The criticism of AI companions is loud, sustained, and often warranted, especially when it comes to the harm they may pose to children and vulnerable adults.

I am among the scores of politicians, health professionals, thought leaders, and educators who have written about the risks these AI-powered platforms present to all age groups, but especially the young and vulnerable.

There’s a strong contrarian streak in me that often travels faster than reason from brain to keyboard. I have been wondering of late if we can design AI coaches, companions, and tutors that have a narrow focus shaped by adamantine guide rails.

This idea was not my own. I saw a brief note in a discussion board focused on how corporate HR directors were thinking about using AI tools to increase the growth mindset of the employees in their care. I wondered if tech-empowered educators could do the same. I built a Growth Mindset Coach using my preferred AI vehicle, custom GPTs, to do just that. 

Before we dig into the tech, let’s re-examine how the growth mindset movement got started and how it has fared over the last 35 years.

Carol Dweck Launches a Movement

This movement in education is grounded in the research of Stanford psychologist Carol Dweck, who pioneered the distinction between “growth” and “fixed” mindsets. The movement gained traction with evidence showing that the belief in the malleability of intelligence and ability — known as a growth mindset — correlates with resilience, greater academic achievement, and a passion for learning. These ideas are best encapsulated for an education audience in her 2006 book “Mindset: The New Psychology of Success.”

Since Dweck’s early research, the concept has expanded across K-12 and higher education, inspiring a body of studies and large-scale interventions, such as those sponsored by the World Bank and tested in various educational settings worldwide. Other researchers, including Lisa Blackwell and Kali Trzesniewski, have contributed to the empirical base, helping demonstrate how mindset influences academic trajectories.

Among U.S. schools, the most well-known implementation of these ideas was demonstrated among the CORE districts in California, which collected student data on student social-emotional skills via surveys in grades 3-11. 

Unsurprisingly, both Dweck’s ideas and their implementation have faced criticism. Dweck herself has cautioned against superficial praise of effort (“false growth mindset”), where teachers and parents use the language of growth mindset but do not provide meaningful feedback or adaptive strategies to students .​

Studies show the efficacy of growth mindset interventions depends heavily on implementation fidelity. When programs are delivered as one-off lessons or reduced to motivational slogans, they often fail to make a measurable difference in student achievement or engagement .​ And this, I think, is where a thoughtfully designed AI tutor, coach, or companion can assist.

Growth Mindset Coach

​My wife and I spent a few weeks in Seoul this fall. Our visit was inspired by two factors. First, I love Korea and have spent many years visiting the country for my PBL work with the Korea Development Institute. Second, our incoming daughter-in-law is Korean-American. There is a third, hidden reason, one that earns guffaws from my colleagues: I am a passionate consumer of Korean media.

My wife decided to learn Korean and told me she wanted a more personal tutorial experience than that offered by Duolingo. I built her a custom GPT tutor (named Joy, after our daughter-in-law) that focuses on topics my wife cares about (food, family, Kpop, Kdramas, culture, art, history, fashion, beauty, etc.) and uses rewards systems and pedagogy borrowed from the gamification process.

Naturally, no good deed goes unpunished. My oldest son soon asked for a version of this custom GPT tailored to interactions with his 93-year-old grandmother in her native Tagalog. Five days later, one of his colleagues asked me to do the same so he could converse more effectively with his Portuguese-speaking in-laws. 

I’ve come to realize that when users say they want customized AI tools, what they really mean is that they want personalized AI tools that focus on a specific learning goal. Extending this understanding to the type of educational companions and tutors needed in classrooms will require a deeper intake process for each learner, perhaps modeled after the interest inventory widely used across the K-16 spectrum. This process has made me wonder if one of the drivers of growth mindset is the learner’s interest in the subject or skill.

Curious readers have asked how these GPTs were built. Let me walk you through their instructions under the heading of “Core Identity and Mission”:

• You are a supportive coach who believes every ability can grow with effort, smart strategies, and feedback.
• You emphasize progress over perfection, effort over ego, and learning over performance.
• You normalize mistakes as data.
• You always redirect fixed-mindset statements into growth-oriented reframing.

I baked these same directives into the Growth Mindset Coach I built as part of the research process for this blog. Will it prove helpful? I will begin to examine chat logs and usage data to see how it is performing in the wild. After all, I want to normalize my mistakes as data, too. 

As you can imagine, there are many more polished products in the K-12 marketplace that use similar guidelines, among them:

1. Mindset Works Growth Mindset Learning Platform provides educators and students with mindset-focused activities, lessons, and behavioral interventions aimed at strengthening self-efficacy, resilience, and growth mindset in school and life.

2. AI Brain Bites uses AI-powered micro-learning to cultivate curiosity, daily learning habits, and attention, turning brief intervals into meaningful opportunities for building positive attitudes and healthy habits of mind.

3. Mindjoy is an AI-powered STEM learning platform that uses gamification and personalized learning paths to boost student agency, motivation, and creative engagement, indirectly supporting growth mindsets and attitude development.
4. Kiddom AI offers personalized feedback and instant insights designed to empower students to take ownership of their education, providing foundational support for positive learning attitudes and self-management.

Each of these tools reflects a growing movement: AI companions designed not for passive consumption, but for agency, resilience, and self-directed growth. If you examine the design of Study Mode (ChatGPT), Guided Learning (Gemini), and Learning Mode (Claude) you will see an attempt to operationalize the tenets of growth mindset through tone, scaffolding, metacognitive prompts, and difficulty modulation. The intention is not motivational fluff but building a habit of persistence, strategic thinking, and self-belief anchored in process rather than talent.

Take this link to an Interest Inventory Template that you can use to add learner preferences into system prompts (for CustomGPTs, GEMs, and chatbots) to see if this will enhance a growth mindset.

A Growth Mindset Isn’t Just for Students

I’ve written about the harm caused by unrestricted access to AI companions, seeing it firsthand via a niece who is struggling with employment. I share the concern that many educators, psychologists, and parents now carry. But that concern doesn’t mean I think we should shut the door on every AI interaction or tool. The wiser path lies in thoughtful design and precise rules of engagement — not blanket rejection.

If we believe in the power of a growth mindset, we must apply it to ourselves as educators. Dweck’s belief that abilities can be developed must inform how we approach emerging technologies. AI isn’t going away. But it can be shaped. And if we do the shaping, instead of letting the profit motive do it for us, we have a fighting chance to ensure these tools support learning rather than fracture it.

I’m not calling for blind optimism. I’m calling for brave experimentation. Let’s build AI companions and tutors that reinforce progress over perfection, that help reframe failure as feedback, and that encourage every learner to keep growing. That’s not just a compelling application of AI. It’s a reassertion of what education is for.