MEDFORD, Mass. >> Amory Kahan, 7, wanted to know when it would be snack time. Harvey Borisy, 5, complained about a scrape on his elbow. And Declan Lewis, 8, was wondering why the two-wheeled wooden robot he was programming to do the Hokey Pokey wasn’t working. He sighed, “Forward, backward, and it stops.”

Declan tried it again, and this time the robot shook back and forth on the gray rug. “It did it!” he cried. Amanda Sullivan, a camp coordinator and a postdoctoral researcher in early childhood technology, smiled. “They’ve been debugging their Hokey Pokeys,” she said.

The children, at a summer camp last month run by the Developmental Technologies Research Group at Tufts University, were learning typical kid skills: building with blocks, taking turns, persevering through frustration. They were also, researchers say, learning the skills necessary to succeed in an automated economy.

Technological advances have rendered an increasing number of jobs obsolete in the past decade, and researchers say parts of most jobs will eventually be automated. What the labor market will look like when today’s young children are old enough to work is perhaps harder to predict than at any time in recent history. Jobs are likely to be very different, but we don’t know which will still exist, which will be done by machines and which new ones will be created.

To prepare, children need to start as early as preschool, educators say. Foundational skills that affect whether people thrive or fall behind in the modern economy are developed early, and achievement gaps appear before kindergarten.

Nervous about the future, some parents are pushing children to learn to code as early as age 2, and advocates say it’s as important as learning letters and numbers. But many researchers and educators say that the focus on coding is misplaced and that the more important skills to teach have to do with playing with other children and nothing to do with machines: human skills that machines can’t easily replicate, like empathy, collaboration and problem-solving.

“It’s a real misnomer that simply learning to code is the answer,” said Ken Goldberg, a chairman in engineering at the University of California, Berkeley. “We don’t need everybody to be extremely capable Python coders. It’s a way of understanding what machines are good at and what they’re not good at — that’s something everybody needs to learn.”

It’s not that technology should be avoided; many researchers say children should be exposed to it. But we don’t know what machines will be able to do in two decades, let alone which programming languages software engineers will use. And children learn better, they say, by playing and building instead of sitting behind screens.

“We don’t want all these young kids sitting in front of a computer,” said Marina Umaschi Bers, a professor of computer science and child development who directs the Tufts research group. “We want them to move around and work with each other.”

Umaschi Bers developed the Kibo robot that Declan was using and ScratchJr, a programming language for the under-7 set. But she says the bigger point is to teach computational thinking. That essentially means breaking down problems into smaller parts and creating plans to solve them — with prototypes, feedback and revisions — in all parts of life.

“That’s key for programming, and it’s key for life,” she said. Her curriculum, used in schools nationwide and abroad, teaches skills like sharing and perseverance, and is woven into all subjects in the school day: Children program robots to act out a story they’re reading, for example.

These ideas reach back five decades to Seymour Papert, a mathematician and educational theorist. Children learn best not when a teacher or computer feeds them knowledge, he said, but when following their curiosity and making things, whether a sand castle or a robot. Like computer programmers, children make mistakes and solve problems along the way.

In 2006, Jeannette Wing, a computer scientist at Columbia, revisited the notion of computational thinking as a skill for everyone to learn and use. “Computational thinking is a way humans solve problems,” she wrote. She gave everyday examples: “When your daughter goes to school in the morning, she puts in her backpack the things she needs for the day; that’s prefetching and caching.”

At the Tufts summer camp, the children were programming the robot — which has no screen but uses colored wooden blocks and a bar code scanner — by building a sequence of blocks labeled with commands like “turn right” or “spin.”

They had already learned an important piece of syntax: The program must begin with a green “begin” block and end with a red “end” one. Nico Luker, 6, decided to test what would happen if he scanned a program without the “end” block. “It won’t work without an end,” predicted Noam Webber, 6. Sure enough, the robot sat still.

Meanwhile, the children were learning about teamwork, failure and sharing. “Technology can be a vehicle to help people create and collaborate better, but at the end of the day, people need to learn to work with people,” Umaschi Bers said.

To begin, teachers place all the materials on tables, and children take what they need. Some try to take everything, prompting a conversation about sharing and ethical choices.

There’s a testing area, where children get points for how many times they try something that fails. “We don’t create an artificial environment where everything will work,” Umaschi Bers said. “We let them find frustration because they will only learn to manage frustration if they encounter it.”

Teaching social and emotional skills is fashionable in education right now, but it’s been part of high-quality teaching for decades, and randomized trials over time have shown how important it is to adult success, said Stephanie M. Jones, a professor of education at Harvard who studies social and emotional development.

“If you raise and educate kids to be flexible, problem solvers and good communicators, they can adapt to a world that is new,” she said.

This is natural to the way preschoolers learn, said David Deming, a professor of public policy, education and economics at Harvard. They flexibly move from the art area to the block area during free play; they build structures and make collages; and they share toys and try again when they mess up.

A big challenge — and one he said is essential to preparing children for a labor market in which routine and individualized tasks are being automated — is making sure this style of education is not lost in higher grades, when teachers turn to lecturing and standardized curriculums. Just as preschoolers learn math by operating a pretend store instead of doing work sheets, he said, high schoolers should learn government by staging a mock Congress rather than reading a textbook.

“You’re learning to work in groups and be creative, and that this problem you’re facing today looks like a problem you faced in a different context a year ago,” he said. “That is a process that is very hard for artificial intelligence to replicate.”

© 2017 The New York Times Company