Serve and Return: Why Kids Need Human Back-and-Forth, Not Bots

A toddler points at a dog and says “Da!” Her dad looks, grins and says “Yes, a dog! A big brown dog.” She points again. That tiny exchange, repeated thousands of times, is what Harvard’s Center on the Developing Child calls serve and return: the child “serves” with a sound, a look or a gesture, and a caring adult “returns” it. The center says these exchanges help “build and strengthen neural connections in the child’s brain.”

So here is the question many parents are now asking: if a smart speaker answers every question and a chatbot replies to every message, isn’t that back-and-forth too? Partly. A machine can respond. What it cannot do is look at the dog with your child, notice she is getting tired, and care that she is learning. This article explains the developmental science of contingent response and why a bot’s version is not the same thing.

Key Takeaways

• Serve and return is responsive, back-and-forth interaction between a young child and a caring adult. Harvard’s Center on the Developing Child calls it “essential for healthy development.”
• In a classic 2003 study, 9-month-olds learned foreign-language sounds from a live speaker but not from the same speaker on video or audio recordings.
• In a 2014 study, toddlers learned new verbs from a live person and from a responsive video chat, but not from a matched, non-responsive video. Responsiveness was the key.
• The American Academy of Pediatrics calls safe, stable, nurturing relationships “biological necessities for all children.”
• Chatbots respond to content, but miss shared attention, emotional attunement and a lasting relationship. There is no evidence yet that bot conversation builds the same foundations.

What serve and return is

Serve and return is the term the Harvard Center on the Developing Child uses for “responsive, back-and-forth exchanges between a young child and a caring adult.” When a baby babbles, gestures or cries, and an adult responds in a fitting way, that exchange builds neural connections.

The center’s key claims:

• These interactions “build and strengthen neural connections in the child’s brain.”
• “Healthy brain architecture depends on a sturdy foundation built by stable, interactive relationships.”
• “The human brain expects this game of back and forth, and it is essential for healthy development.”

It also warns about the opposite. When adult responses are inconsistent, missing or harmful, “developing brain architecture may be disrupted,” and persistent absence can activate the body’s stress response. The center is careful to note that caregivers may struggle to respond consistently because of financial strain, isolation or chronic health problems. This is not about blaming busy parents. It is about what children’s brains are built to expect.

What the research says about responsive interaction

Two studies are worth knowing because they isolate the ingredient that matters.

Live beats recorded, even with the same content. In a 2003 study in PNAS, Patricia Kuhl, Feng-Ming Tsao and Huei-Mei Liu exposed 9-month-old American babies to native Mandarin speakers across 12 lab sessions. Babies normally lose the ability to tell apart foreign speech sounds between 6 and 12 months. Live exposure to Mandarin reversed that decline. In a second experiment, babies got the same speakers and materials through audiovisual or audio-only recordings. Recorded Mandarin, “without interpersonal interaction, had no effect.” The authors concluded that early phonetic learning “is enhanced by social interaction.”

Responsiveness is the active ingredient. In 2014, Sarah Roseberry, Kathy Hirsh-Pasek and Roberta Golinkoff published “Skype Me! Socially Contingent Interactions Help Toddlers Learn Language” in Child Development. They taught 36 toddlers aged 24 to 30 months new verbs in three ways: live, over a live video chat, or through a yoked video that showed the same person and words but did not respond to the child. Toddlers learned in the live and video-chat conditions, and not from the non-responsive video. The authors credit “social contingency”: the adult’s response was timed to and shaped by what the child did.

That second finding is important and honest. Screens are not automatically the problem. A grandparent on a responsive video call can be serve and return. What did not work was content that looked social but ignored the child.

Pediatrics has built this into policy. In 2021 the American Academy of Pediatrics published “Preventing Childhood Toxic Stress: Partnering With Families and Communities to Promote Relational Health” by Andrew Garner, Michael Yogman and AAP committees. It calls for a shift toward relational health and states that safe, stable and nurturing relationships “are biological necessities for all children because they mitigate childhood toxic stress responses and proactively build resilience.”

Why a chatbot’s back-and-forth is not the same

This is where an engineer’s view helps. A chatbot or voice assistant is contingent in a narrow, technical sense: its output depends on your input. You say something, it replies to that thing. On paper, that looks like the yoked-video problem solved.

But look at what serve and return involves in practice, and what the system is actually doing:

• Shared attention. The toddler points at the dog; the adult looks at the same dog. A chatbot has no access to the world your child is looking at, only to the words typed or spoken to it.
• Attunement. A parent notices tiredness, frustration or delight and adjusts. A language model predicts likely next words from text. It does not see your child’s face or hear the wobble in their voice unless a product adds those inputs, and even then it is pattern-matching, not caring.
• A lasting relationship. The AAP’s frame is relationships, not exchanges. A child learns that this person, again and again, comes back. The APA’s 2025 health advisory states that “relationships with AI systems are one-sided, even if the user perceives otherwise.”
• Friction. Real people sometimes misunderstand, disagree or need a turn. UNICEF’s June 2026 policy brief on AI companions warns that “excessive use of always supportive chatbots may displace human interactions needed to build empathy, adaptability, and resilience.”

There is a further complication: kids are not sure what the machine is. In a 2024 University of Edinburgh study of 166 children aged 6 to 11, about two in three thought smart speakers might think like a human. A child who believes the voice understands may invest in it as if it were a relationship.

