No, we cannot code a single, universally correct morality into AI. We can encode values and ethics through objectives, rules, training data and human feedback, but these choices reflect particular people, institutions and cultural contexts.
In IB Digital Society, values and ethics distinguish between right and wrong, fair and unfair, and just and unjust. They guide decisions through frameworks, codes, policies and laws. This concept is shared by SL and HL.
Translating broad principles such as fairness into instructions creates limits:
| Method | Ethical limitation |
|---|---|
| Fixed rules | Rules cannot anticipate every context or conflict between principles. |
| Training data | Historical data may reproduce prejudice and unequal power. |
| Human feedback | Reviewers may disagree and bring cultural values. |
| Optimization objectives | A measurable target may oversimplify complex harm. |
System: Anthropic's Constitutional AI approach trains language models using written principles and AI-generated feedback. It combines artificial intelligence, algorithms and training data.
Example specifics: Research published in 2022 describes a stated constitution that guides the model's criticism and revision of responses. This example postdates the 2022 Digital Society coursebook. Developers still select the constitution, examples and training process, so human choices remain embedded in the system.
Impacts and implications: The approach is intended to produce safer responses and potentially reduce harmful content for users. However, it raises questions about whose values receive priority, whether decisions are transparent and how affected communities can challenge them. Developers, users, regulators and communities may hold competing perspectives.
The common misconception is that mathematical instructions remove human values. In reality, selecting data, objectives and rules redistributes power and can create intended and unintended consequences. Ethics can shape an AI system, but it cannot become a neutral technical property.
In an IB exam, distinguish what can be coded from what must remain contested. For an evaluate or discuss response, compare stakeholder perspectives, weigh strengths against limitations and reach a supported, balanced conclusion rather than claiming there is one correct ethical standard.
