchers, “infants exhibited sophisticated timing behaviors to achieve their goals.”

One might expect that infants would be unlikely to ever smile when mother is already smiling. However, this was not the case. Implementing maximally efficient infant wait times before smiling involves optimal tradeoffs between immediate and long-term goal maximization. An infant strategy of never smiling maximizes immediate time in the mother-only smiling configuration, but at the potential expense of future time in the mother-only smiling configuration (because mother is likely to cease smiling after a number of seconds of smiling alone). Predictions derived from the goal of maximizing “mother-only” smiling appeared to maximize immediate seconds of mother-only smiling while minimizing the probability of mother terminating her smile.

To validate these results, researchers needed a standardized infant to experiment with. You remember Diego-San, I bet:

Diego-San was programmed to implement the goal of maximizing adult-only smiling that the real infants seemed to be doing, using the control policy synthesized from the inverse optimal control analysis of the smile behavior of the infants themselves. It was then thrust in front of 32 UC San Diego undergraduates for four three-minute test periods, testing the infant smile strategy along with a few others as test and control cases. Even though Diego-San is a robot, and the undergrads were (presumably) not the robot’s mother, the result was the same as with the real infant-mother dyads: “the duration of participant-only smiling was significantly longer for the controller based on the inferred infant goals than for each of the other 3 control conditions.”

What this study shows is that mothers time their smiles to maximize the amount of time that they and their infants are smiling, while infants time their smiles to maximize the amount of time that only their mothers are smiling. The study also shows that there is a deliberate and systematic timing going on that’s designed to have a direct influence on the behavior of the other party. What the study doesn’t show is whether the infant (or the adult, for that matter) is conscious of doing this, although it seems likely that at the very least, the infants are altering their behavior to achieve their desired goal state. Tricky little buggers, aren’t they?

[ PLOS One ]

The Conversation (0)