What we do not know matters too. No long-term research has tested whether talking with chatbots helps or harms the development of language or social skills in young children. The evidence we have says the ingredient is responsive interaction with a caring person. A bot provides some of the surface and none of the relationship. Until research shows otherwise, treat it as a supplement for older kids, never a stand-in for people with younger ones.

Feature of serve and return	Live caregiver	Responsive video call with a person	Chatbot or voice assistant	Non-responsive video or TV
Responds to what the child does	Yes	Yes	To words only	No
Looks at the same thing as the child	Yes	Partly	No	No
Reads mood and adjusts	Yes	Mostly	Little or none	No
Ongoing relationship	Yes	Yes	No, one-sided	No
Evidence for early learning	Strong	Some (toddler word learning)	None yet	Weak in studies above

How to protect serve and return at home

Count exchanges, not minutes

Rather than tracking screen minutes alone, notice the back-and-forth. Bath time, car rides and cooking are full of chances: your child names something, you add a word, they answer. Harvard’s center emphasizes everyday moments, not special programs.

Keep the bot out of the “first responder” role for young kids

When a 5-year-old asks “Why is the sky blue?”, the voice assistant will answer faster. Try answering together instead, even if you look it up: “Great question. What do you think?” Then search side by side. The answer matters less than the exchange.

Use video calls for relatives, and make them two-way

The Roseberry study suggests responsive video chat can support learning. A call with grandparents where they ask questions and wait for answers counts. A video your child watches alone does not count the same way.

With older kids, talk about the difference

For tweens and teens, the point shifts from building circuits to protecting relationships. Our article on the chatbot empathy gap has activities to help kids see why a bot’s warm words are not the same as a friend’s. For the science of how empathy grows, see how children develop empathy.

What not to do: feel guilty about every unanswered serve

No parent returns every serve. Harvard’s framing is about patterns over time, and its own page acknowledges the pressures on caregivers. Pick a few protected times each day and do them well.

What to Watch For Over the Next 3 Months

• Week 4: Notice one daily routine, such as dinner, the drive to school or bedtime, where you can add a few extra back-and-forth exchanges. Notice who answers your young child’s questions most often: you or a device.
• Month 2 red flags: In young children, ask your pediatrician about delays in babbling, pointing, words or response to their name. In older kids, watch for preferring to talk to a chatbot over family or friends, which the APA lists as a sign of over-reliance.
• Month 3 self-check: Does your child have a few adults, not just you, who reliably respond to them: a grandparent, a teacher, a coach? That network is what the AAP means by relational health.

Frequently Asked Questions

What is serve and return in simple terms?

It is the back-and-forth between a child and a caring adult. The child “serves” with a sound, word, look or gesture, and the adult “returns” it with attention and a fitting response. Harvard’s Center on the Developing Child says these exchanges build brain connections for communication and social skills.

Can a smart speaker or chatbot do serve and return?

Only partly. It responds to words, but it cannot share attention to the world, read your child’s mood reliably, or form a real relationship. No research yet shows chatbots build the same foundations, while research on live, responsive people is strong.

Are video calls with grandparents okay for toddlers?

Yes, when they are two-way. In a 2014 study, toddlers learned new words through a responsive video chat but not from a pre-recorded video showing the same person. The key is that the person on screen responds to your child.

My child talks to Alexa a lot. Is that harming their development?

There is no evidence that ordinary use causes harm, and no evidence it helps either. The bigger question is what it replaces. If the device answers questions an adult would have discussed, add more human conversation rather than banning the speaker.

Does serve and return matter after the toddler years?

Yes, though it changes shape. For older kids it looks like conversations, shared projects and adults who notice and follow up. The AAP’s relational health framework applies across childhood.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. Center on the Developing Child at Harvard University. “Serve and Return: Back-and-forth exchanges.” Harvard University. https://developingchild.harvard.edu/key-concept/serve-and-return/
2. Kuhl, P. K., Tsao, F.-M., & Liu, H.-M. (2003). “Foreign-language experience in infancy: Effects of short-term exposure and social interaction on phonetic learning.” Proceedings of the National Academy of Sciences. https://doi.org/10.1073/pnas.1532872100
3. Roseberry, S., Hirsh-Pasek, K., & Golinkoff, R. M. (2014). “Skype Me! Socially Contingent Interactions Help Toddlers Learn Language.” Child Development. https://doi.org/10.1111/cdev.12166
4. Garner, A., Yogman, M., et al. (2021). “Preventing Childhood Toxic Stress: Partnering With Families and Communities to Promote Relational Health.” Pediatrics. https://doi.org/10.1542/peds.2021-052582
5. American Psychological Association. (2025, November). “APA Health Advisory on the Use of Generative AI Chatbots and Wellness Applications for Mental Health.” APA. https://www.apa.org/topics/artificial-intelligence-machine-learning/health-advisory-ai-chatbots-wellness-apps-mental-health.pdf
6. UNICEF. (2026, June). “When AI becomes a friend: Child rights risks, harms, and regulatory responses to AI chatbots and companions.” UNICEF. https://www.unicef.org/documents/when-ai-becomes-friend-child-rights-risks
7. University of Edinburgh. (2024, March 20). “Children may overestimate smart speakers’ abilities, study finds.” University of Edinburgh. https://edinburghuni-newsroom.prgloo.com/news/children-may-overestimate-smart-speakers-abilities-study-finds