Good for Whom? The Question We Keep Avoiding About AI in Education
Every conversation about AI in education eventually arrives at the same tired debate.
Is AI good for learning? Is it bad? Should we embrace it? Should we ban it? Are we helping students or harming them?
I’ve grown weary of this binary. Not because the questions don’t matter — they do. But because they’re the wrong questions.
“Good” and “bad” are abstractions. They float above the messy reality of classrooms, schools, and communities. They let us argue endlessly without ever examining the specifics.
The better question — the one that actually gets somewhere — is this: Good for whom? And at what cost?
Once you ask that, everything shifts. You can’t answer with vague pronouncements about technology’s promise or peril. You have to get specific. You have to name who benefits, who pays, and what’s being traded away.
And when you do that, the conversation becomes both harder and more honest.
The Cost We Don’t Talk About
Let me start with something that rarely comes up in edtech discussions: the environmental cost.
AI systems require enormous amounts of energy. Training a single large language model can consume as much electricity as hundreds of homes use in a year. Every query you make — every prompt, every generated response — requires server capacity, cooling systems, and infrastructure that strains power grids and water supplies.
Data centers are being built at unprecedented rates, often in communities that don’t benefit from them. They use water for cooling in areas already facing drought. They consume electricity that could power homes and hospitals.
When we celebrate AI’s potential in education, are we accounting for these costs? Are we asking whether the benefits justify the environmental burden?
I haven’t seen much evidence that we are. We treat AI as if it’s weightless — as if it exists in some abstract digital realm rather than in physical machines that consume real resources.
But students are growing up in a world where these costs will be felt. Climate change is not an abstraction for them. It’s their future. And we’re modeling a relationship with technology that ignores its ecological footprint.
That’s an ethical problem. And it’s one we’re not teaching students to see.
Who Benefits?
Every technology serves someone. The question is who.
When schools adopt AI tools, who profits? Sometimes it’s companies offering free tiers with the expectation of future revenue. Sometimes it’s publishers bundling AI features into existing products. Sometimes it’s platforms collecting student data to train their models.
When we talk about AI “helping teachers,” whose version of help are we assuming?
- Is it reducing workload — or is it standardizing practice in ways that deskill the profession?
- Is it freeing up time for relationships — or is it replacing human judgment with algorithmic efficiency?
When we talk about AI “personalizing learning,” who benefits most?
- Students who need challenge and enrichment? Or institutions looking to reduce staffing costs?
- Students with resources to access premium tools? Or platforms collecting behavioral data for commercial purposes?
These aren’t cynical questions. They’re necessary ones.
I’ve seen genuinely inspiring uses of AI in education — teachers building tools to solve real problems, students creating apps that help their communities, researchers developing approaches that respect teacher expertise. But I’ve also seen products that promise transformation while delivering surveillance. I’ve seen “personalization” that means standardization at scale. I’ve seen “efficiency” that means cutting corners on the things that matter most.
The difference isn’t the technology. It’s the values embedded in how it’s used — and who gets to decide.
The Bias We Import
There’s a dimension to AI ethics that hits education particularly hard: algorithmic bias.
AI systems are trained on data. That data reflects the world as it is — including all its historical injustices, structural inequalities, and unexamined assumptions.
When facial recognition works better on white faces than Black ones, that’s not a technical glitch. It’s a reflection of training data that underrepresents certain populations. When language models associate certain professions with certain genders, that’s not an accident. It’s the result of patterns in the text they’ve absorbed.
When AI is used to assess student writing, predict student success, or recommend student pathways, these biases don’t disappear. They get embedded in decisions that shape young people’s futures.
I’ve written before about how a single word change — classical music versus rap music — produces dramatically different outputs from AI. That’s not a bug. It’s a window into how these systems replicate and amplify existing social hierarchies.
And here’s the thing: students know. They experience bias in their lives long before they encounter it in AI. What they need is the language and framework to name it, understand it, and challenge it.
That’s not just AI literacy. It’s civic education. It’s preparing young people to hold power accountable — including the algorithmic power that increasingly shapes their opportunities.
The Surveillance Question
There’s another cost that’s rarely named: surveillance.
AI in education isn’t just about tools that help students learn. It’s also about tools that monitor, track, and analyze student behavior:
- Eye-tracking to measure attention.
- Keystroke logging to detect cheating.
- Emotion recognition to assess engagement.
- Predictive analytics to flag “at-risk” students.
Each of these raises profound ethical questions. Do students consent to this monitoring? Do they understand what’s being collected and how it’s used? Do parents? Do teachers?
The researcher I mentioned in an earlier piece talked about AI used in different countries to track eye movements — to improve teaching or to surveil students. The difference isn’t technical. It’s philosophical. What do we believe education is for? What do we believe students are?
If education is about developing autonomous, critical thinkers, then surveillance is counterproductive. It teaches compliance, not agency. It creates an environment of suspicion, not trust.
If education is about sorting and selecting — identifying “winners” and “losers” — then surveillance is useful. It generates data that can be used to justify existing hierarchies.
The technology doesn’t decide. We do. But we rarely have that conversation explicitly.
The Access Divide
Let me return to something I raised in the last piece: access.
Get Chalkandchat’s stories in your inbox
Join Medium for free to get updates from this writer.
AI literacy for all is a beautiful aspiration. But access to AI tools is wildly unequal.
