Mercedes Mateo is the Division Chief of Education at the Inter-American Development Bank (IDB), where she leads a team of specialists and professionals to support the transformation of educational systems in Latin America and the Caribbean. Her work covers different areas of social policy, with an emphasis on inequality. She has coordinated the research, design, execution, and evaluation of innovative projects in education, as well as contributions to institutional reforms, education, childcare policies, soft skills development, female labor participation, and social cohesion. She has a PhD in political science from the University of Louvain in Belgium. From 2002 to 2004, she was a Marie Curie Postdoctoral Fellow at the Robert Schumann Center of the European University Institute.
Recently, in an exclusive interview with K12 Digest, Mercedes shared insights into how her path from academia to the IDB revealed that real leverage in education comes not from producing research alone, but from bridging knowledge, political understanding, and implementation design in the rooms where decisions are made. For teachers shifting from knowledge deliverers to learning designers, Mercedes called for a reconceptualization of the profession built on pedagogical content knowledge, collaborative professional time, and metacognitive capacity — skills no algorithm can replicate. Ultimately, she sees the urgent task as ensuring that AI becomes a tool for amplification, not substitution, and that education systems distribute human capability rather than concentrate it. The following excerpts are taken from the interview.
Hi Mercedes. Your work spans institutional reform, inequality, early childhood, and digital skills across a broad region. What moment in the field made you realize the unique leverage you had to influence education systems?
There wasn’t a single moment. It was more of a process, a journey. I came into this work through academia, genuinely in love with research. I believed that rigorous knowledge was a powerful instrument for change. And it is, but the accumulation of evidence is just the beginning. What my academic work taught me, even before I entered the field, was the importance of politics and institutions, and the political economy of reform. I understood it theoretically. But it was when I joined the IDB that I saw it alive.
Take early childhood development. The data is unambiguous. We know what works, we have decades of rigorous evidence, and yet policy doesn’t automatically follow. Sometimes the bottleneck is genuinely knowledge or data. But far more often it is something else entirely: the architecture of the system, the incentive structures, the trust or lack of it between technical teams and decision-makers, and the political economy of reform. In a context of fiscal constraints, how do you convince a Minister of Finance to prioritize education investments over infrastructure, over debt servicing, over a dozen other legitimate competing demands? You have to find a common language between ministries of finance and education. You have to communicate results in terms that speak to their constraints, not just your convictions. And you have to be willing to work through compromise without losing sight of what actually matters.
That experience crystallized something for me. The real leverage is not in producing better research alone. It is in bringing together knowledge, political understanding, and implementation design, and being present in the room where decisions are made with enough credibility in both the technical and political worlds to bridge them. That rare combination of intellectual depth, operational experience, and institutional fluency is what transforms evidence into change. Once you understand that, the whole nature of the work shifts.
As artificial intelligence reshapes how students access information, what do you believe will define a “new education model” that prepares learners to think, not just search?
The new education model will be defined by a fundamental recalibration of what we consider valuable. We have in front of us a complete reshaping of which skills matter and which don’t. For centuries, educational institutions were custodians of knowledge because it was a scarce resource. Today, any student with connectivity has access to more information than any library in history. What remains irreplaceably scarce, and therefore genuinely valuable, is the capacity to do something useful, ethical, and creative with that information.
But when I talk about fundamental recalibration I am not only referring to the shift from accumulating facts to connecting them. I am also talking about a recalibration in the value we assign to different kinds of education and formation. Let me give you a striking example. Daniela Amodei, President of Anthropic, one of the leading AI companies in the world today, studied music and literature. A humanist, leading the frontier of artificial intelligence. That is not a coincidence or an anomaly. It is a signal. Humanistic education develops something that is fundamentally needed in this moment: the capacity to think across boundaries, to hold ethical complexity, to ask not just how to build something but whether it should be built, and for whom. The leaders shaping AI are not only the ones who can code it. They are the ones who can think about what it means.
So what does this new model actually look like? It starts with what neuroscience tells us clearly: you must build the analog skills first. Critical thinking, the ability to construct and deconstruct an argument, to read a complex text and find its fundamental architecture, to tolerate ambiguity and reason under uncertainty. These are not soft skills. They are the cognitive infrastructure upon which everything else is built. At the same time, we cannot isolate students from digital reality. They need to learn with technology, but more importantly they need to learn about it, understanding its logic, its limits, its embedded values and biases. That combination, analog foundation and digital literacy, is what produces a person who uses AI rather than one who is used by it.
