TED: What Young Children Teach Us About Being Scientists
It's not just the What Works Clearinghouse that considers randomized samples the gold standard for research—toddlers do, too.
"Generalizing is the bread and butter of science. This only works if our sample is random," not cherry-picked, said Laura Schulz, a cognitive sciences professor at the Massachusetts Institute of Technology. "It turns out the fascinating thing about science is also a fascinating thing about children, which ... is precisely their ability to pull such rich inferences rapidly and accurately from such sparse and noisy data."
In a presentation Tuesday at the 2015 Technology, Entertainment, and Design conference in Vancouver, Canada, Schulz laid out an ongoing series of experiments showing that children as young as a year and a half react differently to data that seems to be a random, representative sample versus cherry-picked information.
In one experiment, a researcher showed a child a box of mixed blue and yellow balls. She then pulled three blue balls, and showed the child that each ball squeaked. The researcher then pulled a yellow ball—which did not squeak—from the box and gave it to the child.
Children reacted differently to the silent ball depending on whether or not they thought the three blue balls were a random, "representative" sample of the box. When the box appeared to be an equal mix of blue and yellow balls, 80 percent of the babies tried repeatedly to get the yellow ball to squeak and seemed surprised and frustrated when it did not, suggesting they had generalized that "the balls in this box squeak."
However, when it seemed to the baby that the researcher had chosen the balls purposefully rather than at random—when the box appeared to have almost all yellow balls and only a few blue balls—fewer than half of the toddlers tried to squeeze the silent yellow ball. Instead, they cast it aside and tried to reach the blue balls.
"Babies in this respect, like scientists, care whether evidence is randomly sampled or not, and they use this to develop expectations about the world—what squeaks and what doesn't, what to explore and what to ignore," Schulz said.
In a separate experiment, Schulz looked at how children interpret whether a problem is caused by something a person did, or something broken within the object. Two researchers showed toddlers a toy that plays music when a button is pressed—sometimes. In each trial, the toy worked twice and failed twice, before the toy was given to the child.
In the first set-up, the first researcher succeeded, then the second researcher failed twice, and then the first succeeded again. Overwhelmingly, when the child failed to make the toy work, he handed it to Mom, sitting nearby, apparently concluding, "I'm not doing this right."
When instead, the toy worked once and failed once for each of the researchers demonstrating it, the child simply put it away when it didn't work and reached for a different toy, seemingly concluding, "this toy is broken," Schulz said.
Schulz's work adds to growing evidence of complex learning and decision-making even in the youngest children. In a hall full of scientists, entertainers and tech entrepreneurs, Schulz also pointed out the need to invest in the "technology" of early childhood education—not giving toddlers iPads, but nurturing their inherent statistical prowess:
"In the years to come we're going to see technological innovations beyond anything I can even envision, but we are very unlikely to see anything even approximating the computational power of a human child in my lifetime or yours.
If we invest in these most powerful learners and their development—in babies and children and mothers and fathers and caregivers and teachers— in the ways we invest in our other most powerful and elegant forms of technology and engineering and design, we will not just be dreaming of a better future; we will be planning for one."
Schulz's TED talk is not yet available online for non-convention-goers, but below you can see her talk about the work she did with the Boston Children's Museum.