- Some students have personal devices, reliable internet, and parents who can afford premium subscriptions. Others share a single phone with siblings, rely on school equipment, or live in areas without stable connectivity.
- Some schools have teachers with time and training to integrate AI meaningfully. Others have overwhelmed staff who can barely cover the basics.
- Some communities have the political power to demand equitable resources. Others are systematically neglected.
When we talk about AI in education without addressing these disparities, we’re not just ignoring reality — we’re actively widening the gap. We’re designing a future where some students get to shape AI and others only get to be shaped by it.
That’s an ethical failure. And it’s one that affects every country, though the specifics differ.
- In South Africa, it’s the bimodal education system — a small elite of well-resourced schools and a vast majority of underfunded ones.
- In Kenya, it’s the gap between urban and rural infrastructure.
- In Greece, it’s the challenge of reaching remote mountain communities.
- In the United States, it’s the persistent inequities between wealthy suburbs and underfunded urban and rural districts.
The technology isn’t the cause. But it can be an accelerant — making existing inequalities worse, faster.
The Labor Question
There’s one more cost I want to name: the labor implications.
When we talk about AI “supporting teachers,” we often gloss over the fact that technology has a history of being used to deskill, deprofessionalize, and displace workers.
Teachers are already facing unsustainable workloads. They’re being asked to do more with less, to implement new initiatives without adequate support, to navigate political attacks on their profession.
AI could genuinely help — reducing administrative burden, freeing time for relationships, supporting planning and assessment. Or it could be used to justify larger class sizes, reduce staffing, and replace human judgment with algorithmic efficiency.
The difference depends on who controls the technology and what values guide its implementation.
I’ve heard from teachers who feel empowered by AI tools — they’ve built their own apps, customized workflows, reclaimed time for what matters. I’ve also heard from teachers who feel threatened — their expertise dismissed, their autonomy undermined, their concerns about bias and surveillance ignored.
Both experiences are real. Both matter. And the difference between them is not the technology itself. It’s the power dynamics surrounding its use.
What Ethical AI Education Looks Like
So what would it mean to take ethics seriously in AI education?
It would mean teaching students to ask “good for whom?” about every technology they encounter. Not just “Does this work?” but “Who benefits? Who pays? What’s the environmental cost? What biases are embedded? Who has access? Who’s excluded?”
It would mean centering the voices of those most affected by AI — not as an afterthought, but as a starting point. Students from marginalized communities. Teachers in under-resourced schools. Workers whose jobs are being transformed. Communities whose data is being extracted.
It would mean being honest about the trade-offs. AI in education isn’t a simple good. It offers real benefits and imposes real costs. Honest education acknowledges both.
It would mean modeling ethical decision-making ourselves. Not just talking about bias, but examining our own. Not just celebrating innovation, but questioning its consequences. Not just using AI, but asking whether we should.
And it would mean recognizing that ethics isn’t a separate subject. It’s not something you teach in PSHE or citizenship and then set aside. It’s woven through every decision, every lesson, every interaction.
The Question That Changes Everything
I keep coming back to “good for whom?” because it’s the question that refuses easy answers.
- Is AI good for students? For which students? In which contexts? With which support? At what cost?
- Is AI good for teachers? For which teachers? With what training? Under what conditions? With what safeguards?
- Is AI good for education? For what version of education? In service of what goals? Reflecting what values?
These questions don’t lead to neat conclusions. They lead to more questions. To complexity. To uncertainty.
But that’s exactly where we should be. Not in the comfortable binary of pro- and anti-tech, but in the messy, human work of making judgments in specific contexts with real stakes.
That’s what critical thinking looks like in practice. That’s what AI literacy means at its deepest level. That’s what education should be preparing students to do.
A Personal Reckoning
I’ve been trying to apply this to my own practice. Not just writing about ethics, but living it.
When I use AI, I try to ask: Who benefits from this interaction? What data am I generating? What labor was involved in creating this tool? What environmental cost am I incurring? Is there a less extractive way to accomplish this?
I don’t always have good answers. I don’t always make the most ethical choice. But I’m trying to be aware. I’m trying to notice.
That’s the starting point. Not perfection. Awareness.
And I’m trying to bring that awareness into conversations with students. Not preaching, but modeling. Not telling them what to think, but showing them how to ask better questions.
Because that’s what education is for. Not producing compliant consumers or efficient workers. Producing citizens — people who can see the systems shaping their lives and make conscious choices about their place within them.
AI is one of those systems. And it’s shaping our lives more every day.
What I’m Hoping For
I hope we move beyond “is AI good or bad?” to “good for whom, and at what cost?”
I hope we center ethics — not as an add-on, but as the foundation.
I hope we hold tech companies and policymakers accountable for the consequences of their choices.
I hope we listen to the communities most affected by AI, not just the ones with the loudest voices or the biggest budgets.
I hope we recognize that access is an ethical issue, not just a logistical one.
And I hope we remember that the future is not determined. It’s chosen. And choices have consequences — for some more than others.
Discussion Question
Good for whom?
That’s the question I’d invite you to sit with. In your own practice. In your own school. In your own community.
Who benefits from the AI decisions being made? Who pays the costs? Who is included in the conversation? Who is excluded?
The conversation continues in the comments.
Next in this series: What does it mean to be human in an age of AI? Reflecting on embodiment, mortality, and the things that can’t be automated.
Subscribe so you don’t miss it.
