REFLECTION

Through this assignment, I learned that AI can be useful in architectural thinking, but only when it is used critically. Our group compared ChatGPT, Gemini, and Microsoft Copilot through three prompt stages: original, refined, and contextual. The topic was climate-responsive architecture in Malaysia, focusing on how design strategies can improve thermal comfort, reduce cooling loads, and support energy efficiency. The workflow showed that the quality of AI responses improved when the prompts became more specific, moving from general hot-climate strategies to university-building solutions and then to Malaysia-specific campus strategies.

At the beginning, the original prompt produced useful but broad answers. The AI systems mostly gave similar strategies, such as shading, natural ventilation, passive cooling, material selection, and energy efficiency. However, because the prompt did not mention Malaysia or a university campus, the answers were not fully contextual. This helped me understand that AI does not automatically know the exact design direction unless the user gives clear context.

The refined prompt gave better results because it asked for practical architectural solutions for a university building in a hot and humid climate. At this stage, the answers became more focused on shading, ventilation, roof design, building materials, and energy efficiency. Gemini performed strongly because it gave more detailed climate-based explanations, while ChatGPT was helpful for clear design wording. Copilot was useful for simple and organized summaries.

The contextual prompt was the most effective because it directly mentioned a university campus in Malaysia. This produced the strongest and most relevant responses. Gemini performed best overall because it showed the strongest balance between contextual accuracy, technical depth, and climate-specific reasoning. ChatGPT was still useful for design communication and architectural language, while Copilot helped organize practical ideas clearly.

This assignment also taught me that AI should not be treated as the final authority. Some responses sounded confident even when they were still too general or needed checking. Human judgement was needed to decide which ideas were realistic, buildable, ethical, and suitable for Malaysia’s climate. As architecture students, we still had to compare, question, refine, and select the best ideas instead of simply copying the AI outputs.

Overall, I learned that AI is strongest when used as a support tool, not as a replacement for architectural thinking. Different AI systems have different strengths: Gemini was better for technical and contextual accuracy, ChatGPT was better for design explanation, and Copilot was useful for structured summaries. The final decision still depended on human interpretation, discussion, and responsibility. This reflects the main lesson of the assignment: AI does not replace architectural thinking, but it can expand reflective possibilities when used carefully.