The Western Filter: When Your AI Study Buddy Embeds Your Own Cultural Biases

My days in the IBDP Mathematics classroom are all about guiding, grinding, and practice. My evenings, however, are a different story. They are currently spent building wonderfully chaotic worlds with my three-year-old, worlds that—thanks to basic generative AI—are perfect allegories for the classroom.

The other evening, we were creating a simple bedtime story using a custom Gem. The main character, a "Snackler," was described with a magnificently ridiculous feature set, including 20 trunks. But as the AI generated the second image, those 20 trunks morphed into a single elephant-like proboscis.

My son pointed at the screen and asked: "Why do the pictures show different versions of the Snackler? He doesn't look the same in the second one."

That simple question is the ultimate challenge posed by agentic AI. If an AI struggles with basic consistency in a fictional story, how much harder is it to spot the subtle, but costly, inconsistencies in a complex mathematical model? This childhood curiosity—this demand for accountability—is the very thing we must foster in our students.

The issue isn't the AI's power; it's the bias we code into it.

The Global-Local Conflict: Our Western Data DNA

Before we empower our students to use these tools, we must first confront the uncomfortable truth about our own teaching materials. I’ve been wrestling with this as part of my Masters research into globalisation and glocalisation.

The mathematical models are universal truths, but the teaching of the IBDP in an international setting is often a form of policy borrowing. The cultural context we apply to it—the specific examples, methods, and pedagogies we privilege—is the Western transplant. I’ve seen first-hand the challenges in places like the UAE and Singapore when a Western-centric curriculum is rigidly applied.

When I look at my own bank of IBDP notes—curated over years in the UAE, UK, Singapore and Hong Kong—I realise my teaching is steeped in this bias. This is my Data DNA. My notes lean heavily on Western examples of loans, annuities, and specific UK-favoured mathematical notations. Critically, this inherited curriculum often reinforces the deeply ingrained notion that there is a singular correct method for solving a problem.

The Agent Architect: Automating Our Biases

The best use of agentic AI is as a proactive, out-of-class personalised tutor. I’ve created a custom Gem trained exclusively on my notes and past examples for Arithmetic and Geometric sequences and series, contextualised.

The custom AI now operates as the ultimate Western-centric study buddy, diligently reproducing my methodological and contextual preferences. A student working on homework might prompt it with a complex problem:

"Help me model a personal retirement savings plan using a geometric series, and suggest two potential pitfalls based on my teacher's previous examples."

The AI's answer will not only be mathematically sound but also perfectly biased. It will favour the financial product examples I've always used, and it will likely lean toward the precise algebraic methods I've historically preferred.

The student is now ready for the most critical part of their education: interrogation. The teacher’s new role is to teach the students to critique the AI’s source code (our bias) and output using the essential compass of the ABCE framework.

The Interrogation: Dismantling Our Own Biases with ABCE

We must make the critique of the tool the main learning objective, turning passive users into active IB Inquirers.

1. Accuracy

• The Question: "Is the AI’s formula correct, and is the real-world data it used (e.g., interest rates) sourced reliably?"

• The Interrogation: We demand intellectual honesty, refusing to let the AI's authoritative voice bypass basic mathematical rigour.

2. Bias

• The Question: "Why did the AI suggest that specific solution method and that specific financial scenario? Did its training data (my UK/HK-centric notes) lead it to favour a UK annuity model over a more relevant local investment? Did it prioritise an algebraic proof over a simpler, visual proof?".

• The Interrogation: This is the core lesson. Students must infer the AI's Bias—both contextual and methodological—and challenge the dominant narrative. They must actively seek out methods and examples that were excluded by the curriculum's Data DNA.

3. Completeness

• The Question: "Does the AI’s model account for all the real-world factors, such as inflation, hidden fees, or variable interest rates? What essential context is missing from this mathematically clean answer?".

• The Interrogation: By questioning Completeness, students develop metacognitive skills, learning that the AI's output is limited by its parameters, not a definitive truth.

4. Ethics

• The Question: "What are the ethical implications of using an AI to generate this advice? Am I relying on this tool in a way that risks academic dishonesty, and does the model's bias unfairly privilege one group over another?".

• The Interrogation: We move beyond functionality to morality, ensuring the use of the AI promotes learning and ethical responsibility.

The Full Circle: From Snacklers to Solutions

We started with a child observing a simple inconsistency: a creature with 20 trunks should not suddenly morph into a one-trunked elephant. We end with our IBDP students doing the same for a complex financial model—pointing out where the model is inconsistent, incomplete, or biased by the teacher's own Data DNA.

The teacher’s new role is not to be replaced by the personalised tutor, but to be the coach of the interrogation. The AI is a tool of empowerment, but only if we teach our students to point out and dismantle the flaws we ourselves built into it.

The disruptive energy of AI is here. Let's ensure our students are not just using the tools, but actively challenging them.

Have you used a custom agent in your classroom? What were the unseen methodological biases you discovered in your own Data DNA?

For more on analytical thinking, questioning skills, and how we develop Inquirers in the age of AI, check out my previous posts at www.edtechequation.com.