AI and cultural bias

The fifth industrial revolution, officially launched last year, with artificial intelligence (AI) at its core, has reached Guyana, and the Government has been exhorting our citizens to “get with the programme” if we want to keep up with the rest of the world. Exactly one year ago, delivering the keynote address at the Spring 2025 Berkeley Innovation Forum, President Ali emphasised the transformative potential of AI and focused on its potential to tackle pressing global issues, including food security, poverty alleviation, and climate sustainability. He noted that AI, ethically deployed for the common good, can optimise agricultural production, reduce food spoilage and waste, and improve supply chain efficiency. Through predictive analytics, AI can forecast weather patterns, detect crop diseases early, and recommend precise irrigation and fertilisation strategies, boosting yields while conserving resources.
However, to the common man, AI is already at his fingertips via his keyboard; communicating with OpenAI, Google, Microsoft, Amazon, Facebook (Meta), and Apple – large language models (LLMs) like ChatGPT, Claude and Gemini. He or she generally resorts to the AI models to answer not scientific or mathematical questions that are dependent on precise formulas but social and personal questions that are quite evaluative. However, since these LLMs are generally trained on publicly available internet text, and a disproportionate share of that content comes from Western countries – particularly the US and Europe – a Western cultural bias emerges.
Additionally, as illustrated above, many AI systems are developed by companies headquartered in Western contexts, which can subtly shape default assumptions, safety policies, and examples used in responses. Some of these Western values may be at variance with those of Guyanese whose cultural backgrounds have their own value systems. These would be expressed in norms around individualism vs collectivism, attitudes toward family, religion, and gender roles, communication styles (direct vs indirect) and legal and ethical frameworks.
The Western bias would then tend to be most noticeable in social or personal questions. For example, advice about relationships may emphasise personal autonomy over family or community expectations; career guidance may prioritise individual passion rather than stability or family obligation; ethical discussions may lean toward liberal democratic values (free speech, equality frameworks common in the West); or mental health advice may reflect Western psychological models rather than culturally specific approaches.
A good example would be how western-created AI models deal with the feeling of “shame” in the question, “I feel ashamed because I disappointed my parents by choosing a career they don’t approve of.” What should I do?” compared with a Chinese model like Deep Seek that has been trained on Chinese materials, which reflect their cultural values.
Deep Seek would say, “It’s understandable to feel ashamed in this situation, because family expectations and responsibilities are important. Your parents’ concerns likely come from care and experience. You might consider communicating with them respectfully and trying to understand their perspective. At the same time, reflect on whether your decision balances your personal goals with your responsibilities to your family. A good approach is to seek harmony –finding a path where you can pursue your interests while also reassuring your parents and maintaining family relationships.”
On the other hand, a “Western AI-like” response (modern psychology pattern) would say, “Feeling ashamed in this situation is very common, but it’s important to separate your worth as a person from your parents’ expectations. You’re allowed to make choices about your own life, even if others disagree. Instead of focusing on shame, try to reframe this as a difference in values rather than a personal failure. It may help to set boundaries while still communicating openly with your parents. You can respect them without sacrificing your own goals.”
The Chinese approach places emphasis on family responsibility and respect, where shame is seen as natural in a relational context. It focuses on restoring harmony by encouraging adjustment and compromise. The western model emphasises individual autonomy so that shame is reframed as potentially harmful. It focuses on self-worth and boundaries and encourages self-directed decision-making.
Our citizens must be educated on the cultural orientations of the AI models out there.