And we must be honest that this transformation is not incremental. It is not about integrating AI tools into existing classrooms and calling it innovation. It requires reimagining the purpose of learning itself. But perhaps the most urgent dimension of this recalibration is equity. It cannot be that children from wealthy families learn about AI while children from poor families simply learn with it. The first group develops agency and judgment. The second risks becoming a passive consumer of intelligence designed elsewhere. That is not a digital divide. It is a human one.
Finally, the new model demands that we change what we evaluate and reward. We have spent generations celebrating the correct answer. We must now develop the capacity to ask better questions, to analyze, to synthesize across disciplines. We should be designing curricula around problems that do not respect disciplinary boundaries, because the most important challenges of our time certainly don’t.
We’ve spent decades closing the digital divide, yet AI tools are now widely available. Why might the next major inequality be cognitive rather than digital, and how would we recognize it in schools today?
The digital divide was always about access. Who had the device, the connection, the platform. And closing that gap was and it is still today necessary and important. But I think we have made a quiet assumption along the way: that access to tools was equivalent to access to opportunity. The AI moment holds that assumption up to a mirror, and what we see reflected is not comfortable.
The inequality I worry about most is not who has access to an AI assistant. It is who has developed the cognitive foundation to use that assistant as a tool for amplification rather than a substitute for thought. And that divide is already forming, not along lines of connectivity, but along lines of educational philosophy and family resources. What we are seeing is that families with greater means are choosing schools that offer rich humanistic formation. Schools that prioritize reading on paper, the arts, music, physical engagement with the natural world. These environments develop the foundational analog capacities that become decisive in an AI-agentic world. Meanwhile, more affordable and scalable educational solutions tend to rely heavily on technology-mediated instruction, which is not without value, but which, if applied without the right foundations, risks producing what I call passive consumers of intelligence rather than active generators of it. We need to make sure that when students are exposed to AI as a pedagogical tool have the cognitive architecture that would allow them to use it well.
And the neuroscience here is sobering and we need to take it seriously. In the first years of life, when the brain is in its most critical period of development, indiscriminate exposure to screens floods the developing nervous system with stimuli it is not yet equipped to process. Neuroscientists describe what follows as accelerated maturation: the brain, overwhelmed by input, skips foundational developmental stages in order to cope. What sounds like precocity is in fact a deficit. Those skipped stages show up later as reduced resilience, lower cognitive flexibility, and diminished capacity for sustained attention, precisely the capacities most needed in a complex world. Jean Twenge’s longitudinal research on adolescents documents the sharp correlation between rising screen time and rising rates of anxiety, depression, and loneliness among teenagers, particularly girls. More recently, Jonathan Haidt’s work in The Anxious Generation synthesizes a broad body of evidence linking smartphone adoption and social media use in early adolescence to deteriorating mental health outcomes across multiple countries and demographic groups.
At the university level, the evidence is equally striking. A study led by Nataliya Kosmyna and colleagues at MIT’s Media Lab, titled Your Brain on ChatGPT, tracked students over four months using EEG technology to measure brain activity while they wrote essays under three conditions: using ChatGPT, using a standard search engine, or working with no tools at all. The AI-assisted group showed the lowest levels of neural engagement, and when tested afterward on retention and comprehension, they remembered less and showed weaker understanding than students who had written independently. A striking 83% of ChatGPT users were unable to quote from the essays they had just written. The researchers coined the term cognitive debt to describe this phenomenon: while AI tools spare the user mental effort in the short term, they generate long-term costs including diminished critical thinking, reduced creativity, increased vulnerability to bias, and shallow information processing.
Perhaps most telling was a detail about subjective experience: the AI-using students reported feeling more confident about their work, even as less actual learning had taken place. This is why the emerging cognitive inequality is so difficult to see and so dangerous to ignore. The smooth, effortless quality of AI-assisted thinking masked the fact that the cognitive work required for genuine understanding had simply not occurred. The brain, like any complex system, develops through use. Capacities that are not exercised do not simply remain dormant. They atrophy.
Teachers are moving from knowledge deliverers to learning designers. In a world where cognitive capacity may become the new equity line, what support do educators need to lead that shift?
If we are serious about recalibrating what students learn, what we consider valuable, and which skills matter, then we need an equally fundamental reconceptualization of the role of the teacher. And that begins with how systems value what teachers do.
In the new education model I described earlier, the goal is no longer the transmission of codified knowledge. It is the development of cognitive infrastructure: critical thinking, the capacity to construct and deconstruct arguments, to tolerate ambiguity, to reason under uncertainty, to ask better questions rather than retrieve stored answers. If that is what we want students to develop, then the teacher can no longer be a deliverer of content. They must become designers of learning environments, an architect of experiences where curiosity, judgment, and synthesis are not occasional activities but daily practice. That requires preparation, support, and recognition.
For too long, teaching has been assessed by coverage: how much of the curriculum was transmitted, and how students performed on standardized recall. That model made sense when knowledge was scarce and the school was its custodian. Now, if we want teachers to lead the shift toward cognitive formation, we have to evaluate and reward them accordingly. That means measuring not just what students know at the end of a unit, but whether they can think independently, construct original arguments, and engage productively with problems that have no single correct answer. And for that, teachers need pedagogical content knowledge, collaborative professional time, and the development of their own metacognitive capacity. To achieve that, we need to fundamentally rethink preservice and in-service training.
Pedagogical content knowledge requires not just mastery of subject matter, but genuine understanding of how learners develop and understanding it. Where are the conceptual sticking points? How do you design a sequence of experiences that moves a student from surface familiarity to genuine intellectual ownership? This kind of knowledge is not intuitive and it is not acquired in a single training session. It requires sustained professional formation and practice.
Collaborative professional time is structurally rare. The most effective school systems in the world treat teacher learning as an ongoing collective practice. Teachers co-design lessons, observe each other in classrooms, and reflect together on what is working and why. What builds teaching quality is structured, disciplined collaboration between practitioners who trust each other enough to be honest.
Teachers’ own metacognitive capacity is the ability to observe and articulate one’s own thinking processes, to notice when reasoning is going well and when it is being shortcut or influenced. It is not just a skill we want students to develop. It is a skill teachers must embody and model. Because the most powerful thing a teacher can demonstrate in an AI-agentic world is not how to find the answer. It is how to reason through genuine uncertainty, how to recognize when one’s own judgment is being shaped by unconscious biases, and how to maintain intellectual agency in a world specifically designed to shortcut it. That capacity, metacognitive awareness as a lived practice and a condition of freedom, is what great teachers will transmit to their students. And it is something no algorithm can replicate or replace.
Books shape how leaders think about power, policy, and people. What book on development, politics, or learning has the most notes in your margins, and what idea from it shows up in your work?
Two intellectual obsessions have genuinely shaped how I think about education and development, and I find them inseparable.
The first is the question of why some countries create wealth, and others don’t, and more specifically, why some nations that started with very little became engines of prosperity, while others richly endowed with natural resources remained trapped in low growth and inequality. I remember being in Gothenburg during part of my PhD program, deeply puzzled by the Nordic countries. I asked a professor why Sweden had become so equal, and his answer stopped me: we were poor. We had no natural resources. When you have a lot to extract, some people own the resources and the rest work. Here, we had to do something else. That answer sent me into the literature that traces the institutional and political foundations of prosperity. Hirschman’s Exit, Voice, and Loyalty gave me a framework for understanding how individuals and institutions respond to decline and opportunity. Acemoglu and Robinson’s Why Nations Fail extended that into a systematic argument about inclusive versus extractive institutions. The idea that what separates prosperous from failing states is not geography or culture, but whether political and economic institutions distribute power broadly or concentrate it. Ray Dalio’s The Changing World Order adds the temporal dimension, showing how even great empires follow cycles of rise and decline, often determined by the same institutional dynamics. These books are present in how I think about education systems. They are not just about schooling. They are about whether a society is investing in the distributed development of human capability or extracting value from the many to concentrate it among the few.
The second obsession then follows naturally from it: distribution. Not just how wealth is created, but who benefits from it. And here the most influential text for me has been John Rawls’s A Theory of Justice. The concept of the original position and the veil of ignorance is one of the most powerful thought experiments in the history of political philosophy. If you did not know which position you would occupy in a society, rich or poor, majority or minority, born in a capital city or a remote rural community, how would you design its institutions? Rawls argues that behind that veil, rational people would choose a society that protects basic liberties for all and arranges inequalities only to the benefit of the least advantaged. That framework lives in my work every day. When I think about where to direct investment in education, whose learning outcomes we measure and whose we ignore, which children have access to the kind of formation that unlocks human potential, I am always, somewhere in the background, behind that veil. The question is not what is most efficient on average. It is what is just for the child who had no say in where she was born.
Rest fuels sustainable leadership. What hobby or ritual helps you return to work with fresh energy for policy, partners, and teams?
I believe the answer to sustaining energy rests in a combination between the self and the others. Resilience lives in the tension between two needs that can feel contradictory: the need for stillness and reflection and the need for belonging and connection. To store energy and resource, one needs time to be alone without feeling abandoned, and time to be with others without feeling lonely.
The first part therefore is time with myself. Not idle time, but reflective time. I talked earlier about metacognition as something we need to cultivate in teachers, the capacity to observe one’s own thinking, to notice when reasoning is going well and when it is being distorted by fatigue, pressure, or bias. That practice is not only pedagogical. It is deeply personal. Leadership without self-reflection is leadership on autopilot, and autopilot in complex, high-stakes environments eventually leads somewhere you didn’t intend to go. The moments I carve out for myself, whether in the early morning, in movement, or simply in silence, are not indulgences. They are the conditions under which I can think clearly, reset my judgment, and return to work with genuine presence rather than just physical attendance.
The second part is the people. I feel very fortunate, and sometimes we don’t say that enough. My children are my greatest source of strength. They challenge me every single day in ways that no professional experience ever could. They demand authenticity. They notice inconsistency. They remind me, constantly, of what actually matters and of the world they will inherit from decisions being made right now. Being their mother has made me a better leader in ways I could never have anticipated.
My parents and friends are another irreplaceable source of strength. Through good times and bad, they hold the fort. They tell me the hard truths, but they also know how to celebrate life. That combination of honest counsel and shared joy is rare, and I do not take it for granted.
At work, I have had the privilege of building relationships with people who read me and I can read. People with whom I can genuinely think through hard problems rather than simply get things done. There is a profound difference between those two things. A team that simply gets things done collectively is exhausting. A team built on deep professional trust, mutual inspiration, where you can think out loud, be wrong, change your mind, and still be taken seriously, is generative. That kind of trust does not happen automatically. It is built over time, through consistency and through the courage to be honest even when it costs something.
That network, family, close friends, and colleagues with whom you feel genuinely safe, is not a peripheral benefit of a good life. It is the infrastructure that sustains life itself and leadership. Without it, you can perform for a while. You can push through on willpower and discipline and a sense of duty. But you cannot endure. Reflection and connection can be sacrificed when urgency takes over. And in this kind of work, urgency is always present. Protecting the conditions that make us capable of sustained, high-quality work as individuals and teams is not selfishness. It is responsibility.
From Ph.D. to Chief, you’ve built technical and political capital. What advice would you give a researcher who wants to move into implementation?
The journey from academia to public policy is fascinating. Development is one of the “messiest jobs”, to borrow Luis Garicano’s concept. Not the clean, well-defined, single-task role. Rather the kind of work that requires you to navigate ambiguity, combine multiple disciplines, read people and contexts, and make judgment calls that no algorithm can make for you.
The first thing I would say to that researcher is that the skills that make you excellent in academia are a great asset for policy. Analytical skills, the ability to hold complexity, and intellectual honesty about uncertainty: these are all enormously valuable qualities. But they need to be translated. Not simplified. Translated. Policymakers are operating under fundamentally different constraints, with different incentive structures, shorter time horizons, and the weight of political accountability that no academic paper ever carries. Your job is not to water down your ideas. It is to learn to communicate their essential logic in a language that speaks to the decisions actually being made. That requires a kind of situational intelligence that no PhD program teaches explicitly, but that you need to develop deliberately.
The second thing I would say is develop your political intelligence without sacrificing your intellectual integrity. Understanding power, who holds it, what they want to protect, what they might be willing to risk, is not cynicism. It is the condition of effectiveness. The researchers who move successfully into policy are those who learn to navigate institutional complexity without losing clarity about what actually matters and why. Walking into a Ministry of Finance with a stack of evidence is not enough. You have to understand what that minister is trying to protect, and the pressures they are navigating. Change happens when technical credibility and political fluency meet in the same person or the same team. Building both, without letting either crowd out the other is the real work of change management.
Third: take implementation seriously as an intellectual discipline. There is sometimes an implicit hierarchy in academic culture that treats ideas as noble and execution as administrative. That hierarchy is wrong and costly. The distance between a well-designed policy and a well-implemented one is where most reforms go to die. Understanding how institutions actually work, how incentives shape behavior at every level of a system, how trust is built between technical teams and decision-makers, how you sequence reforms in a context of fiscal constraints: these are genuinely complex intellectual problems. Treat them that way.
And finally, find your people. Leadership in implementation can be demanding in ways that research rarely is, because the feedback loops are slower, the variables are harder to control, and the stakes are human and immediate. What sustains you through that is not a hobby or a ritual. It is the people around you, at work and at home, whom you genuinely trust, admire and can think alongside. That network is not a peripheral benefit. It is the infrastructure that makes everything else possible over the long run. As messy as this work is, as demanding and unpredictable and humbling as it can be, I would not trade it for anything.